[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_611":612,"_613":614},"routes/_app.x.$handle.$name",{"_103":102,"_32":46,"_112":91,"_27":113,"_13":14,"_114":-5,"_11":115,"_116":117,"_118":119,"_120":121,"_122":123,"_20":-5,"_124":-5,"_125":126,"_127":25,"_128":129,"_130":60,"_131":-5,"_132":-5,"_133":134},"routes/_app.x.$handle.$name.branch.$branch.version.$version.diff",{"_11":12,"_13":14,"_15":16,"_17":-5,"_18":-5},"project",{"_44":45,"_32":46,"_39":47,"_48":-5,"_49":25,"_50":51,"_52":51,"_53":54,"_55":-5,"_18":56,"_57":58,"_59":60,"_61":-5,"_62":-5,"_63":64,"_65":66,"_67":68,"_69":70,"_71":72,"_73":-5,"_74":75,"_76":75,"_77":78,"_79":78,"_80":78,"_81":75,"_82":75,"_83":84,"_85":86,"_87":88,"_89":25},"version",2,"changes",[19],"aiChatInfo","userId",{"_20":21,"_22":23,"_24":25,"_26":25,"_27":28,"_29":28},"file",{"_30":31,"_32":33,"_34":35,"_36":38,"_39":40,"_41":14,"_42":43,"_24":25},"previous",{"_30":31,"_32":33,"_34":35,"_36":37},"deleted",false,"deletedHere","path","frontend/Forecast.tsx","latestPath","fileId","01a033f6-8ab0-7225-822e-d3bfbaed3b11","name","Forecast.tsx","fileType","script","content","/** @jsxImportSource https://esm.sh/react@18.2.0 */\n// Onglet Forecast — proposition de prévision à 6 mois, référence par référence.\n//\n// L'outil ne remplace pas le jugement commercial : il produit une proposition chiffrée,\n// affiche ce sur quoi elle repose, et laisse corriger chaque ligne. La colonne finale\n// exportée est TOUJOURS celle que vous avez validée, jamais la proposition brute.\nimport { useEffect, useMemo, useState } from \"https://esm.sh/react@18.2.0\";\nimport * as XLSX from \"https://esm.sh/xlsx@0.18.5\";\nimport {\n classifyDemand, DEFAULT_FORECAST_PARAMS, findAnalogues, forecastNewProduct,\n forecastQuality, forecastSku, lifeStage, reliabilityOf, reliabilityReason,\n type ForecastParams, type LifeStage,\n} from \"/frontend/forecasting.ts\";\nimport {\n DEFAULT_CUTOVER as CUTOVER, displayUser, fmtInt, fmtEur, GROUPS, type Product,\n} from \"/frontend/helpers.ts\";\n\nconst NAVY = \"#1F3864\";\nconst ACCENT = \"#4F46E5\";\nconst HORIZON = 6;\n\ninterface DpSku {\n code: number; name: string; brand: string; pl: string; pipo: string;\n top80: boolean; eta: string | null;\n hist: number[]; fc: number[]; fcm1: number[]; avail: number[];\n}\ninterface DpPayload {\n source_file: string; generated_at: string; months: string[];\n skus: DpSku[];\n outside: { code: number; months: string[]; hist: number[] }[];\n /** Date de dernière facturation dans l'export BI — le mois qui la contient est partiel. */\n bi_asof: string | null;\n bi_months: string[];\n}\n\n/** Ordre des niveaux de fiabilité, pour que le tri suive l'échelle et non l'alphabet. */\nconst RELIABILITY_RANK: Record = {\n \"high\": 4, \"medium\": 3, \"low\": 2, \"very low\": 1, \"n/a\": 0,\n};\n\n/** Valeur triable d'une ligne pour une colonne donnée. Les colonnes mensuelles portent\n * la clé m0…m5 : c'est la valeur finale retenue, corrections manuelles comprises. */\nfunction sortValue(r: any, k: string): number | string {\n if (/^m\\d+$/.test(k)) return r.final[Number(k.slice(1))] ?? 0;\n switch (k) {\n case \"code\": return r.s.code;\n case \"name\": return String(r.s.name ?? \"\").toLowerCase();\n case \"cat\": return String(r.p?.cat ?? r.s.pl ?? \"\").toLowerCase();\n case \"stage\": return String(r.stage);\n case \"dtoc\": return r.ch.dtoc;\n case \"mp\": return r.ch.mp;\n case \"tot90\": return r.ch.dtoc + r.ch.mp;\n case \"hist\": return r.total;\n // Les lignes sans historique (analogue, fin de vie) n'ont pas de fiabilité : elles\n // se regroupent au bas du classement plutôt que de s'intercaler au hasard.\n case \"relia\": return r.basis === \"history\" ? (RELIABILITY_RANK[reliabilityOf(r.cls.kind)] ?? 0) : 0;\n case \"fc\": return r.sumCurrent;\n case \"total6\": return r.sumProposal;\n case \"gap\": return r.sumCurrent > 0 ? (r.sumProposal - r.sumCurrent) / r.sumCurrent : -Infinity;\n default: return 0;\n }\n}\n\nconst monthLabel = (iso: string) => {\n const [y, m] = iso.split(\"-\").map(Number);\n return new Date(Date.UTC(y, m - 1, 1)).toLocaleDateString(\"fr-FR\", { month: \"short\", year: \"2-digit\" });\n};\n\nexport default function Forecast(\n { products, canEdit = true }: { products: Product[]; canEdit?: boolean },\n) {\n const [dp, setDp] = useState(null);\n const [err, setErr] = useState(null);\n const [orders, setOrders] = useState(undefined);\n\n const [brand, setBrand] = useState(null);\n const [grp, setGrp] = useState(null);\n const [q, setQ] = useState(\"\");\n const [onlyGaps, setOnlyGaps] = useState(false);\n // Par défaut on masque les fins de vie : elles ne reçoivent aucune prévision et\n // n'encombreraient la liste que pour rappeler qu'il n'y a rien à décider.\n const [hideEol, setHideEol] = useState(true);\n const [params, setParams] = useState(DEFAULT_FORECAST_PARAMS);\n const [overrides, setOverrides] = useState>({});\n const [detail, setDetail] = useState(null);\n // Tri par colonne, comme dans Products. Par défaut sur le total proposé, décroissant :\n // on veut voir en premier les références qui pèsent le plus dans la prévision.\n const [sort, setSort] = useState<{ k: string; dir: 1 | -1 }>({ k: \"total6\", dir: -1 });\n // Objectif de CA par mois. null = pas d'objectif, la prévision reste celle du modèle.\n const [revTarget, setRevTarget] = useState<(number | null)[]>(Array(HORIZON).fill(null));\n // Campagnes : arbitrages mémorisés, une campagne par fichier DP.\n const [campaigns, setCampaigns] = useState(null);\n const [campaignMeta, setCampaignMeta] = useState(null);\n const [loadedId, setLoadedId] = useState(null);\n const [saveState, setSaveState] = useState<\"idle\" | \"saving\" | \"saved\" | \"error\">(\"idle\");\n const [carryNote, setCarryNote] = useState(null);\n\n useEffect(() => {\n fetch(\"/api/forecast\").then(async (r) => {\n if (!r.ok) { setErr((await r.json().catch(() => ({}))).message ?? `Error ${r.status}`); return; }\n setDp(await r.json());\n }).catch((e) => setErr(String(e)));\n fetch(\"/api/business\").then((r) => r.ok ? r.json() : null).then(setOrders).catch(() => setOrders(null));\n }, []);\n\n const byCode = useMemo(() => {\n const m: Record = {};\n for (const p of products) m[String(p.code)] = p;\n return m;\n }, [products]);\n\n // Expéditions Shopify : par SKU et par mois pour l'historique, et sur 90 jours par canal\n // pour l'affichage. Shopify est la seule source qui distingue DtoC et marketplace, le\n // fichier Demand Planning ne portant qu'une ligne « DtoC » globale.\n const shop = useMemo(() => {\n const monthly: Record> = {};\n const last90: Record = {};\n // Prix de vente moyen réellement constaté, HT et hors éco-participation : la même\n // définition que le CA du reste du dashboard. Bien plus juste que le prix affiché,\n // qui ignore remises et tarifs marketplace.\n const sold: Record = {};\n let dtoc = 0, mp = 0;\n if (!orders?.lines) return { monthly, last90, sold, shareMp: 0 };\n const from = new Date(Date.now() - 90 * 864e5).toISOString().slice(0, 10);\n for (const l of orders.lines) {\n if (l.k !== \"ful\") continue;\n const d = l.f || l.d;\n if (!d) continue;\n const sku = String(l.sku ?? \"\");\n if (!sku) continue;\n (monthly[sku] ??= {})[d.slice(0, 7)] = (monthly[sku][d.slice(0, 7)] ?? 0) + l.q;\n const deee = byCode[sku]?.deee ?? 0;\n const s = (sold[sku] ??= { rev: 0, units: 0 });\n s.rev += (l.n != null ? l.n : l.p * l.q) / 1.2 - deee * l.q;\n s.units += l.q;\n if (d >= from) {\n const isMp = /but_mirakl|cdiscount|boulanger|showroomprive/.test((l.t ?? \"\").toLowerCase());\n const e = (last90[sku] ??= { dtoc: 0, mp: 0 });\n if (isMp) { e.mp += l.q; mp += l.q; } else { e.dtoc += l.q; dtoc += l.q; }\n }\n }\n return { monthly, last90, sold, shareMp: dtoc + mp > 0 ? mp / (dtoc + mp) : 0 };\n }, [orders, byCode]);\n\n /** Mois réellement couverts par Shopify, au-delà de la bascule. Mesuré plutôt que\n * supposé : si un mois n'a rien, on garde la BI pour ce mois. */\n const shopMonths = useMemo(() => {\n const s = new Set();\n for (const m of Object.values(shop.monthly)) {\n for (const [k, v] of Object.entries(m)) if (v > 0 && k >= CUTOVER.slice(0, 7)) s.add(k);\n }\n return s;\n }, [shop]);\n\n useEffect(() => {\n if (shop.shareMp > 0) setParams((p) => ({ ...p, mpShareCurrent: shop.shareMp }));\n }, [shop.shareMp]);\n\n // Mois échus (historique) et mois à prévoir. Le dernier mois d'historique non nul\n // marque la frontière : au-delà, le fichier DP ne porte que de la prévision.\n const split = useMemo(() => {\n if (!dp) return null;\n let last = -1;\n for (let i = 0; i < dp.months.length; i++) {\n if (dp.skus.some((s) => (s.hist[i] ?? 0) > 0)) last = i;\n }\n const target = dp.months.slice(last + 1, last + 1 + HORIZON);\n return { lastHist: last, target, targetIdx: target.map((_, i) => last + 1 + i) };\n }, [dp]);\n\n // Seul le mois calendaire EN COURS est incomplet une fois Shopify utilisé : Shopify est\n // à jour au jour près, contrairement à la BI qui s'arrête à sa date d'export. On ramène\n // ce mois au mois plein au prorata des jours écoulés, sinon le niveau récent — celui qui\n // pilote les premiers mois de prévision — serait sous-estimé d'autant.\n const partial = useMemo(() => {\n const now = new Date();\n const days = now.getDate();\n const of = new Date(now.getFullYear(), now.getMonth() + 1, 0).getDate();\n const month = `${now.getFullYear()}-${String(now.getMonth() + 1).padStart(2, \"0\")}`;\n return { month, factor: days < of ? of / days : 1, days, of };\n }, []);\n\n const rows = useMemo(() => {\n if (!dp || !split) return [];\n\n // ── Composition de l'historique, alignée sur Business et Scorecards ──\n // Avant la bascule du 1er mars 2026, Shopify ne couvre presque rien (26 unités en\n // janvier contre 1 081 facturées) : seule la BI fait foi. Après, les deux sources\n // s'accordent à 1-6 % près et Shopify est plus fraîche — mesuré en juillet, Shopify\n // relève 6 248 unités contre 4 623 pour une BI arrêtée au 25.\n const histOf = (s: DpSku) => {\n const sm = shop.monthly[String(s.code)] ?? {};\n const h = dp.months.slice(0, split.lastHist + 1).map((m, i) => {\n const useShop = m >= CUTOVER.slice(0, 7) && shopMonths.has(m);\n const v = useShop ? (sm[m] ?? 0) : (s.hist[i] ?? 0);\n return m === partial.month ? v * partial.factor : v;\n });\n return h;\n };\n\n const pool = dp.skus\n .filter((s) => histOf(s).reduce((a, b) => a + b, 0) >= 5)\n .map((s) => ({\n code: s.code, name: s.name, brand: s.brand, pl: s.pl,\n hist: histOf(s),\n price: byCode[String(s.code)]?.price ?? null,\n grp: byCode[String(s.code)]?.grp ?? null,\n }));\n\n return dp.skus.map((s) => {\n const p = byCode[String(s.code)];\n const hist = histOf(s);\n const total = hist.reduce((a, b) => a + b, 0);\n const cls = classifyDemand(hist);\n const ch = shop.last90[String(s.code)] ?? { dtoc: 0, mp: 0 };\n const stage = lifeStage(p?.lc, s.pipo);\n\n // « Nouveau » se juge sur le CYCLE DE VIE seul, jamais sur le faible volume.\n //\n // La version précédente déclenchait la recherche d'analogues dès qu'une référence\n // avait moins de 5 ventes. Mesuré sur le fichier : 176 références sont dans ce cas,\n // et AUCUNE n'est une arrivée récente — ce sont des produits en catalogue depuis un\n // an qui ne décollent pas. Leur prêter la courbe de démarrage d'un best-seller leur\n // inventait un volume qu'ils n'ont jamais fait. Un produit qui ne se vend pas n'est\n // pas un produit neuf.\n // Âge commercial : mois écoulés depuis la première vente.\n const firstSale = hist.findIndex((v) => v > 0);\n const commercialAge = firstSale === -1 ? 0 : hist.length - firstSale;\n // L'analogue ne sert qu'à combler une absence de signal. Dès deux mois de vie\n // commerciale, la référence parle d'elle-même — même à une unité par mois, c'est\n // une information plus fiable que la courbe empruntée à des voisins.\n const isNew = stage === \"launch\" && total < 5 && commercialAge <= 1;\n let proposal: number[];\n let basis: \"history\" | \"analogue\" | \"none\" | \"end of life\" = \"history\";\n let analogues: ReturnType = [];\n let newIndex: number | null = null; // rythme du produit rapporté à ses analogues\n let newAge = 0;\n\n if (params.zeroEndOfLife && stage === \"end of life\") {\n proposal = Array(HORIZON).fill(0);\n basis = \"end of life\";\n } else if (isNew) {\n // La fenêtre de démarrage doit couvrir l'âge ET l'horizon, sinon les analogues\n // s'arrêtent avant les mois que l'on cherche à prévoir.\n const age = commercialAge;\n analogues = findAnalogues(\n { pl: s.pl, brand: s.brand, price: p?.price ?? null, grp: p?.grp ?? null },\n pool, 5, age + HORIZON,\n );\n const np = forecastNewProduct(analogues, p?.price ?? null, HORIZON, { age, observed: total });\n proposal = np.values;\n newIndex = np.index;\n newAge = age;\n basis = np.basedOn > 0 ? \"analogue\" : \"none\";\n } else {\n proposal = forecastSku({ hist, horizon: HORIZON, stage, params }).values;\n }\n\n // Alerte approvisionnement : on ne corrige jamais la prévision, on signale que le\n // stock annoncé ne suivra pas. Confondre les deux reviendrait à prévoir ce que\n // l'on sait produire plutôt que ce que le marché demande.\n const shortage = split.targetIdx.map((idx, i) => {\n const a = s.avail[idx] ?? 0;\n return a > 0 && proposal[i] > a;\n });\n\n const current = split.targetIdx.map((i) => s.fc[i] ?? 0);\n const quality = forecastQuality(s.hist, s.fc, split.lastHist + 1);\n\n // Prix unitaire retenu pour convertir les unités en CA : le prix réellement\n // constaté quand la référence a vendu, sinon le prix du site ramené au HT hors\n // éco-participation. Deux ventes suffisent — en dessous, la moyenne serait le\n // reflet d'une promotion isolée.\n const s90 = shop.sold[String(s.code)];\n const asp = s90 && s90.units >= 2 && s90.rev > 0\n ? s90.rev / s90.units\n : Math.max(0, (p?.price ?? 0) / 1.2 - (p?.deee ?? 0));\n\n return {\n s, p, hist, total, cls, ch, isNew, basis, analogues, stage, shortage, newIndex, newAge,\n proposal, current, quality, asp,\n sumCurrent: current.reduce((a, b) => a + b, 0),\n };\n });\n }, [dp, split, byCode, shop, shopMonths, params, partial]);\n\n // ── Objectif de CA → coefficient par mois ──\n // Le CA de référence se calcule sur la proposition BRUTE du modèle, jamais sur la\n // valeur déjà mise à l'échelle : sans cela, saisir un objectif déplacerait sa propre\n // base et le coefficient partirait en boucle.\n const scoped = useMemo(() => {\n const ql = q.trim().toLowerCase();\n return rows.filter((r) => {\n if (brand && r.s.brand !== brand) return false;\n if (grp && (r.p?.grp ?? \"\") !== grp) return false;\n if (hideEol && r.stage === \"end of life\") return false;\n if (onlyGaps && Math.abs(r.sumProposal - r.sumCurrent) < Math.max(3, r.sumCurrent * 0.2)) return false;\n if (ql && !String(r.s.code).includes(ql) && !r.s.name.toLowerCase().includes(ql)) return false;\n return true;\n });\n }, [rows, brand, grp, q, onlyGaps, hideEol]);\n\n /** CA que produirait la proposition brute du modèle, mois par mois. Sert de base au\n * champ objectif : c'est la valeur pré-remplie et le dénominateur du coefficient. */\n const baseRevenue = useMemo(\n () => Array.from({ length: HORIZON }, (_, h) =>\n scoped.reduce((s, r) => s + Math.round(r.proposal[h] ?? 0) * r.asp, 0)),\n [scoped],\n );\n\n /** Coefficient appliqué à chaque mois pour atteindre l'objectif saisi. */\n const revFactor = useMemo(\n () => revTarget.map((t, h) =>\n t != null && t > 0 && baseRevenue[h] > 0 ? t / baseRevenue[h] : 1),\n [revTarget, baseRevenue],\n );\n\n // Quantités affichées : proposition × coefficient d'objectif, sauf saisie manuelle qui\n // l'emporte toujours. Corriger une ligne à la main reste possible après avoir posé un\n // objectif — le CA réellement atteint est alors réaffiché sous le champ.\n const filtered = useMemo(() => {\n return scoped.map((r) => {\n const final = r.proposal.map((v, i) =>\n overrides[`${r.s.code}|${i}`] ?? Math.round(v * revFactor[i]));\n return { ...r, final, sumProposal: final.reduce((a, b) => a + b, 0) };\n }).sort((a, b) => {\n const va = sortValue(a, sort.k), vb = sortValue(b, sort.k);\n if (typeof va === \"string\" || typeof vb === \"string\") {\n return String(va).localeCompare(String(vb)) * sort.dir;\n }\n return ((va as number) - (vb as number)) * sort.dir;\n });\n }, [scoped, overrides, revFactor, sort]);\n\n /** CA effectivement porté par les quantités affichées, corrections comprises. */\n const finalRevenue = useMemo(\n () => Array.from({ length: HORIZON }, (_, h) =>\n filtered.reduce((s, r) => s + (r.final[h] ?? 0) * r.asp, 0)),\n [filtered],\n );\n\n // ── Campagne courante ──\n // Identifiée par le fichier DP ET son premier mois à prévoir : un même fichier redéposé\n // reprend la même campagne, un fichier d'un nouveau mois en ouvre une autre.\n const campaignId = useMemo(\n () => (dp && split?.target.length ? `${dp.source_file}|${split.target[0]}` : null),\n [dp, split],\n );\n\n // Chargement des arbitrages de la campagne. Tant que ce n'est pas fait, on n'enregistre\n // rien — sinon le premier rendu écraserait la sauvegarde par un état vide.\n useEffect(() => {\n if (!campaignId) return;\n fetch(`/api/forecast/campaigns?id=${encodeURIComponent(campaignId)}`)\n .then((r) => r.ok ? r.json() : null)\n .then((j) => {\n setCampaigns(j?.campaigns ?? []);\n if (j?.current) {\n setOverrides(j.current.overrides ?? {});\n setRevTarget(\n Array.from({ length: HORIZON }, (_, i) => j.current.revTarget?.[i] ?? null));\n if (j.current.params) setParams((p) => ({ ...p, ...j.current.params }));\n setCampaignMeta(j.current);\n }\n setLoadedId(campaignId);\n })\n .catch(() => setLoadedId(campaignId));\n }, [campaignId]);\n\n // Enregistrement différé : on attend une seconde de calme pour ne pas écrire à chaque\n // frappe. `mpShareCurrent` est exclu des réglages sauvegardés — c'est une mesure, pas\n // une décision, et elle doit se recalculer sur les données du moment.\n useEffect(() => {\n if (!campaignId || loadedId !== campaignId || !canEdit || !split) return;\n setSaveState(\"saving\");\n const t = setTimeout(() => {\n const { mpShareCurrent: _m, ...saved } = params as any;\n fetch(\"/api/forecast/campaigns\", {\n method: \"PUT\",\n headers: { \"Content-Type\": \"application/json\" },\n body: JSON.stringify({\n id: campaignId, source_file: dp!.source_file, months: split.target,\n overrides, revTarget, params: saved,\n }),\n })\n .then((r) => r.ok ? r.json() : Promise.reject())\n .then((c) => { setCampaignMeta(c); setSaveState(\"saved\"); })\n .catch(() => setSaveState(\"error\"));\n }, 1000);\n return () => clearTimeout(t);\n }, [overrides, revTarget, params, campaignId, loadedId, canEdit, split, dp]);\n\n /** Campagne précédente exploitable : la plus récente autre que celle en cours. */\n const previousCampaign = useMemo(\n () => (campaigns ?? []).find((c) => c.id !== campaignId) ?? null,\n [campaigns, campaignId],\n );\n\n /** Reprise explicite des arbitrages de la campagne précédente, réalignée par mois. */\n const carryOver = async () => {\n if (!previousCampaign || !split) return;\n const r = await fetch(`/api/forecast/campaigns?id=${encodeURIComponent(previousCampaign.id)}`);\n if (!r.ok) return;\n const prev = (await r.json())?.current;\n if (!prev) return;\n const from: string[] = prev.months ?? [];\n const to = split.target;\n let carried = 0, dropped = 0;\n const next: Record = {};\n for (const [key, value] of Object.entries(prev.overrides ?? {})) {\n const [code, idx] = key.split(\"|\");\n const month = from[Number(idx)];\n const t = month ? to.indexOf(month) : -1;\n if (t === -1) { dropped++; continue; }\n next[`${code}|${t}`] = value as number;\n carried++;\n }\n setOverrides(next);\n setRevTarget(to.map((m) => {\n const i = from.indexOf(m);\n return i === -1 ? null : (prev.revTarget?.[i] ?? null);\n }));\n setCarryNote(\n `${carried} correction(s) reprise(s) de ${prev.source_file}` +\n (dropped ? `, ${dropped} abandonnée(s) — le mois concerné n'est plus dans l'horizon.` : \".\"),\n );\n };\n\n if (err) {\n return (\n \n

{err}

\n This tab is fed from Forecast/ValidationDP-*.xlsm when\n push_dashboard.py runs.\n

\n \n );\n }\n if (!dp || !split) return

Loading…

;\n\n const tot = {\n proposal: filtered.reduce((s, r) => s + r.sumProposal, 0),\n current: filtered.reduce((s, r) => s + r.sumCurrent, 0),\n actual: filtered.reduce((s, r) => s + r.quality.actual, 0),\n fcPast: filtered.reduce((s, r) => s + r.quality.forecast, 0),\n };\n const brands = [...new Set(dp.skus.map((s) => s.brand).filter(Boolean))].sort();\n\n /** En-tête cliquable. Premier clic : décroissant, ce qu'on veut presque toujours sur\n * des volumes ; second clic : croissant. */\n const Th = ({ k, children, right = true, extra = \"\" }: any) => (\n setSort((s) => ({ k, dir: s.k === k && s.dir === -1 ? 1 : -1 }))}\n className={`px-2 cursor-pointer select-none whitespace-nowrap hover:text-slate-700 ${\n right ? \"text-right\" : \"text-left\"} ${extra} ${sort.k === k ? \"text-slate-800 font-semibold\" : \"\"}`}\n >\n {children}{sort.k === k ? (sort.dir === -1 ? \" ▼\" : \" ▲\") : \"\"}\n \n );\n\n return (\n

\n \n {canEdit && (\n \n {saveState === \"saving\" ? \"saving…\"\n : saveState === \"saved\" ? \"saved\"\n : saveState === \"error\" ? \"save failed\"\n : campaignMeta ? \"saved\" : \"\"}\n {campaignMeta?.updated_by && saveState !== \"saving\" && (\n · {displayUser(campaignMeta.updated_by)}\n )}\n {\" ·\"}\n \n )}\n {dp.source_file} · imported {new Date(dp.generated_at).toLocaleDateString(\"fr-FR\")}\n {/* Le mois en cours est mis à l'échelle du mois plein : le taire ferait\n apparaître un écart inexplicable avec les autres onglets. */}\n {partial.factor !== 1 && (\n \n {\" \"}· current month × {partial.factor.toFixed(2)}\n \n )}\n }>\n

\n \n setParams({ ...params, window: Math.max(1, Number(e.target.value)) })}\n className=\"w-20 px-2 py-1 border rounded text-right\" />\n \n\n \n

\n setParams({\n ...params,\n mpShareTarget: e.target.value === \"\" ? null : Number(e.target.value) / 100,\n })}\n className=\"w-20 px-2 py-1 border rounded text-right\" />\n %\n {params.mpShareTarget != null && params.mpShareTarget > params.mpShareCurrent && (\n \n → +{(((1 - params.mpShareCurrent) / (1 - params.mpShareTarget) - 1) * 100).toFixed(1)}% volume\n \n )}\n

\n \n\n \n setParams({ ...params, zeroEndOfLife: e.target.checked })} />\n \n

\n \n\n {/* Reprise d'une campagne précédente — jamais automatique. Si le demand planning\n n'a pas validé la proposition d'août, ses corrections ne doivent pas revenir\n déguisées en acquis dans la campagne de septembre. */}\n {canEdit && previousCampaign && Object.keys(overrides).length === 0\n && !revTarget.some((t) => t != null) && !carryNote && (\n

\n \n New campaign — nothing entered yet. The previous one ({previousCampaign.source_file},\n {\" \"}{previousCampaign.months?.[0]} → {previousCampaign.months?.[previousCampaign.months.length - 1]})\n holds your last arbitrations\n {previousCampaign.exported_at\n ? `, exported on ${new Date(previousCampaign.exported_at).toLocaleDateString(\"fr-FR\")}.`\n : \", never exported.\"}\n \n \n

\n )}\n {carryNote && (\n

\n {carryNote} Check each one against what the model now proposes.\n \n

\n )}\n\n {tot.actual > 0 && (\n

\n Accuracy of the forecast in place. Over the {split.lastHist + 1} elapsed months,\n the Commercial Forecast totals {fmtInt(tot.fcPast)} units against {fmtInt(tot.actual)} actually\n invoiced — a gap of {((tot.fcPast - tot.actual) / tot.actual * 100).toFixed(0)}%.\n

\n )}\n\n \n

\n setQ(e.target.value)} placeholder=\"SKU or name…\"\n className=\"px-3 py-1.5 border rounded-lg text-sm w-full sm:w-56\" />\n setBrand(null)}>All brands\n {brands.map((b) => (\n setBrand(b)}>{b}\n ))}\n |\n setGrp(null)}>All categories\n {GROUPS.map((g) => (\n setGrp(g.key)}>{g.label}\n ))}\n \n setOnlyGaps(e.target.checked)} />\n Significant gaps only\n \n \n setHideEol(e.target.checked)} />\n Hide end of life ({rows.filter((r) => r.stage === \"end of life\").length})\n \n \n

\n \n \n \n 0 ? `${((tot.proposal - tot.current) / tot.current * 100).toFixed(0)}%` : \"—\"} />\n

\n \n\n \n

\n \n \n {/* Objectif de CA par mois : pré-rempli avec la projection du modèle, libre\n à la saisie. Modifier une case remet à l'échelle toute la colonne. */}\n \n \n {split.target.map((m, i) => {\n const set = revTarget[i] != null;\n const drift = set && baseRevenue[i] > 0\n ? finalRevenue[i] / (revTarget[i] as number) - 1 : 0;\n return (\n \n );\n })}\n \n \n \n \n \n \n \n \n \n \n \n \n \n {split.target.map((m, i) => (\n \n ))}\n \n \n \n \n \n {filtered.slice(0, 400).map((r) => {\n const gap = r.sumCurrent > 0 ? (r.sumProposal - r.sumCurrent) / r.sumCurrent : null;\n return (\n \n \n \n \n \n \n \n \n \n \n \n {r.final.map((v, i) => (\n \n ))}\n \n \n \n );\n })}\n \n

\n Revenue target (excl. VAT)\n \n — on the {fmtInt(filtered.length)} products displayed; editing a month\n rescales its column\n \n	\n {\n const v = e.target.value.replace(/[^\\d]/g, \"\");\n const next = [...revTarget];\n next[i] = v === \"\" ? null : Number(v);\n setRevTarget(next);\n }}\n title={`Model projection: ${fmtInt(baseRevenue[i])} €. Enter a target to rescale this month.`}\n className={`w-20 px-1 py-0.5 border rounded text-right tabular-nums text-[11px] ${\n set ? \"border-indigo-400 bg-white font-semibold text-indigo-800\" : \"border-transparent bg-transparent text-slate-500\"\n }`} />\n {set && Math.abs(drift) > 0.01 && (\n \n {fmtInt(finalRevenue[i])} €\n \n )}\n	\n {fmtInt(finalRevenue.reduce((a, b) => a + b, 0))} €\n {revTarget.some((t) => t != null) && (\n \n )}\n
SKU	Product	Category	Lifecycle	DtoC 90d	MP 90d	Total 90d	Hist. 12m	Reliability	Current FC	{monthLabel(m)}	Total 6m	Gap
\n \n	\n {r.p?.img && $\"\"$ }\n {r.s.name}\n {r.isNew && NEW}\n	{r.p?.cat ?? r.s.pl}	\n \n	{fmtInt(r.ch.dtoc)}	{fmtInt(r.ch.mp)}	{fmtInt(r.ch.dtoc + r.ch.mp)}	{fmtInt(r.total)}		{fmtInt(r.sumCurrent)}	\n {canEdit ? (\n setOverrides({\n ...overrides, [`${r.s.code}\|${i}`]: Number(e.target.value) \|\| 0,\n })}\n className={`w-12 px-1 py-0.5 border rounded text-right tabular-nums ${\n overrides[`${r.s.code}\|${i}`] != null ? \"border-indigo-400 bg-indigo-50\"\n : r.shortage[i] ? \"border-amber-400 bg-amber-50\"\n : \"border-transparent hover:border-slate-200\"\n }`} />\n ) : v}\n	{fmtInt(r.sumProposal)}	0.2 ? \"text-emerald-700\" : gap < -0.2 ? \"text-rose-700\" : \"text-slate-500\"}`}>\n {gap == null ? \"—\" : `${(gap * 100).toFixed(0)} %`}\n

\n {filtered.length > 400 && (\n

\n Showing the first 400 rows of {filtered.length}. The Excel export contains them all.\n

\n )}\n

\n \n\n {detail != null && (() => {\n // IMPORTANT : chercher dans `filtered`, pas dans `rows`. C'est `filtered` qui\n // ajoute `final` (proposition mise à l'échelle, corrections manuelles comprises).\n // Une ligne prise dans `rows` n'a pas cette propriété : le panneau lisait alors\n // `r.final[i]` sur `undefined`, ce qui faisait tomber toute l'application et\n // affichait une page blanche — visible uniquement sur les références établies,\n // les nouveautés empruntant l'autre branche du panneau.\n const r = filtered.find((x) => x.s.code === detail);\n if (!r) return null;\n return (\n // En surimpression, et non à la suite du tableau : celui-ci occupant 70 % de la\n // hauteur d'écran, le panneau s'ouvrait hors champ et cliquer sur un SKU semblait\n // ne rien faire.\n

setDetail(null)}>\n

e.stopPropagation()}>\n setDetail(null)}\n className=\"px-2 py-1 rounded border border-slate-200 text-xs text-slate-500 hover:bg-slate-50\">\n Close\n \n }>\n

Invoiced history

\n {r.hist.map((v, i) => {\n const max = Math.max(...r.hist, 1);\n return (\n

\n {v || \"\"}\n

\n );\n })}\n

\n {dp.months[0]} → {dp.months[r.hist.length - 1]} ·{\" \"}\n {reliabilityOf(r.cls.kind)} reliability — {r.cls.kind} demand\n {r.cls.adi > 0 && ` (ADI ${r.cls.adi.toFixed(2)}, CV² ${r.cls.cv2.toFixed(2)})`}\n

{reliabilityReason(r.cls.kind)}

\n {r.quality.bias != null && (\n

\n Past forecast: {fmtInt(r.quality.forecast)} predicted for {fmtInt(r.quality.actual)} actual\n ({(r.quality.bias * 100).toFixed(0)}%).\n

\n )}\n

\n {r.isNew ? (\n <>\n

\n Analogue products used\n

\n {r.analogues.length === 0 ? (\n

\n No analogue found. Without a comparable range that has actually started\n selling, the tool proposes nothing rather than inventing a figure.\n

\n ) : (\n \n \n \n \n \n \n {r.analogues.map((m) => (\n \n \n \n \n \n \n ))}\n \n

SKU	Name	Price	Why
{m.candidate.code}	{m.candidate.name}	{fmtEur(m.candidate.price)}	{m.reasons.join(\", \")}

\n )}\n

\n The analogues' ramp-up curves are averaged, then scaled by the price ratio\n (elasticity -0.7 — an assumption, not a measurement).\n

\n \n ) : (\n <>\n

Announced availability

\n {split.target.map((m, i) => {\n const a = r.s.avail[split.targetIdx[i]] ?? 0;\n const f = r.final[i];\n return (\n

\n {monthLabel(m)}\n 0 && f > a ? \"text-rose-600 font-medium\" : \"\"}>\n {fmtInt(f)} forecast / {a > 0 ? fmtInt(a) : \"—\"} available\n \n

\n );\n })}\n

\n In red: forecast above the announced availability. Flagged but never capped —\n that is a supply matter, not a forecasting one.\n

\n \n )}\n

\n \n

\n );\n })()}\n\n {dp.outside?.length > 0 && (\n \n

\n These products were invoiced in 2026 but do not appear in the validation file. They\n represent {fmtInt(dp.outside.reduce((s, o) => s + o.hist.reduce((a, b) => a + b, 0), 0))} units\n and currently receive no forecast at all.\n

\n \n \n \n \n \n \n {dp.outside.slice(0, 25).map((o) => (\n \n \n \n \n \n ))}\n \n

SKU	Product	Invoiced 2026
{o.code}	{byCode[String(o.code)]?.name ?? \"—\"}	{fmtInt(o.hist.reduce((a, b) => a + b, 0))}

\n \n )}\n

\n );\n}\n\n/** Export au format attendu par l'équipe demand planning : une ligne par référence,\n * une colonne par mois, plus les éléments de contexte qui permettent de challenger\n * la proposition sans revenir vers nous. */\nfunction exportXlsx(rows: any[], target: string[], dp: DpPayload) {\n const data = rows.map((r) => {\n const o: Record = {\n \"Product Code\": r.s.code,\n \"Product Description\": r.s.name,\n \"Brand\": r.s.brand,\n \"PL Name\": r.s.pl,\n \"PIPO\": r.s.pipo,\n \"Catalogue lifecycle\": r.p?.lc ?? \"\",\n \"Lifecycle used\": r.stage,\n \"Account\": \"DtoC\",\n \"Source Table\": \"Commercial Forecast\",\n \"Forecast reliability\": r.basis === \"history\" ? reliabilityOf(r.cls.kind) : \"n/a\",\n \"Demand profile\": r.cls.kind,\n \"Basis\": r.basis,\n \"DtoC sales 90d\": Math.round(r.ch.dtoc),\n \"Marketplace sales 90d\": Math.round(r.ch.mp),\n \"Total sales 90d\": Math.round(r.ch.dtoc + r.ch.mp),\n \"History 12m\": Math.round(r.total),\n \"Current FC 6m\": Math.round(r.sumCurrent),\n };\n o[\"Avg selling price (excl. VAT)\"] = Math.round(r.asp * 100) / 100;\n target.forEach((m, i) => { o[m] = r.final[i]; });\n o[\"Total 6 months\"] = r.sumProposal;\n o[\"Revenue 6 months (excl. VAT)\"] = Math.round(r.sumProposal * r.asp);\n o[\"Gap vs current FC\"] = r.sumCurrent > 0\n ? Math.round((r.sumProposal - r.sumCurrent) / r.sumCurrent * 100) / 100 : null;\n return o;\n });\n\n const wb = XLSX.utils.book_new();\n XLSX.utils.book_append_sheet(wb, XLSX.utils.json_to_sheet(data), \"Forecast\");\n\n const notes = [\n [\"D2C France forecast export\"],\n [\"Generated\", new Date().toLocaleString(\"fr-FR\")],\n [\"Source DP file\", dp.source_file],\n [\"Products\", data.length],\n [],\n [\"Method\"],\n [\"Level\", \"Average of sales over the chosen history window, weighted towards recent months at short horizon and smoothed at long horizon\"],\n [\"End of life\", \"Set to zero: a discontinued product is not meant to be replenished\"],\n [\"New products\", \"Average of analogue products' ramp-up curves, scaled by the price ratio\"],\n [\"No decay\", \"No ageing coefficient is applied: the lifecycle explains the phenomenon better\"],\n [],\n [\"Reliability\", \"Derived from how spaced out and how uneven the sales history is (Syntetos-Boylan-Croston). High = sells steadily every month; very low = sporadic sales of very uneven size, worth overriding by hand\"],\n [],\n [\"Known limitations\"],\n [\"Seasonality\", \"Not applied — weak on large appliances\"],\n [\"Marketplace\", \"Ramp-up entered manually, not inferred from history\"],\n [\"Availability\", \"Flagged when the forecast exceeds it, never used to cap the forecast\"],\n [\"Scope\", `${dp.outside?.length ?? 0} invoiced products outside the DP file are not forecast here`],\n ];\n XLSX.utils.book_append_sheet(wb, XLSX.utils.aoa_to_sheet(notes), \"Method\");\n\n const d = new Date().toISOString().slice(0, 10);\n XLSX.writeFile(wb, `forecast_D2C_${d}.xlsx`);\n}\n\nfunction StageBadge({ stage }: { stage: LifeStage }) {\n const tone = stage === \"active\" ? \"bg-emerald-100 text-emerald-700\"\n : stage === \"launch\" ? \"bg-indigo-100 text-indigo-700\"\n : stage === \"end of life\" ? \"bg-slate-200 text-slate-600\"\n : \"bg-slate-100 text-slate-400\";\n return {stage};\n}\n\n/** Ce que l'on peut attendre du chiffre proposé, plutôt que la forme de la série.\n * Le profil de demande sous-jacent reste dans l'infobulle : c'est la justification,\n * pas la conclusion. */\nfunction ClassBadge({ kind, basis }: { kind: any; basis: string }) {\n if (basis === \"end of life\") {\n return not forecast;\n }\n if (basis === \"analogue\") {\n return (\n \n from analogue\n \n );\n }\n if (basis === \"none\") {\n return (\n \n no basis\n \n );\n }\n const level = reliabilityOf(kind);\n const tone = level === \"high\" ? \"bg-emerald-100 text-emerald-700\"\n : level === \"medium\" ? \"bg-amber-100 text-amber-700\"\n : level === \"low\" ? \"bg-orange-100 text-orange-700\"\n : level === \"very low\" ? \"bg-rose-100 text-rose-700\"\n : \"bg-slate-100 text-slate-400\";\n return (\n \n {level}\n \n );\n}\n\nfunction Field({ label, help, children }: any) {\n return (\n

{label}

\n {children}\n

{help}

\n );\n}\n\nfunction Stat({ label, value, accent }: any) {\n return (\n

{label}

{value}

\n );\n}\n\nfunction Pill({ active, onClick, children }: any) {\n return (\n \n );\n}\n\nfunction Card({ title, children, right }: any) {\n return (\n

\n {title && (\n

{title}

\n {right}\n

\n )}\n

{children}

\n );\n}\n","/** @jsxImportSource https://esm.sh/react@18.2.0 */\n// Onglet Forecast — proposition de prévision à 6 mois, référence par référence.\n//\n// L'outil ne remplace pas le jugement commercial : il produit une proposition chiffrée,\n// affiche ce sur quoi elle repose, et laisse corriger chaque ligne. La colonne finale\n// exportée est TOUJOURS celle que vous avez validée, jamais la proposition brute.\nimport { useEffect, useMemo, useState } from \"https://esm.sh/react@18.2.0\";\nimport * as XLSX from \"https://esm.sh/xlsx@0.18.5\";\nimport {\n classifyDemand, DEFAULT_FORECAST_PARAMS, findAnalogues, forecastNewProduct,\n forecastQuality, forecastSku, lifeStage, reliabilityOf, reliabilityReason,\n type ForecastParams, type LifeStage,\n} from \"/frontend/forecasting.ts\";\nimport {\n DEFAULT_CUTOVER as CUTOVER, displayUser, fmtInt, fmtEur, GROUPS, type Product,\n} from \"/frontend/helpers.ts\";\n\nconst NAVY = \"#1F3864\";\nconst ACCENT = \"#4F46E5\";\nconst HORIZON = 6;\n\ninterface DpSku {\n code: number; name: string; brand: string; pl: string; pipo: string;\n top80: boolean; eta: string | null;\n hist: number[]; fc: number[]; fcm1: number[]; avail: number[];\n}\ninterface DpPayload {\n source_file: string; generated_at: string; months: string[];\n skus: DpSku[];\n outside: { code: number; months: string[]; hist: number[] }[];\n /** Date de dernière facturation dans l'export BI — le mois qui la contient est partiel. */\n bi_asof: string | null;\n bi_months: string[];\n}\n\n/** Ordre des niveaux de fiabilité, pour que le tri suive l'échelle et non l'alphabet. */\nconst RELIABILITY_RANK: Record = {\n \"high\": 4, \"medium\": 3, \"low\": 2, \"very low\": 1, \"n/a\": 0,\n};\n\n/** Valeur triable d'une ligne pour une colonne donnée. Les colonnes mensuelles portent\n * la clé m0…m5 : c'est la valeur finale retenue, corrections manuelles comprises. */\nfunction sortValue(r: any, k: string): number | string {\n if (/^m\\d+$/.test(k)) return r.final[Number(k.slice(1))] ?? 0;\n switch (k) {\n case \"code\": return r.s.code;\n case \"name\": return String(r.s.name ?? \"\").toLowerCase();\n case \"cat\": return String(r.p?.cat ?? r.s.pl ?? \"\").toLowerCase();\n case \"stage\": return String(r.stage);\n case \"dtoc\": return r.ch.dtoc;\n case \"mp\": return r.ch.mp;\n case \"tot90\": return r.ch.dtoc + r.ch.mp;\n case \"hist\": return r.total;\n // Les lignes sans historique (analogue, fin de vie) n'ont pas de fiabilité : elles\n // se regroupent au bas du classement plutôt que de s'intercaler au hasard.\n case \"relia\": return r.basis === \"history\" ? (RELIABILITY_RANK[reliabilityOf(r.cls.kind)] ?? 0) : 0;\n case \"fc\": return r.sumCurrent;\n case \"total6\": return r.sumProposal;\n case \"gap\": return r.sumCurrent > 0 ? (r.sumProposal - r.sumCurrent) / r.sumCurrent : -Infinity;\n default: return 0;\n }\n}\n\nconst monthLabel = (iso: string) => {\n const [y, m] = iso.split(\"-\").map(Number);\n return new Date(Date.UTC(y, m - 1, 1)).toLocaleDateString(\"fr-FR\", { month: \"short\", year: \"2-digit\" });\n};\n\nexport default function Forecast(\n { products, canEdit = true }: { products: Product[]; canEdit?: boolean },\n) {\n const [dp, setDp] = useState(null);\n const [err, setErr] = useState(null);\n const [orders, setOrders] = useState(undefined);\n\n const [brand, setBrand] = useState(null);\n const [grp, setGrp] = useState(null);\n const [q, setQ] = useState(\"\");\n const [onlyGaps, setOnlyGaps] = useState(false);\n // Par défaut on masque les fins de vie : elles ne reçoivent aucune prévision et\n // n'encombreraient la liste que pour rappeler qu'il n'y a rien à décider.\n const [hideEol, setHideEol] = useState(true);\n const [params, setParams] = useState(DEFAULT_FORECAST_PARAMS);\n const [overrides, setOverrides] = useState>({});\n const [detail, setDetail] = useState(null);\n // Tri par colonne, comme dans Products. Par défaut sur le total proposé, décroissant :\n // on veut voir en premier les références qui pèsent le plus dans la prévision.\n const [sort, setSort] = useState<{ k: string; dir: 1 | -1 }>({ k: \"total6\", dir: -1 });\n // Objectif de CA par mois. null = pas d'objectif, la prévision reste celle du modèle.\n const [revTarget, setRevTarget] = useState<(number | null)[]>(Array(HORIZON).fill(null));\n // Campagnes : arbitrages mémorisés, une campagne par fichier DP.\n const [campaigns, setCampaigns] = useState(null);\n const [campaignMeta, setCampaignMeta] = useState(null);\n const [loadedId, setLoadedId] = useState(null);\n const [saveState, setSaveState] = useState<\"idle\" | \"saving\" | \"saved\" | \"error\">(\"idle\");\n const [carryNote, setCarryNote] = useState(null);\n\n useEffect(() => {\n fetch(\"/api/forecast\").then(async (r) => {\n if (!r.ok) { setErr((await r.json().catch(() => ({}))).message ?? `Error ${r.status}`); return; }\n setDp(await r.json());\n }).catch((e) => setErr(String(e)));\n fetch(\"/api/business\").then((r) => r.ok ? r.json() : null).then(setOrders).catch(() => setOrders(null));\n }, []);\n\n const byCode = useMemo(() => {\n const m: Record = {};\n for (const p of products) m[String(p.code)] = p;\n return m;\n }, [products]);\n\n // Expéditions Shopify : par SKU et par mois pour l'historique, et sur 90 jours par canal\n // pour l'affichage. Shopify est la seule source qui distingue DtoC et marketplace, le\n // fichier Demand Planning ne portant qu'une ligne « DtoC » globale.\n const shop = useMemo(() => {\n const monthly: Record> = {};\n const last90: Record = {};\n // Prix de vente moyen réellement constaté, HT et hors éco-participation : la même\n // définition que le CA du reste du dashboard. Bien plus juste que le prix affiché,\n // qui ignore remises et tarifs marketplace.\n const sold: Record = {};\n let dtoc = 0, mp = 0;\n if (!orders?.lines) return { monthly, last90, sold, shareMp: 0 };\n const from = new Date(Date.now() - 90 * 864e5).toISOString().slice(0, 10);\n for (const l of orders.lines) {\n if (l.k !== \"ful\") continue;\n const d = l.f || l.d;\n if (!d) continue;\n const sku = String(l.sku ?? \"\");\n if (!sku) continue;\n (monthly[sku] ??= {})[d.slice(0, 7)] = (monthly[sku][d.slice(0, 7)] ?? 0) + l.q;\n const deee = byCode[sku]?.deee ?? 0;\n const s = (sold[sku] ??= { rev: 0, units: 0 });\n s.rev += (l.n != null ? l.n : l.p * l.q) / 1.2 - deee * l.q;\n s.units += l.q;\n if (d >= from) {\n const isMp = /but_mirakl|cdiscount|boulanger|showroomprive/.test((l.t ?? \"\").toLowerCase());\n const e = (last90[sku] ??= { dtoc: 0, mp: 0 });\n if (isMp) { e.mp += l.q; mp += l.q; } else { e.dtoc += l.q; dtoc += l.q; }\n }\n }\n return { monthly, last90, sold, shareMp: dtoc + mp > 0 ? mp / (dtoc + mp) : 0 };\n }, [orders, byCode]);\n\n /** Mois réellement couverts par Shopify, au-delà de la bascule. Mesuré plutôt que\n * supposé : si un mois n'a rien, on garde la BI pour ce mois. */\n const shopMonths = useMemo(() => {\n const s = new Set();\n for (const m of Object.values(shop.monthly)) {\n for (const [k, v] of Object.entries(m)) if (v > 0 && k >= CUTOVER.slice(0, 7)) s.add(k);\n }\n return s;\n }, [shop]);\n\n useEffect(() => {\n if (shop.shareMp > 0) setParams((p) => ({ ...p, mpShareCurrent: shop.shareMp }));\n }, [shop.shareMp]);\n\n // Mois échus (historique) et mois à prévoir. Le dernier mois d'historique non nul\n // marque la frontière : au-delà, le fichier DP ne porte que de la prévision.\n const split = useMemo(() => {\n if (!dp) return null;\n let last = -1;\n for (let i = 0; i < dp.months.length; i++) {\n if (dp.skus.some((s) => (s.hist[i] ?? 0) > 0)) last = i;\n }\n const target = dp.months.slice(last + 1, last + 1 + HORIZON);\n return { lastHist: last, target, targetIdx: target.map((_, i) => last + 1 + i) };\n }, [dp]);\n\n // Seul le mois calendaire EN COURS est incomplet une fois Shopify utilisé : Shopify est\n // à jour au jour près, contrairement à la BI qui s'arrête à sa date d'export. On ramène\n // ce mois au mois plein au prorata des jours écoulés, sinon le niveau récent — celui qui\n // pilote les premiers mois de prévision — serait sous-estimé d'autant.\n const partial = useMemo(() => {\n const now = new Date();\n const days = now.getDate();\n const of = new Date(now.getFullYear(), now.getMonth() + 1, 0).getDate();\n const month = `${now.getFullYear()}-${String(now.getMonth() + 1).padStart(2, \"0\")}`;\n return { month, factor: days < of ? of / days : 1, days, of };\n }, []);\n\n const rows = useMemo(() => {\n if (!dp || !split) return [];\n\n // ── Composition de l'historique, alignée sur Business et Scorecards ──\n // Avant la bascule du 1er mars 2026, Shopify ne couvre presque rien (26 unités en\n // janvier contre 1 081 facturées) : seule la BI fait foi. Après, les deux sources\n // s'accordent à 1-6 % près et Shopify est plus fraîche — mesuré en juillet, Shopify\n // relève 6 248 unités contre 4 623 pour une BI arrêtée au 25.\n const histOf = (s: DpSku) => {\n const sm = shop.monthly[String(s.code)] ?? {};\n const h = dp.months.slice(0, split.lastHist + 1).map((m, i) => {\n const useShop = m >= CUTOVER.slice(0, 7) && shopMonths.has(m);\n const v = useShop ? (sm[m] ?? 0) : (s.hist[i] ?? 0);\n return m === partial.month ? v * partial.factor : v;\n });\n return h;\n };\n\n const pool = dp.skus\n .filter((s) => histOf(s).reduce((a, b) => a + b, 0) >= 5)\n .map((s) => ({\n code: s.code, name: s.name, brand: s.brand, pl: s.pl,\n hist: histOf(s),\n price: byCode[String(s.code)]?.price ?? null,\n grp: byCode[String(s.code)]?.grp ?? null,\n }));\n\n return dp.skus.map((s) => {\n const p = byCode[String(s.code)];\n const hist = histOf(s);\n const total = hist.reduce((a, b) => a + b, 0);\n const cls = classifyDemand(hist);\n const ch = shop.last90[String(s.code)] ?? { dtoc: 0, mp: 0 };\n const stage = lifeStage(p?.lc, s.pipo);\n\n // « Nouveau » se juge sur le CYCLE DE VIE seul, jamais sur le faible volume.\n //\n // La version précédente déclenchait la recherche d'analogues dès qu'une référence\n // avait moins de 5 ventes. Mesuré sur le fichier : 176 références sont dans ce cas,\n // et AUCUNE n'est une arrivée récente — ce sont des produits en catalogue depuis un\n // an qui ne décollent pas. Leur prêter la courbe de démarrage d'un best-seller leur\n // inventait un volume qu'ils n'ont jamais fait. Un produit qui ne se vend pas n'est\n // pas un produit neuf.\n // Âge commercial : mois écoulés depuis la première vente.\n const firstSale = hist.findIndex((v) => v > 0);\n const commercialAge = firstSale === -1 ? 0 : hist.length - firstSale;\n // L'analogue ne sert qu'à combler une absence de signal. Dès deux mois de vie\n // commerciale, la référence parle d'elle-même — même à une unité par mois, c'est\n // une information plus fiable que la courbe empruntée à des voisins.\n const isNew = stage === \"launch\" && total < 5 && commercialAge <= 1;\n let proposal: number[];\n let basis: \"history\" | \"analogue\" | \"none\" | \"end of life\" = \"history\";\n let analogues: ReturnType = [];\n let newIndex: number | null = null; // rythme du produit rapporté à ses analogues\n let newAge = 0;\n\n if (params.zeroEndOfLife && stage === \"end of life\") {\n proposal = Array(HORIZON).fill(0);\n basis = \"end of life\";\n } else if (isNew) {\n // La fenêtre de démarrage doit couvrir l'âge ET l'horizon, sinon les analogues\n // s'arrêtent avant les mois que l'on cherche à prévoir.\n const age = commercialAge;\n analogues = findAnalogues(\n { pl: s.pl, brand: s.brand, price: p?.price ?? null, grp: p?.grp ?? null },\n pool, 5, age + HORIZON,\n );\n const np = forecastNewProduct(analogues, p?.price ?? null, HORIZON, { age, observed: total });\n proposal = np.values;\n newIndex = np.index;\n newAge = age;\n basis = np.basedOn > 0 ? \"analogue\" : \"none\";\n } else {\n proposal = forecastSku({ hist, horizon: HORIZON, stage, params }).values;\n }\n\n // Alerte approvisionnement : on ne corrige jamais la prévision, on signale que le\n // stock annoncé ne suivra pas. Confondre les deux reviendrait à prévoir ce que\n // l'on sait produire plutôt que ce que le marché demande.\n const shortage = split.targetIdx.map((idx, i) => {\n const a = s.avail[idx] ?? 0;\n return a > 0 && proposal[i] > a;\n });\n\n const current = split.targetIdx.map((i) => s.fc[i] ?? 0);\n const quality = forecastQuality(s.hist, s.fc, split.lastHist + 1);\n\n // Prix unitaire retenu pour convertir les unités en CA : le prix réellement\n // constaté quand la référence a vendu, sinon le prix du site ramené au HT hors\n // éco-participation. Deux ventes suffisent — en dessous, la moyenne serait le\n // reflet d'une promotion isolée.\n const s90 = shop.sold[String(s.code)];\n const asp = s90 && s90.units >= 2 && s90.rev > 0\n ? s90.rev / s90.units\n : Math.max(0, (p?.price ?? 0) / 1.2 - (p?.deee ?? 0));\n\n return {\n s, p, hist, total, cls, ch, isNew, basis, analogues, stage, shortage, newIndex, newAge,\n proposal, current, quality, asp,\n sumCurrent: current.reduce((a, b) => a + b, 0),\n };\n });\n }, [dp, split, byCode, shop, shopMonths, params, partial]);\n\n // ── Objectif de CA → coefficient par mois ──\n // Le CA de référence se calcule sur la proposition BRUTE du modèle, jamais sur la\n // valeur déjà mise à l'échelle : sans cela, saisir un objectif déplacerait sa propre\n // base et le coefficient partirait en boucle.\n const scoped = useMemo(() => {\n const ql = q.trim().toLowerCase();\n return rows.filter((r) => {\n if (brand && r.s.brand !== brand) return false;\n if (grp && (r.p?.grp ?? \"\") !== grp) return false;\n if (hideEol && r.stage === \"end of life\") return false;\n if (onlyGaps && Math.abs(r.sumProposal - r.sumCurrent) < Math.max(3, r.sumCurrent * 0.2)) return false;\n if (ql && !String(r.s.code).includes(ql) && !r.s.name.toLowerCase().includes(ql)) return false;\n return true;\n });\n }, [rows, brand, grp, q, onlyGaps, hideEol]);\n\n /** CA que produirait la proposition brute du modèle, mois par mois. Sert de base au\n * champ objectif : c'est la valeur pré-remplie et le dénominateur du coefficient. */\n const baseRevenue = useMemo(\n () => Array.from({ length: HORIZON }, (_, h) =>\n scoped.reduce((s, r) => s + Math.round(r.proposal[h] ?? 0) * r.asp, 0)),\n [scoped],\n );\n\n /** Coefficient appliqué à chaque mois pour atteindre l'objectif saisi. */\n const revFactor = useMemo(\n () => revTarget.map((t, h) =>\n t != null && t > 0 && baseRevenue[h] > 0 ? t / baseRevenue[h] : 1),\n [revTarget, baseRevenue],\n );\n\n // Quantités affichées : proposition × coefficient d'objectif, sauf saisie manuelle qui\n // l'emporte toujours. Corriger une ligne à la main reste possible après avoir posé un\n // objectif — le CA réellement atteint est alors réaffiché sous le champ.\n const filtered = useMemo(() => {\n return scoped.map((r) => {\n const final = r.proposal.map((v, i) =>\n overrides[`${r.s.code}|${i}`] ?? Math.round(v * revFactor[i]));\n return { ...r, final, sumProposal: final.reduce((a, b) => a + b, 0) };\n }).sort((a, b) => {\n const va = sortValue(a, sort.k), vb = sortValue(b, sort.k);\n if (typeof va === \"string\" || typeof vb === \"string\") {\n return String(va).localeCompare(String(vb)) * sort.dir;\n }\n return ((va as number) - (vb as number)) * sort.dir;\n });\n }, [scoped, overrides, revFactor, sort]);\n\n /** CA effectivement porté par les quantités affichées, corrections comprises. */\n const finalRevenue = useMemo(\n () => Array.from({ length: HORIZON }, (_, h) =>\n filtered.reduce((s, r) => s + (r.final[h] ?? 0) * r.asp, 0)),\n [filtered],\n );\n\n // ── Campagne courante ──\n // Identifiée par le fichier DP ET son premier mois à prévoir : un même fichier redéposé\n // reprend la même campagne, un fichier d'un nouveau mois en ouvre une autre.\n const campaignId = useMemo(\n () => (dp && split?.target.length ? `${dp.source_file}|${split.target[0]}` : null),\n [dp, split],\n );\n\n // Chargement des arbitrages de la campagne. Tant que ce n'est pas fait, on n'enregistre\n // rien — sinon le premier rendu écraserait la sauvegarde par un état vide.\n useEffect(() => {\n if (!campaignId) return;\n fetch(`/api/forecast/campaigns?id=${encodeURIComponent(campaignId)}`)\n .then((r) => r.ok ? r.json() : null)\n .then((j) => {\n setCampaigns(j?.campaigns ?? []);\n if (j?.current) {\n setOverrides(j.current.overrides ?? {});\n setRevTarget(\n Array.from({ length: HORIZON }, (_, i) => j.current.revTarget?.[i] ?? null));\n if (j.current.params) setParams((p) => ({ ...p, ...j.current.params }));\n setCampaignMeta(j.current);\n }\n setLoadedId(campaignId);\n })\n .catch(() => setLoadedId(campaignId));\n }, [campaignId]);\n\n // Enregistrement différé : on attend une seconde de calme pour ne pas écrire à chaque\n // frappe. `mpShareCurrent` est exclu des réglages sauvegardés — c'est une mesure, pas\n // une décision, et elle doit se recalculer sur les données du moment.\n useEffect(() => {\n if (!campaignId || loadedId !== campaignId || !canEdit || !split) return;\n setSaveState(\"saving\");\n const t = setTimeout(() => {\n const { mpShareCurrent: _m, ...saved } = params as any;\n fetch(\"/api/forecast/campaigns\", {\n method: \"PUT\",\n headers: { \"Content-Type\": \"application/json\" },\n body: JSON.stringify({\n id: campaignId, source_file: dp!.source_file, months: split.target,\n overrides, revTarget, params: saved,\n }),\n })\n .then((r) => r.ok ? r.json() : Promise.reject())\n .then((c) => { setCampaignMeta(c); setSaveState(\"saved\"); })\n .catch(() => setSaveState(\"error\"));\n }, 1000);\n return () => clearTimeout(t);\n }, [overrides, revTarget, params, campaignId, loadedId, canEdit, split, dp]);\n\n /** Campagne précédente exploitable : la plus récente autre que celle en cours. */\n const previousCampaign = useMemo(\n () => (campaigns ?? []).find((c) => c.id !== campaignId) ?? null,\n [campaigns, campaignId],\n );\n\n /** Reprise explicite des arbitrages de la campagne précédente, réalignée par mois. */\n const carryOver = async () => {\n if (!previousCampaign || !split) return;\n const r = await fetch(`/api/forecast/campaigns?id=${encodeURIComponent(previousCampaign.id)}`);\n if (!r.ok) return;\n const prev = (await r.json())?.current;\n if (!prev) return;\n const from: string[] = prev.months ?? [];\n const to = split.target;\n let carried = 0, dropped = 0;\n const next: Record = {};\n for (const [key, value] of Object.entries(prev.overrides ?? {})) {\n const [code, idx] = key.split(\"|\");\n const month = from[Number(idx)];\n const t = month ? to.indexOf(month) : -1;\n if (t === -1) { dropped++; continue; }\n next[`${code}|${t}`] = value as number;\n carried++;\n }\n setOverrides(next);\n setRevTarget(to.map((m) => {\n const i = from.indexOf(m);\n return i === -1 ? null : (prev.revTarget?.[i] ?? null);\n }));\n setCarryNote(\n `${carried} correction(s) reprise(s) de ${prev.source_file}` +\n (dropped ? `, ${dropped} abandonnée(s) — le mois concerné n'est plus dans l'horizon.` : \".\"),\n );\n };\n\n if (err) {\n return (\n \n

{err}

\n This tab is fed from Forecast/ValidationDP-*.xlsm when\n push_dashboard.py runs.\n

\n \n );\n }\n if (!dp || !split) return

Loading…

\n \n setParams({ ...params, window: Math.max(1, Number(e.target.value)) })}\n className=\"w-20 px-2 py-1 border rounded text-right\" />\n \n\n \n

\n \n\n \n setParams({ ...params, zeroEndOfLife: e.target.checked })} />\n \n

\n )}\n {carryNote && (\n

\n {carryNote} Check each one against what the model now proposes.\n \n

\n )}\n\n {tot.actual > 0 && (\n

\n )}\n\n \n

\n \n \n \n 0 ? `${((tot.proposal - tot.current) / tot.current * 100).toFixed(0)}%` : \"—\"} />\n

\n \n\n \n

\n Revenue target (excl. VAT)\n \n — on the {fmtInt(filtered.length)} products displayed; editing a month\n rescales its column\n \n	\n {\n const v = e.target.value.replace(/[^\\d]/g, \"\");\n const next = [...revTarget];\n next[i] = v === \"\" ? null : Number(v);\n setRevTarget(next);\n }}\n title={`Model projection: ${fmtInt(baseRevenue[i])} €. Enter a target to rescale this month.`}\n className={`w-20 px-1 py-0.5 border rounded text-right tabular-nums text-[11px] ${\n set ? \"border-indigo-400 bg-white font-semibold text-indigo-800\" : \"border-transparent bg-transparent text-slate-500\"\n }`} />\n {set && Math.abs(drift) > 0.01 && (\n \n {fmtInt(finalRevenue[i])} €\n \n )}\n	\n {fmtInt(finalRevenue.reduce((a, b) => a + b, 0))} €\n {revTarget.some((t) => t != null) && (\n \n )}\n
SKU	Product	Category	Lifecycle	DtoC 90d	MP 90d	Total 90d	Hist. 12m	Reliability	Current FC	{monthLabel(m)}	Total 6m	Gap
\n \n	\n {r.p?.img && $\"\"$ }\n {r.s.name}\n {r.isNew && NEW}\n	{r.p?.cat ?? r.s.pl}	\n \n	{fmtInt(r.ch.dtoc)}	{fmtInt(r.ch.mp)}	{fmtInt(r.ch.dtoc + r.ch.mp)}	{fmtInt(r.total)}		{fmtInt(r.sumCurrent)}	\n {canEdit ? (\n setOverrides({\n ...overrides, [`${r.s.code}\|${i}`]: Number(e.target.value) \|\| 0,\n })}\n className={`w-12 px-1 py-0.5 border rounded text-right tabular-nums ${\n overrides[`${r.s.code}\|${i}`] != null ? \"border-indigo-400 bg-indigo-50\"\n : r.shortage[i] ? \"border-amber-400 bg-amber-50\"\n : \"border-transparent hover:border-slate-200\"\n }`} />\n ) : v}\n	{fmtInt(r.sumProposal)}	0.2 ? \"text-emerald-700\" : gap < -0.2 ? \"text-rose-700\" : \"text-slate-500\"}`}>\n {gap == null ? \"—\" : `${(gap * 100).toFixed(0)} %`}\n

\n {filtered.length > 400 && (\n

\n Showing the first 400 rows of {filtered.length}. The Excel export contains them all.\n

\n )}\n

setDetail(null)}>\n

e.stopPropagation()}>\n setDetail(null)}\n className=\"px-2 py-1 rounded border border-slate-200 text-xs text-slate-500 hover:bg-slate-50\">\n Close\n \n }>\n

Invoiced history

\n {/* Hauteurs en PIXELS, pas en pourcentage.\n Le conteneur est en `items-end`, ce qui dimensionne chaque colonne\n sur son contenu : sa hauteur devient indéfinie, et une barre en\n `height: X%` ne peut alors pas se résoudre — elle tombait à zéro.\n Seules les étiquettes restaient visibles, sans aucune barre.\n On calcule donc les hauteurs en dur sur une piste de 84 px, les\n 12 px restants revenant à l'étiquette. */}\n {(() => {\n const TRACK = 84; // hauteur utile des barres, en pixels\n const max = Math.max(...r.hist, 1);\n return (\n

\n {r.hist.map((v, i) => {\n // Une valeur non nulle garde 2 px : sans ce plancher, une vente\n // isolée face à un pic à 300 serait rigoureusement invisible et\n // se confondrait avec un mois sans vente.\n const h = v > 0 ? Math.max(2, Math.round(v / max * TRACK)) : 0;\n return (\n

\n \n {v || \"\"}\n \n {/* Piste grise sur toute la hauteur : un mois sans vente se\n lit comme un zéro mesuré, et non comme une donnée absente. */}\n

\n );\n })}\n

\n );\n })()}\n

{reliabilityReason(r.cls.kind)}

\n {r.quality.bias != null && (\n

\n Past forecast: {fmtInt(r.quality.forecast)} predicted for {fmtInt(r.quality.actual)} actual\n ({(r.quality.bias * 100).toFixed(0)}%).\n

\n )}\n

\n {r.isNew ? (\n <>\n

\n Analogue products used\n

\n {r.analogues.length === 0 ? (\n

\n No analogue found. Without a comparable range that has actually started\n selling, the tool proposes nothing rather than inventing a figure.\n

\n ) : (\n \n \n \n \n \n \n {r.analogues.map((m) => (\n \n \n \n \n \n \n ))}\n \n

SKU	Name	Price	Why
{m.candidate.code}	{m.candidate.name}	{fmtEur(m.candidate.price)}	{m.reasons.join(\", \")}

\n )}\n

\n The analogues' ramp-up curves are averaged, then scaled by the price ratio\n (elasticity -0.7 — an assumption, not a measurement).\n

\n \n ) : (\n <>\n

Announced availability

\n {split.target.map((m, i) => {\n const a = r.s.avail[split.targetIdx[i]] ?? 0;\n const f = r.final[i];\n return (\n

\n {monthLabel(m)}\n 0 && f > a ? \"text-rose-600 font-medium\" : \"\"}>\n {fmtInt(f)} forecast / {a > 0 ? fmtInt(a) : \"—\"} available\n \n

\n );\n })}\n

\n In red: forecast above the announced availability. Flagged but never capped —\n that is a supply matter, not a forecasting one.\n

\n \n )}\n

\n \n

\n );\n })()}\n\n {dp.outside?.length > 0 && (\n \n

\n \n \n \n \n \n \n {dp.outside.slice(0, 25).map((o) => (\n \n \n \n \n \n ))}\n \n

SKU	Product	Invoiced 2026
{o.code}	{byCode[String(o.code)]?.name ?? \"—\"}	{fmtInt(o.hist.reduce((a, b) => a + b, 0))}

\n \n )}\n

{label}

\n {children}\n

{help}

\n );\n}\n\nfunction Stat({ label, value, accent }: any) {\n return (\n

{label}

{value}

\n );\n}\n\nfunction Pill({ active, onClick, children }: any) {\n return (\n \n );\n}\n\nfunction Card({ title, children, right }: any) {\n return (\n

\n {title && (\n

{title}

\n {right}\n

\n )}\n

{children}

\n );\n}\n","createdAt",["D",1787578219985],"branchVersion","branchId","6e69a154-9fd0-4350-9b40-821b369159a3","id","cf2b2da1-d8e7-495b-9b80-5e85202b2429","d2c-dashboard",["D",1785139688485],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","D2C France e-commerce catalogue dashboard (Haier/Candy/Hoover)","imageUrl","9dc0f674-f4af-4f43-82f0-60e027cb7685","pinnedFileId","4da8d688-8992-11f1-bb40-1607ee4eb77e","pinPreview",true,"pinMigrated","screenshotVersion","likeCount",0,"orgId","019f70ba-a157-71ef-a091-fba3f4e32aa1","tursoName","cf2b2da1d8e7495b9b805e85202b2429","tursoHostname","cf2b2da1d8e7495b9b805e85202b2429-val-town.aws-us-east-2.turso.io","tursoMeasuredAt",["D",1789474853273],"pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_103":102,"_111":107},"org",{"_103":102,"_104":105,"_106":107,"_18":56,"_108":109,"_110":25},"branch",{"_13":90,"_32":91,"_44":43,"_92":45,"_93":94,"_95":96,"_97":51,"_98":45,"_99":66,"_100":46,"_101":102},"isOwned",3,"forecast-histogram-fix","projectId","forkedBranchId","36a2c3f0-828a-4176-bacb-c93668eda600","forkedBranchVersion",808,"forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","nicolasduchene","handle","displayName","Nicolas Duchene","avatar","https://lh3.googleusercontent.com/a/ACg8ocIezwTGgRq3xwm4fhgneTl1bk1gYVw-Hs87-EJKQoj3b5WO_2hc=s96-c","tier","free","canRestrictVals","profileImageUrl","branchName","","adoptOrgId",{"_44":45,"_32":46,"_39":606,"_48":-5,"_49":25,"_50":51,"_52":51,"_53":54,"_55":-5,"_18":56,"_57":58,"_59":60,"_61":-5,"_62":-5,"_63":64,"_65":66,"_67":68,"_69":70,"_71":607,"_73":-5,"_74":75,"_76":75,"_77":78,"_79":78,"_80":78,"_81":75,"_82":75,"_83":608,"_85":609,"_87":610,"_89":25},"tree",[236,251,279,280,281,282,283,284,285,286,287],"files",[236,237,238,239,240,241,242,243,244,245,246,247,248,249,250,251,252,253,254,255,256,257,258,259,260,261,262,263,264,265,266,267,268,269,270,271,272,273,274,275,276,277,278,279,280,281,282,283,284,285,286,287],"counts",{"_224":64,"_225":226,"_227":64,"_228":229,"_230":231},"contents",[142,143,144,145,146,147,148,149,150,151,152],"upstreamProject","upstreamBranch",{"_44":94,"_39":135,"_136":137,"_32":138,"_92":45,"_13":139,"_93":-5,"_95":-5,"_140":141,"_48":-5,"_89":25,"_50":51,"_101":102,"_65":66,"_100":46},"emptyEnvVars","flags",[],"hasHttpFiles","grants","orgs","customDomains",{},["D",1785139688511],"updatedAt",["D",1787661119567],"main",809,"lockfileKey","1787661119730-esm.town_v_nicolasduchene_d2c-dashboard@809-main",{"_44":219,"_32":220,"_30":221,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":222,"_13":64,"_34":217,"_36":-5,"_24":25,"_161":-5,"_27":223},{"_44":213,"_32":214,"_30":215,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":216,"_13":64,"_34":217,"_36":-5,"_24":25,"_161":-5,"_27":218},{"_44":207,"_32":208,"_30":209,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":210,"_13":64,"_34":35,"_36":211,"_24":25,"_161":-5,"_27":212},{"_44":201,"_32":202,"_30":203,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":204,"_13":64,"_34":35,"_36":205,"_24":25,"_161":-5,"_27":206},{"_44":195,"_32":196,"_30":197,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":198,"_13":64,"_34":35,"_36":199,"_24":25,"_161":-5,"_27":200},{"_44":188,"_32":189,"_30":190,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":191,"_13":64,"_34":192,"_36":193,"_24":25,"_161":-5,"_27":194},{"_44":182,"_32":183,"_30":184,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":185,"_13":64,"_34":35,"_36":186,"_24":25,"_161":-5,"_27":187},{"_44":176,"_32":177,"_30":178,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":179,"_13":64,"_34":35,"_36":180,"_24":25,"_161":-5,"_27":181},{"_44":169,"_32":170,"_30":171,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":172,"_13":64,"_34":173,"_36":174,"_24":25,"_161":-5,"_27":175},{"_44":163,"_32":164,"_30":165,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":166,"_13":64,"_34":35,"_36":167,"_24":25,"_161":-5,"_27":168},{"_44":153,"_32":154,"_30":155,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":159,"_13":64,"_34":35,"_36":160,"_24":25,"_161":-5,"_27":162},"a8debe5e-3605-460e-a905-bab92866634c","params.ts","01a033f6-8ab1-728d-ba87-379e18ed26d1","parentId",1,"latest",["D",1787578190526],"// Paramètres économiques — source unique de vérité.\n//\n// Ces neuf valeurs pilotaient auparavant trois copies indépendantes : les constantes de\n// `scripts/push_dashboard.py`, celles de `frontend/pricing.ts`, et l'onglet Parameters\n// du fichier Excel. Toute divergence produisait silencieusement deux breakevens\n// différents pour le même produit. Elles vivent désormais ici, dans le blob `params`,\n// et les trois consommateurs les lisent :\n// • l'onglet Pricing Simulator, via /api/params\n// • l'onglet Parameters, qui les édite via PUT /api/params\n// • le pipeline Python, qui les récupère avant de calculer breakeven et logistique\n//\n// Un changement n'a donc PAS d'effet rétroactif sur les colonnes Breakeven et Logistics\n// de l'onglet Products : celles-ci sont calculées par le pipeline et ne bougeront qu'à\n// son prochain passage. L'onglet Parameters le dit explicitement à l'utilisateur.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nexport interface ParamDef {\n key: keyof Params;\n label: string;\n help: string;\n /** \"pct\" est saisi en pourcentage mais stocké en fraction (8,6 % → 0.086). */\n unit: \"pct\" | \"eur\";\n min: number;\n max: number;\n /** Valeur d'origine, reprise de l'onglet Parameters du fichier Excel. */\n fallback: number;\n /** Cellule d'origine dans le classeur, pour retrouver la trace. */\n cell: string;\n}\n\nexport interface Params {\n OPEX: number;\n MKTG: number;\n VAT: number;\n WEEE_GROSS: number;\n TAKEBACK: number;\n RET_MDA: number;\n RET_SDA: number;\n SHIP_SDA: number;\n SHIP_MDA: number;\n}\n\n/** Bornes volontairement larges : elles n'existent que pour bloquer une faute de frappe\n * (un OPEX à 86 % au lieu de 8,6 %), pas pour imposer une politique commerciale. */\nexport const PARAM_DEFS: ParamDef[] = [\n {\n key: \"OPEX\", label: \"OPEX\", cell: \"B5\", unit: \"pct\", min: 0, max: 0.5, fallback: 0.086,\n help: \"Operating expenses, as a % of the excl. VAT price. Deducted to reach breakeven.\",\n },\n {\n key: \"MKTG\", label: \"Marketing\", cell: \"B15\", unit: \"pct\", min: 0, max: 0.5, fallback: 0.10,\n help: \"Marketing budget, as a % of the excl. VAT price.\",\n },\n {\n key: \"VAT\", label: \"VAT\", cell: \"B16\", unit: \"pct\", min: 0, max: 0.3, fallback: 0.20,\n help: \"VAT rate. Used to convert the site's incl. VAT prices into excl. VAT amounts.\",\n },\n {\n key: \"WEEE_GROSS\", label: \"WEEE take-back (gross)\", cell: \"B11\", unit: \"eur\", min: 0, max: 100, fallback: 10.584,\n help: \"Cost of taking back one old appliance, per unit taken back. Not to be confused with \"\n + \"the eco-participation, which comes from the official scale and is product-specific.\",\n },\n {\n key: \"TAKEBACK\", label: \"Take-back rate\", cell: \"B12\", unit: \"pct\", min: 0, max: 1, fallback: 0.15,\n help: \"Share of deliveries that involve a take-back. Multiplied by the cost above, it feeds \"\n + \"the CCV logistics cost.\",\n },\n {\n key: \"RET_MDA\", label: \"Return rate MDA\", cell: \"B13\", unit: \"pct\", min: 0, max: 1, fallback: 0.037,\n help: \"Large appliances, delivered by CCV. Increases the logistics cost.\",\n },\n {\n key: \"RET_SDA\", label: \"Return rate SDA\", cell: \"B14\", unit: \"pct\", min: 0, max: 1, fallback: 0.10,\n help: \"Small appliances, delivered by DPD.\",\n },\n {\n key: \"SHIP_SDA\", label: \"Delivery charged SDA\", cell: \"B17\", unit: \"eur\", min: 0, max: 200, fallback: 6.49,\n help: \"Delivery contribution paid by the customer, incl. VAT, on small appliances.\",\n },\n {\n key: \"SHIP_MDA\", label: \"Delivery charged MDA\", cell: \"B18\", unit: \"eur\", min: 0, max: 500, fallback: 27.31,\n help: \"Same, on large appliances.\",\n },\n];\n\nexport const DEFAULT_PARAMS: Params = Object.fromEntries(\n PARAM_DEFS.map((d) => [d.key, d.fallback]),\n) as unknown as Params;\n\nexport interface ParamsRecord {\n values: Params;\n updated_at: string | null;\n updated_by: string | null;\n /** Les 50 dernières modifications, plus récente d'abord. */\n history: {\n at: string;\n user: string;\n key: string;\n from: number;\n to: number;\n }[];\n}\n\nconst KEY = \"params\";\n\n/** Lecture. Toute clé absente retombe sur sa valeur d'origine : ajouter un paramètre\n * au code ne doit jamais casser une configuration déjà enregistrée. */\nexport async function getParams(): Promise {\n const stored = (await blob.getJSON(KEY)) as Partial | null;\n const values = { ...DEFAULT_PARAMS };\n for (const d of PARAM_DEFS) {\n const v = (stored?.values as any)?.[d.key];\n if (typeof v === \"number\" && isFinite(v)) values[d.key] = v;\n }\n return {\n values,\n updated_at: stored?.updated_at ?? null,\n updated_by: stored?.updated_by ?? null,\n history: Array.isArray(stored?.history) ? stored!.history! : [],\n };\n}\n\nexport interface SaveResult {\n ok: boolean;\n changed: string[];\n errors: string[];\n record: ParamsRecord;\n}\n\n/** Écriture. Seules les clés connues sont retenues, et chacune doit tenir dans ses\n * bornes — une valeur hors limites est refusée nommément plutôt qu'ignorée en silence. */\nexport async function saveParams(\n user: string,\n incoming: Record,\n): Promise {\n const current = await getParams();\n const next = { ...current.values };\n const changed: string[] = [];\n const errors: string[] = [];\n const now = new Date().toISOString();\n const entries: ParamsRecord[\"history\"] = [];\n\n for (const d of PARAM_DEFS) {\n if (!(d.key in incoming)) continue;\n const raw = incoming[d.key];\n const v = typeof raw === \"number\" ? raw : parseFloat(String(raw).replace(\",\", \".\"));\n if (!isFinite(v)) { errors.push(`${d.label}: value not readable`); continue; }\n if (v < d.min || v > d.max) {\n const fmt = (n: number) => d.unit === \"pct\" ? `${(n * 100).toFixed(1)}%` : `${n} €`;\n errors.push(`${d.label}: ${fmt(v)} out of range (${fmt(d.min)} – ${fmt(d.max)})`);\n continue;\n }\n // Comparaison arrondie : une saisie identique ne doit pas polluer l'historique\n if (Math.abs(v - current.values[d.key]) < 1e-9) continue;\n entries.push({ at: now, user, key: d.key, from: current.values[d.key], to: v });\n next[d.key] = v;\n changed.push(d.key);\n }\n\n if (errors.length > 0) return { ok: false, changed: [], errors, record: current };\n if (changed.length === 0) return { ok: true, changed: [], errors: [], record: current };\n\n const record: ParamsRecord = {\n values: next,\n updated_at: now,\n updated_by: user,\n history: [...entries, ...current.history].slice(0, 50),\n };\n await blob.setJSON(KEY, record);\n return { ok: true, changed, errors: [], record };\n}\n","label",[154],"1a9075f9-e84a-4979-b063-170a5614d15a","orders_sync.ts","01a033f6-8ab1-728d-ba87-3177c0a2f599",["D",1787578190526],"// Fetches the daily Matrixify ORDERS exports (stable URLs) and stores compact\n// order lines in blob \"orders_lines\" — powering the Business tab and the\n// shipping stats in the Logistic tab.\n//\n// Matrixify order exports repeat the order columns on every row, while the\n// line-item / fulfilment / refund blocks live on DIFFERENT rows of the same order:\n// • \"Line Item\" → the ordered item, positive quantity\n// • \"Fulfillment Line\" → the same SKU shipped, quantity and total in reverse, carrying\n// \"Fulfillment: Created At\" — this is what dates invoiced revenue\n// • \"Refund Line\" → the refunded items, in reverse, with their own \"Refund: Created At\"\n// • \"Discount\" → ORDER-LEVEL discount, on its own row. Critically, it is NOT\n// deducted from the Line Item totals, so it has to be allocated\n// back across the lines (pro rata) to get real net revenue.\n// • \"Shipping Line\" / \"Refund Shipping\" → delivery charged / refunded\n// • \"Transaction\" → payment rows, ignored\n// Older exports had no \"Line: Type\" column at all: fulfilments then appeared as untyped\n// negative-quantity rows. Both shapes are handled.\n// We group rows by order ID, then emit one record per line with\n// k = \"ord\" (ordered), \"ful\" (shipped) or \"ref\" (refunded), quantities always positive.\n//\n// ACCOUNTING RULE: a refund is booked on the REFUND date, not on the date of the\n// original order — a June order refunded in July is deducted from July. Refund lines\n// therefore carry their own date (rf) and are subtracted in that month.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { feeds } from \"./feeds.ts\";\n\n// Les URLs viennent des variables d'environnement — voir feeds.ts.\n\n// Historique conservé dans le blob, indépendamment de la profondeur de l'export.\nconst RETENTION_DAYS = 400;\n\n// ── FUSION INCRÉMENTALE ───────────────────────────────────────────────────────\n// Le sync ne remplace plus le blob : il le MET À JOUR commande par commande.\n// • une commande présente dans l'export → ses lignes remplacent celles stockées\n// • une commande absente → ses lignes stockées sont conservées\n// Cela permet de limiter l'export Matrixify à une fenêtre glissante (90 jours\n// suffisent : le délai commande → remboursement le plus long observé est de 50 j)\n// sans jamais perdre l'historique. Effet de bord bienvenu : si un flux tombe en\n// erreur, les commandes de cette marque restent intactes au lieu de disparaître.\n//\n// Limite assumée : une commande supprimée dans Shopify disparaît de l'export mais\n// reste stockée jusqu'à expiration de la rétention.\nconst orderKey = (l: OrderLine) => `${l.b}|${l.o}`;\n\nfunction parseCSV(text: string): string[][] {\n // Matrixify propose « CSV with included BOM » pour Excel. Le BOM se colle au premier\n // en-tête (\"ID\" au lieu de \"ID\"), ce qui fait échouer la recherche de colonne.\n if (text.charCodeAt(0) === 0xFEFF) text = text.slice(1);\n const rows: string[][] = [];\n let row: string[] = [], field = \"\", inQ = false;\n for (let i = 0; i < text.length; i++) {\n const c = text[i];\n if (inQ) {\n if (c === '\"') {\n if (text[i + 1] === '\"') { field += '\"'; i++; }\n else inQ = false;\n } else field += c;\n } else if (c === '\"') inQ = true;\n else if (c === \",\") { row.push(field); field = \"\"; }\n else if (c === \"\\n\" || c === \"\\r\") {\n if (c === \"\\r\" && text[i + 1] === \"\\n\") i++;\n row.push(field); field = \"\";\n if (row.length > 1 || row[0] !== \"\") rows.push(row);\n row = [];\n } else field += c;\n }\n if (field !== \"\" || row.length) { row.push(field); rows.push(row); }\n return rows;\n}\n\n/** Compact order line.\n * k=\"ord\" (ordered) | \"ful\" (shipped) | \"ref\" (refunded)\n * o=order name · b=brand · d=order date · f=fulfilment date · t=tags (marketplace)\n * rf=refund date — on k=\"ref\" lines this is THE date the amount is booked on\n * sku · q=qty (>0) · p=unit price BEFORE discount\n * n=net line total AFTER discount (when the export provides \"Line: Total\" /\n * \"Line: Discount\"; undefined otherwise)\n * ship=order-level shipping charged (marketplace fees appear here) */\nexport interface OrderLine {\n k: \"ord\" | \"ful\" | \"ref\";\n o: string; b: string; d: string; f?: string | null;\n t?: string | null; rf?: string | null;\n sku?: string | null; q: number; p: number; n?: number; ship?: number;\n}\n\n/** AVOIR SANS DÉTAIL ARTICLE.\n *\n * Un remboursement Shopify ne crée de `refund_line_items` que si une quantité est\n * saisie sur les lignes d'articles. Remboursé « au montant » depuis l'admin, il n'en\n * a aucun : Matrixify émet alors une ligne « Refund Line » sans SKU, sans quantité et\n * sans total — seule la date est renseignée. C'est le cas de la quasi-totalité des\n * avoirs Online Store ; les avoirs marketplace, créés par API ligne à ligne, sortent\n * complets et continuent de passer par le chemin normal (k = \"ref\").\n *\n * Ces avoirs étaient jusqu'ici éliminés en silence par le filtre « pas de quantité,\n * donc pas une ligne produit », ce qui laissait le CA d'une commande intégralement\n * remboursée compté à plein tarif. Sur juillet 2026 cela représentait 51 653 € et\n * expliquait l'essentiel de l'écart avec le BI.\n *\n * On les collecte donc à part plutôt que dans `lines` : ils ne sont rattachables à\n * aucune référence, et les injecter parmi les lignes produit fausserait marge et\n * classements. `amt` reste null tant que la colonne « Transaction: Amount » n'est pas\n * activée dans l'export — le nombre d'avoirs est alors connu, pas leur montant. */\nexport interface OrphanRefund {\n o: string; // nom de commande\n b: string; // marque\n d: string; // date de commande\n rf: string; // date de remboursement — c'est sur elle que le montant s'impute\n t?: string | null; // tags (canal)\n amt: number | null; // montant TTC remboursé, null si la colonne n'est pas exportée\n}\n\nexport async function syncOrders(): Promise<{\n ok: boolean;\n feeds: Record;\n lines: number;\n fulfilled: number;\n orphanRefunds: { count: number; withAmount: number };\n merge: { fromExport: number; keptFromHistory: number; ordersInExport: number; dropped: number };\n}> {\n const feedResults: Record = {};\n const all: OrderLine[] = [];\n const allOrphans: OrphanRefund[] = [];\n let anyOk = false, nReturns = 0;\n\n for (const { brand, url, varName } of feeds(\"ORDERS\")) {\n if (!url) { feedResults[brand] = `variable ${varName} non renseignée`; continue; }\n try {\n const res = await fetch(url, { redirect: \"follow\" });\n if (!res.ok) { feedResults[brand] = `HTTP ${res.status}`; continue; }\n const rows = parseCSV(await res.text());\n if (rows.length < 2) { feedResults[brand] = \"empty\"; anyOk = true; continue; }\n\n const header = rows[0].map((h) => h.trim().toLowerCase());\n const idx = (...names: string[]) => {\n for (const n of names) { const i = header.indexOf(n); if (i !== -1) return i; }\n return -1;\n };\n const iId = idx(\"id\");\n const iName = idx(\"name\");\n const iCreated = idx(\"created at\", \"processed at\");\n const iFulDate = idx(\"fulfillment: created at\", \"fulfilled at\");\n const iFulSt = idx(\"fulfillment: status\", \"fulfillment status\");\n const iSrc = idx(\"source name\", \"source\");\n const iTags = idx(\"tags\");\n const iRefund = idx(\"refund: created at\");\n const iSku = idx(\"line: sku\", \"lineitem sku\", \"line: variant sku\");\n const iQty = idx(\"line: quantity\", \"lineitem quantity\");\n const iPrice = idx(\"line: price\", \"lineitem price\");\n // Optional columns — present once the exports carry discounts (added 28/07/2026)\n const iDisc = idx(\"line: discount\");\n const iTotal = idx(\"line: total\");\n const iType = idx(\"line: type\");\n // Colonnes de transaction — facultatives. Elles seules portent le montant d'un\n // remboursement saisi sans détail article. Tant qu'elles ne sont pas cochées dans\n // l'export Matrixify, on sait compter ces avoirs mais pas les chiffrer.\n const iTxKind = idx(\"transaction: kind\");\n const iTxAmount = idx(\"transaction: amount\", \"transaction: shop currency amount\");\n if (iName === -1 || iCreated === -1) { feedResults[brand] = \"missing Name/Created At columns\"; continue; }\n\n // ── Garde-fou : un export appauvri ne doit JAMAIS écraser des données riches ──\n // Sans \"Line: Type\" on perd expéditions et avoirs ; sans \"Line: Total\" le CA\n // repasse en brut, remises non déduites. Plutôt que d'ingérer silencieusement\n // une version dégradée, on refuse le flux : la fusion conserve alors l'existant.\n const missing = [\n iType === -1 ? \"Line: Type\" : null,\n iTotal === -1 ? \"Line: Total\" : null,\n iDisc === -1 ? \"Line: Discount\" : null,\n ].filter(Boolean) as string[];\n if (missing.length > 0) {\n feedResults[brand] = `REFUSÉ — colonnes manquantes : ${missing.join(\", \")}`;\n console.warn(`[${brand}] export ignoré, colonnes manquantes : ${missing.join(\", \")}`);\n anyOk = true; // le flux a répondu : ce n'est pas une panne réseau\n continue;\n }\n\n // ── Pass 1: order-level attributes, merged across all rows of the order\n interface OrderMeta {\n name: string; d: string; src: string | null; t: string | null; rf: string | null;\n ship: number; // order-level shipping (marketplace delivery fees land here)\n disc: number; // order-level discount, POSITIVE amount to be spread over the lines\n // Montant des transactions de remboursement, indexé par date de remboursement.\n // Une même date peut porter plusieurs transactions (paiement scindé) : on cumule.\n refTx: Record;\n }\n const meta: Record = {};\n const keyOf = (row: string[]) =>\n (iId !== -1 ? (row[iId] ?? \"\").trim() : \"\") || (row[iName] ?? \"\").trim();\n\n for (let r = 1; r < rows.length; r++) {\n const row = rows[r];\n const key = keyOf(row);\n if (!key) continue;\n const m = meta[key] ?? {\n name: (row[iName] ?? \"\").trim(),\n d: (row[iCreated] ?? \"\").slice(0, 10),\n src: null, t: null, rf: null, ship: 0, disc: 0, refTx: {},\n };\n const rowType = iType !== -1 ? (row[iType] ?? \"\").trim().toLowerCase() : \"\";\n // Transaction de remboursement : une ligne \"Transaction\" qui porte une date de\n // remboursement. C'est le seul endroit où figure le montant rendu au client quand\n // l'avoir n'a pas de détail article. `Transaction: Kind` lève l'ambiguïté quand la\n // colonne est présente ; sinon la date de remboursement suffit à identifier la ligne.\n if (rowType === \"transaction\" && iRefund !== -1 && row[iRefund]?.trim() && iTxAmount !== -1) {\n const kind = iTxKind !== -1 ? (row[iTxKind] ?? \"\").trim().toLowerCase() : \"refund\";\n if (kind === \"refund\" || kind === \"\") {\n const amt = parseFloat(row[iTxAmount]);\n if (!isNaN(amt) && amt !== 0) {\n const day = row[iRefund].slice(0, 10);\n m.refTx[day] = Math.round(((m.refTx[day] ?? 0) + Math.abs(amt)) * 100) / 100;\n }\n }\n }\n // \"Shipping Line\" rows carry the delivery charged to the customer.\n if (rowType === \"shipping line\") {\n const amt = iTotal !== -1 ? parseFloat(row[iTotal]) : parseFloat(row[iPrice] ?? \"\");\n if (!isNaN(amt)) m.ship += amt;\n }\n // \"Discount\" rows carry an order-level promo code. Shopify exports it as a\n // negative total that is NOT reflected in the Line Item totals — so we bank it\n // here and spread it across the order's lines below.\n if (rowType === \"discount\") {\n const amt = iTotal !== -1 ? parseFloat(row[iTotal]) : NaN;\n if (!isNaN(amt)) m.disc += Math.abs(amt);\n }\n if (!m.name) m.name = (row[iName] ?? \"\").trim();\n if (!m.d) m.d = (row[iCreated] ?? \"\").slice(0, 10);\n if (iSrc !== -1 && !m.src && row[iSrc]?.trim()) m.src = row[iSrc].trim();\n if (iTags !== -1 && !m.t && row[iTags]?.trim()) m.t = row[iTags].trim();\n if (iRefund !== -1 && !m.rf && row[iRefund]?.trim()) m.rf = row[iRefund].slice(0, 10);\n meta[key] = m;\n }\n\n // ── Pass 2: emit one record per product line, grouped by order so that the\n // order-level discount can be allocated afterwards.\n const perOrder: Record = {};\n const orphansHere: OrphanRefund[] = [];\n for (let r = 1; r < rows.length; r++) {\n const row = rows[r];\n const key = keyOf(row);\n const m = meta[key];\n if (!m || !m.d) continue;\n\n const lType = iType !== -1 ? (row[iType] ?? \"\").trim().toLowerCase() : \"\";\n const fulSt = iFulSt !== -1 ? (row[iFulSt] ?? \"\").trim().toLowerCase() : \"\";\n const q = iQty !== -1 ? parseFloat(row[iQty]) : NaN;\n\n // ── Avoir sans détail article ────────────────────────────────────────\n // À intercepter AVANT le filtre de quantité ci-dessous : c'est précisément\n // parce qu'une ligne « Refund Line » n'a ni SKU ni quantité qu'elle tombait\n // dans le « ce n'est pas une ligne produit » et disparaissait sans trace.\n if (lType === \"refund line\" && (isNaN(q) || q === 0)) {\n const rd = iRefund !== -1 && row[iRefund]?.trim() ? row[iRefund].slice(0, 10) : null;\n if (rd) {\n orphansHere.push({\n o: m.name, b: brand, d: m.d, rf: rd, t: m.t,\n amt: m.refTx[rd] ?? null,\n });\n }\n continue;\n }\n\n if (isNaN(q) || q === 0) continue; // not a product row\n\n let kind: \"ord\" | \"ful\" | \"ref\";\n if (lType) {\n // Typed export: the line type is authoritative\n if (lType === \"line item\") kind = \"ord\";\n else if (lType === \"fulfillment line\") kind = \"ful\";\n else if (lType === \"refund line\") kind = \"ref\";\n else continue; // discount / shipping / transaction rows\n } else {\n // Legacy export with no \"Line: Type\": shipped items were negative-quantity rows\n kind = (q < 0 || fulSt === \"success\" || fulSt === \"fulfilled\") ? \"ful\" : \"ord\";\n }\n if (kind === \"ful\") nReturns++;\n\n // The refund date on the row itself is what the amount is booked on\n const rowRefDate = iRefund !== -1 && row[iRefund]?.trim() ? row[iRefund].slice(0, 10) : null;\n if (kind === \"ref\" && !rowRefDate) continue; // undatable refund → cannot be booked\n\n const p = iPrice !== -1 ? parseFloat(row[iPrice]) : NaN;\n // Line value: \"Line: Total\" is net of any LINE-level discount but not of the\n // order-level one, which is applied further down.\n let net: number | undefined;\n if (iTotal !== -1 && row[iTotal]?.trim()) {\n const t = parseFloat(row[iTotal]);\n if (!isNaN(t)) net = Math.abs(t);\n } else if (iDisc !== -1 && row[iDisc]?.trim() && !isNaN(p)) {\n const d = parseFloat(row[iDisc]);\n if (!isNaN(d)) net = Math.abs(Math.abs(q) * p + d); // discount is negative\n }\n\n (perOrder[key] ??= []).push({\n k: kind,\n o: m.name, b: brand, d: m.d,\n f: iFulDate !== -1 && row[iFulDate]?.trim() ? row[iFulDate].slice(0, 10) : null,\n t: m.t, rf: kind === \"ref\" ? rowRefDate : m.rf,\n sku: iSku !== -1 ? (row[iSku] ?? \"\").trim() || null : null,\n q: Math.abs(q), p: isNaN(p) ? 0 : p,\n ...(net !== undefined ? { n: Math.round(net * 100) / 100 } : {}),\n ...(m.ship ? { ship: Math.round(m.ship * 100) / 100 } : {}),\n });\n }\n\n // ── Pass 3: spread each order's discount over its lines, pro rata of line value.\n // Done per kind so that the ordered and the shipped view are both net of it.\n let count = 0, discApplied = 0;\n for (const [key, lines] of Object.entries(perOrder)) {\n const disc = meta[key]?.disc ?? 0;\n if (disc > 0) {\n for (const kind of [\"ord\", \"ful\", \"ref\"] as const) {\n const set = lines.filter((l) => l.k === kind && l.n != null);\n const gross = set.reduce((s, l) => s + (l.n ?? 0), 0);\n if (gross <= 0) continue;\n // A discount can never exceed the goods it applies to\n const applied = Math.min(disc, gross);\n for (const l of set) {\n l.n = Math.round(((l.n ?? 0) - applied * ((l.n ?? 0) / gross)) * 100) / 100;\n }\n if (kind === \"ord\") discApplied += applied;\n }\n }\n for (const l of lines) { all.push(l); count++; }\n }\n for (const o of orphansHere) allOrphans.push(o);\n feedResults[brand] = count;\n anyOk = true;\n } catch (e) {\n feedResults[brand] = String(e);\n }\n }\n\n const merge = { fromExport: all.length, keptFromHistory: 0, ordersInExport: 0, dropped: 0 };\n\n // Rien de exploitable : on ne touche pas au blob plutôt que d'écraser l'historique\n // avec du vide (un flux en panne ne doit jamais effacer des données).\n const orphanStats = { count: 0, withAmount: 0 };\n\n if (!anyOk || all.length === 0) {\n return { ok: anyOk, feeds: feedResults, lines: 0, fulfilled: nReturns, orphanRefunds: orphanStats, merge };\n }\n\n // ── Fusion avec l'historique déjà stocké ───────────────────────────────────\n const stored = (await blob.getJSON(\"orders_lines\")) as any;\n const previous: OrderLine[] = stored?.lines ?? [];\n const previousOrphans: OrphanRefund[] = stored?.orphan_refunds ?? [];\n // Les commandes rafraîchies sont repérées sur les DEUX collections : une commande\n // dont l'export ne contient plus que l'avoir doit quand même voir ses données\n // remplacées, sinon un avoir supprimé dans Shopify survivrait indéfiniment.\n const refreshed = new Set([\n ...all.map(orderKey),\n ...allOrphans.map((o) => `${o.b}|${o.o}`),\n ]);\n merge.ordersInExport = refreshed.size;\n\n // On conserve les lignes des commandes que l'export ne couvre plus\n const carried = previous.filter((l) => !refreshed.has(orderKey(l)));\n merge.keptFromHistory = carried.length;\n\n const combined = [...carried, ...all];\n\n // Rétention sur la date la plus tardive : une vieille commande remboursée\n // récemment doit survivre, son avoir étant imputé au mois courant.\n const cutoff = new Date(Date.now() - RETENTION_DAYS * 864e5).toISOString().slice(0, 10);\n const kept = combined.filter((l) => (l.k === \"ref\" ? (l.rf ?? l.d) : l.d) >= cutoff);\n merge.dropped = combined.length - kept.length;\n\n // Même traitement pour les avoirs sans détail : fusion incrémentale puis rétention\n // sur la date de remboursement, celle sur laquelle le montant s'impute.\n const carriedOrphans = previousOrphans.filter((o) => !refreshed.has(`${o.b}|${o.o}`));\n const keptOrphans = [...carriedOrphans, ...allOrphans].filter((o) => o.rf >= cutoff);\n orphanStats.count = keptOrphans.length;\n orphanStats.withAmount = keptOrphans.filter((o) => o.amt != null).length;\n\n await blob.setJSON(\"orders_lines\", {\n updated_at: new Date().toISOString(),\n feeds: feedResults,\n retention_days: RETENTION_DAYS,\n lines: kept,\n // Avoirs impossibles à rattacher à une référence — voir OrphanRefund.\n // Champ à part et non fusionné dans `lines` : les mêler aux lignes produit\n // fausserait marge, quantités et classements.\n orphan_refunds: keptOrphans,\n });\n return {\n ok: anyOk, feeds: feedResults, lines: kept.length, fulfilled: nReturns,\n orphanRefunds: orphanStats, merge,\n };\n}\n",[164],"d41e286a-4245-42c2-b202-16f85f834b81","main.tsx","01a033f6-8ab1-728d-ba87-2f0595be2e6d",["D",1787578190526],"http","// D2C France e-commerce dashboard — Haier / Candy / Hoover\n// Data pushed by local Python pipeline via POST /api/ingest (token-protected).\n// UI gated behind Val Town login + email allowlist.\nimport { Hono } from \"npm:hono\";\nimport { parseVal, serveFile } from \"https://esm.town/v/std/utils/index.ts\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { getOAuthUserData, oauthMiddleware } from \"https://esm.town/v/std/oauth/middleware.ts\";\nimport { syncCms } from \"./cms_sync.ts\";\nimport { addChanges, buildFeed, cancelChange, listChanges } from \"./changes.ts\";\nimport { getSkuHistory, getTrends } from \"./history.ts\";\nimport { getParams, PARAM_DEFS, saveParams } from \"./params.ts\";\nimport { getCampaign, listCampaigns, saveCampaign } from \"./campaigns.ts\";\n\nconst VAT = 0.20;\n\nconst app = new Hono();\napp.onError((err) => Promise.reject(err));\n\n// ── Access control ────────────────────────────────────────────────────────────\n// Two allowlists, both comma-separated email addresses held in env vars:\n// ALLOWED_EMAILS → full access: can read and queue price / status changes\n// VIEWER_EMAILS → read-only: sees everything, cannot write anything\n// The val owner (username below) always has full access. An address present in both\n// lists is treated as an editor.\n// Read-only is enforced HERE, on every write route — the frontend also hides the edit\n// controls, but that is cosmetic and must never be the only barrier.\nconst OWNER_USERNAME = \"nicolasduchene\";\n\nexport type Role = \"owner\" | \"editor\" | \"viewer\";\n\nfunction emailList(varName: string): string[] {\n return (Deno.env.get(varName) ?? \"\")\n .split(/[,;\\s]+/).map((e) => e.trim().toLowerCase()).filter(Boolean);\n}\n\ninterface Access {\n ok: boolean;\n user?: any;\n role?: Role;\n reason?: string;\n}\n\nasync function checkAccess(req: Request): Promise {\n const session = await getOAuthUserData(req);\n if (!session?.user) return { ok: false, reason: \"unauthenticated\" };\n const u = session.user;\n if (u.username === OWNER_USERNAME) return { ok: true, user: u, role: \"owner\" };\n const email = (u.email ?? \"\").toLowerCase();\n if (!email) return { ok: false, user: u, reason: \"not_allowed\" };\n if (emailList(\"ALLOWED_EMAILS\").includes(email)) return { ok: true, user: u, role: \"editor\" };\n if (emailList(\"VIEWER_EMAILS\").includes(email)) return { ok: true, user: u, role: \"viewer\" };\n return { ok: false, user: u, reason: \"not_allowed\" };\n}\n\n/** Identité inscrite dans la piste d'audit. On privilégie le pseudo Val Town et on ne\n * retient que la partie locale d'une adresse e-mail en dernier recours : la traçabilité\n * est conservée sans écrire d'adresse personnelle dans un journal partagé. */\nfunction auditName(u: any): string {\n if (u?.username) return String(u.username);\n const email = String(u?.email ?? \"\");\n return email ? email.split(\"@\")[0] : \"?\";\n}\n\n/** Constant-time string comparison — a plain `!==` leaks the length of the matching\n * prefix through response timing, which is enough to reconstruct a token. */\nfunction safeEqual(a: string, b: string): boolean {\n if (a.length !== b.length) return false;\n let diff = 0;\n for (let i = 0; i < a.length; i++) diff |= a.charCodeAt(i) ^ b.charCodeAt(i);\n return diff === 0;\n}\n\n/** Checks a shared secret sent by a machine caller (pipeline, Matrixify). */\nfunction tokenOk(provided: string | undefined | null, varName: string): boolean {\n const expected = Deno.env.get(varName);\n if (!expected || !provided) return false;\n return safeEqual(provided, expected);\n}\n\n/** Append-only trail of feed fetches. The feed URL carries its token in the path, so\n * it can leak through proxy logs or browser history — this makes any unexpected\n * consumption visible after the fact. Kept to the last 500 entries. */\nasync function logFeedAccess(entry: Record) {\n try {\n const prev = (await blob.getJSON(\"feed_access_log\")) as any[] | null;\n const list = Array.isArray(prev) ? prev : [];\n list.unshift({ at: new Date().toISOString(), ...entry });\n await blob.setJSON(\"feed_access_log\", list.slice(0, 500));\n } catch { /* le journal ne doit jamais faire échouer la livraison du flux */ }\n}\n\n/** Guard for write routes: resolves access and rejects read-only accounts. */\nasync function requireWrite(req: Request): Promise<{ access: Access; deny?: Response }> {\n const access = await checkAccess(req);\n if (!access.ok) {\n return { access, deny: Response.json({ error: \"unauthorized\" }, { status: 401 }) };\n }\n if (access.role === \"viewer\") {\n return {\n access,\n deny: Response.json(\n { error: \"read_only\", message: \"This account has read-only access to the dashboard.\" },\n { status: 403 },\n ),\n };\n }\n return { access };\n}\n\n// ── Auth-gated pages ──────────────────────────────────────────────────────────\napp.get(\"/\", async (c) => {\n const access = await checkAccess(c.req.raw);\n if (!access.ok) {\n // Show a landing page rather than auto-redirecting: an immediate redirect to\n // /auth/login would silently re-authenticate right after a logout.\n if (access.reason === \"unauthenticated\") return c.html(loginPage());\n return c.html(deniedPage(access.user), 403);\n }\n return serveFile(\"/frontend/index.html\");\n});\n\napp.get(\"/api/data\", async (c) => {\n const access = await checkAccess(c.req.raw);\n if (!access.ok) return c.json({ error: \"unauthorized\" }, 401);\n const data = await blob.getJSON(\"catalog\");\n if (!data) return c.json({ error: \"no_data\", message: \"No catalogue data pushed yet.\" }, 404);\n\n // Merge the daily CMS overlay (Matrixify feeds): fresher price/stock/status/image\n const overlay = await blob.getJSON(\"cms_overlay\");\n if (overlay?.items && Array.isArray(data.products)) {\n for (const p of data.products) {\n const o = overlay.items[String(p.code)];\n if (!o) continue;\n if (o.img && !p.img) p.img = o.img;\n if (o.ss) p.ss = o.ss;\n if (o.sstock != null) p.sstock = o.sstock;\n if (o.price != null && o.price > 0) {\n p.price = o.price;\n // Recompute price-derived metrics with the fresh price\n if (p.cost) {\n const ht = o.price / (1 + VAT);\n p.mcur = Math.round(((ht - p.cost - (p.deee ?? 0)) / ht) * 10000) / 10000;\n }\n if (p.comp && p.comp > 0) {\n p.gap = Math.round(((o.price - p.comp) / p.comp) * 10000) / 10000;\n p.pos = o.price <= p.comp * 0.99 ? \"competitive\" : o.price <= p.comp * 1.05 ? \"average\" : \"above\";\n }\n }\n }\n data.cms_synced_at = overlay.updated_at;\n }\n return c.json({\n ...data,\n // L'adresse e-mail n'est volontairement PAS renvoyée : elle servait uniquement\n // à l'affichage, et il s'agit souvent d'une adresse personnelle. Ne pas l'envoyer\n // du tout évite aussi qu'elle apparaisse dans les outils de développement.\n viewer: {\n username: access.user.username,\n role: access.role,\n canEdit: access.role !== \"viewer\",\n },\n });\n});\n\n// ── Product changes (team editing with audit trail) ───────────────────────────\napp.post(\"/api/changes\", async (c) => {\n const { access, deny } = await requireWrite(c.req.raw);\n if (deny) return deny;\n const body = await c.req.json();\n if (!Array.isArray(body?.changes)) return c.json({ error: \"bad_payload\" }, 400);\n const user = auditName(access.user);\n const result = await addChanges(user, body.changes);\n return c.json(result);\n});\n\napp.get(\"/api/changes\", async (c) => {\n const access = await checkAccess(c.req.raw);\n if (!access.ok) return c.json({ error: \"unauthorized\" }, 401);\n return c.json({ entries: await listChanges() });\n});\n\napp.post(\"/api/changes/cancel\", async (c) => {\n const { access, deny } = await requireWrite(c.req.raw);\n if (deny) return deny;\n const { id } = await c.req.json();\n const user = auditName(access.user);\n return c.json({ ok: await cancelChange(String(id), user) });\n});\n\n// ── Paramètres économiques ────────────────────────────────────────────────────\n// Lecture ouverte à tout compte autorisé (le simulateur en a besoin), écriture\n// réservée aux éditeurs par requireWrite — un accès en lecture seule reçoit un 403.\napp.get(\"/api/params\", async (c) => {\n const access = await checkAccess(c.req.raw);\n if (!access.ok) return c.json({ error: \"unauthorized\" }, 401);\n return c.json({ ...(await getParams()), defs: PARAM_DEFS });\n});\n\napp.put(\"/api/params\", async (c) => {\n const { access, deny } = await requireWrite(c.req.raw);\n if (deny) return deny;\n const body = await c.req.json();\n if (!body || typeof body !== \"object\") return c.json({ error: \"bad_payload\" }, 400);\n const result = await saveParams(auditName(access.user), body.values ?? body);\n return c.json(result, result.ok ? 200 : 400);\n});\n\n/** Lecture par le pipeline Python, qui n'a pas de session OAuth : même jeton que\n * l'ingestion. Les valeurs ne sont pas des secrets, mais rien ne justifie de les\n * exposer anonymement alors que le reste de l'API est fermé. */\napp.get(\"/api/params/pipeline\", async (c) => {\n if (!tokenOk(c.req.header(\"x-ingest-token\"), \"INGEST_TOKEN\")) {\n return c.json({ error: \"forbidden\" }, 403);\n }\n const { values, updated_at } = await getParams();\n return c.json({ values, updated_at });\n});\n\n// ── Forecast (historique Demand Planning) ─────────────────────────────────────\n// Le pipeline pousse ici l'extraction du fichier ValidationDP. On stocke brut : le\n// calcul de prévision vit dans le frontend, pour que changer une hypothèse ne demande\n// pas de relancer le pipeline.\napp.post(\"/api/ingest-forecast\", async (c) => {\n if (!tokenOk(c.req.header(\"x-ingest-token\"), \"INGEST_TOKEN\")) {\n return c.json({ error: \"forbidden\" }, 403);\n }\n const body = await c.req.json();\n if (!Array.isArray(body?.skus) || !Array.isArray(body?.months)) {\n return c.json({ error: \"bad_payload\" }, 400);\n }\n await blob.setJSON(\"forecast_dp\", body);\n return c.json({ ok: true, skus: body.skus.length, months: body.months.length });\n});\n\napp.get(\"/api/forecast\", async (c) => {\n const access = await checkAccess(c.req.raw);\n if (!access.ok) return c.json({ error: \"unauthorized\" }, 401);\n const data = await blob.getJSON(\"forecast_dp\");\n if (!data) {\n return c.json({\n error: \"no_forecast\",\n message: \"Fichier Demand Planning pas encore importé. Lancez push_dashboard.py.\",\n }, 404);\n }\n return c.json(data);\n});\n\n// ── Campagnes de forecast (arbitrages mémorisés) ──────────────────────────────\n// Lecture pour tous les comptes autorisés, écriture réservée aux éditeurs : un compte\n// en lecture seule peut consulter les arbitrages de l'équipe, pas les modifier.\napp.get(\"/api/forecast/campaigns\", async (c) => {\n const access = await checkAccess(c.req.raw);\n if (!access.ok) return c.json({ error: \"unauthorized\" }, 401);\n const id = c.req.query(\"id\");\n return c.json({\n campaigns: await listCampaigns(),\n current: id ? await getCampaign(id) : null,\n });\n});\n\napp.put(\"/api/forecast/campaigns\", async (c) => {\n const { access, deny } = await requireWrite(c.req.raw);\n if (deny) return deny;\n const body = await c.req.json();\n if (!body?.id || !Array.isArray(body?.months)) {\n return c.json({ error: \"bad_payload\" }, 400);\n }\n return c.json(await saveCampaign(auditName(access.user), body));\n});\n\n// ── Trends (daily snapshots) ──────────────────────────────────────────────────\napp.get(\"/api/trends\", async (c) => {\n const access = await checkAccess(c.req.raw);\n if (!access.ok) return c.json({ error: \"unauthorized\" }, 401);\n const days = Math.min(parseInt(c.req.query(\"days\") ?? \"30\") || 30, 180);\n return c.json({ series: await getTrends(days) });\n});\n\napp.get(\"/api/history/:sku\", async (c) => {\n const access = await checkAccess(c.req.raw);\n if (!access.ok) return c.json({ error: \"unauthorized\" }, 401);\n const sku = parseInt(c.req.param(\"sku\"));\n if (!sku) return c.json({ error: \"bad_sku\" }, 400);\n return c.json({ history: await getSkuHistory(sku) });\n});\n\n// ── Amazon Vendor ─────────────────────────────────────────────────────────────\n// Blob SÉPARÉ de `catalog` et `orders_lines`, délibérément. Amazon Vendor est du 1P :\n// Amazon nous achète puis revend. Le « sell-in » y est ce qu'Amazon nous paie, alors\n// que le CA du D2C est ce que paie le client final. Additionner les deux produirait un\n// chiffre sans signification — la séparation des stockages rend ce mélange impossible\n// par construction, plutôt que de compter sur la vigilance de celui qui écrira l'écran.\napp.post(\"/api/ingest-amazon\", async (c) => {\n if (!tokenOk(c.req.header(\"x-ingest-token\"), \"INGEST_TOKEN\")) {\n return c.json({ error: \"forbidden\" }, 403);\n }\n const body = await c.req.json();\n if (!Array.isArray(body?.products)) return c.json({ error: \"bad_payload\" }, 400);\n await blob.setJSON(\"amazon\", body);\n const withSellout = body.products.filter((p: any) => p.sellout).length;\n return c.json({\n ok: true,\n products: body.products.length,\n withSellout,\n periods: Object.keys(body.sellout_periods ?? {}),\n });\n});\n\napp.get(\"/api/amazon\", async (c) => {\n const access = await checkAccess(c.req.raw);\n if (!access.ok) return c.json({ error: \"unauthorized\" }, 401);\n const data = await blob.getJSON(\"amazon\");\n if (!data) {\n return c.json({\n error: \"no_amazon\",\n message: \"Aucune donnée Amazon Vendor. Lancez push_amazon_dashboard.py.\",\n }, 404);\n }\n return c.json(data);\n});\n\n// ── Business data (order lines from Matrixify orders exports) ─────────────────\napp.get(\"/api/business\", async (c) => {\n const access = await checkAccess(c.req.raw);\n if (!access.ok) return c.json({ error: \"unauthorized\" }, 401);\n const data = await blob.getJSON(\"orders_lines\");\n if (!data) return c.json({ error: \"no_orders\", message: \"Orders export not configured yet.\" }, 404);\n return c.json(data);\n});\n\napp.post(\"/api/sync-orders\", async (c) => {\n if (!tokenOk(c.req.header(\"x-ingest-token\"), \"INGEST_TOKEN\")) {\n return c.json({ error: \"forbidden\" }, 403);\n }\n const { syncOrders } = await import(\"./orders_sync.ts\");\n return c.json(await syncOrders());\n});\n\n// ── Matrixify import feeds (one per store, token in path) ─────────────────────\n// GET /feeds//.csv → pending changes as Matrixify import CSV\n// Fetching the feed marks the included changes as \"sent\".\n// Add ?peek=1 to preview without consuming.\napp.get(\"/feeds/:token/:file\", async (c) => {\n const file = c.req.param(\"file\").toLowerCase();\n const brand = file.replace(\".csv\", \"\").toUpperCase();\n // 404 on a bad token, not 403: a \"forbidden\" confirms the endpoint exists and that\n // only the token is missing, which invites guessing. An unknown path says nothing.\n if (!tokenOk(c.req.param(\"token\"), \"FEED_TOKEN\")) {\n await logFeedAccess({\n brand, ok: false, reason: \"bad_token\",\n ip: c.req.header(\"x-forwarded-for\") ?? null,\n ua: c.req.header(\"user-agent\") ?? null,\n });\n return c.notFound();\n }\n if (![\"HOOVER\", \"HAIER\", \"CANDY\"].includes(brand)) return c.notFound();\n\n const consume = c.req.query(\"peek\") !== \"1\";\n const csv = await buildFeed(brand, consume);\n // Count the rows delivered so an unexpected fetch can be spotted and replayed\n const rows = Math.max(0, csv.trim().split(\"\\n\").length - 1);\n await logFeedAccess({\n brand, ok: true, consumed: consume, rows,\n ip: c.req.header(\"x-forwarded-for\") ?? null,\n ua: c.req.header(\"user-agent\") ?? null,\n });\n return new Response(csv, {\n headers: {\n \"Content-Type\": \"text/csv; charset=utf-8\",\n \"Cache-Control\": \"no-store\",\n \"X-Robots-Tag\": \"noindex, nofollow\",\n },\n });\n});\n\n/** Feed access trail — owner and editors only. */\napp.get(\"/api/feed-log\", async (c) => {\n const access = await checkAccess(c.req.raw);\n if (!access.ok) return c.json({ error: \"unauthorized\" }, 401);\n return c.json({ entries: (await blob.getJSON(\"feed_access_log\")) ?? [] });\n});\n\n// Manual CMS sync trigger (same token as ingest)\napp.post(\"/api/sync-cms\", async (c) => {\n if (!tokenOk(c.req.header(\"x-ingest-token\"), \"INGEST_TOKEN\")) {\n return c.json({ error: \"forbidden\" }, 403);\n }\n const result = await syncCms();\n const { verifyChanges } = await import(\"./changes.ts\");\n const verified = await verifyChanges();\n return c.json({ ...result, verified });\n});\n\n// ── Data ingestion (called by the Python pipeline) ────────────────────────────\napp.post(\"/api/ingest\", async (c) => {\n if (!tokenOk(c.req.header(\"x-ingest-token\"), \"INGEST_TOKEN\")) {\n return c.json({ error: \"forbidden\" }, 403);\n }\n const body = await c.req.json();\n if (!Array.isArray(body?.products)) return c.json({ error: \"bad_payload\" }, 400);\n await blob.setJSON(\"catalog\", body);\n return c.json({ ok: true, products: body.products.length, generated_at: body.generated_at ?? null });\n});\n\n// ── Static frontend + source ──────────────────────────────────────────────────\n// Favicon at the root: browsers request /favicon.ico on their own, and the login\n// and denied pages are served before any frontend asset is reachable.\napp.get(\"/favicon.svg\", () => serveFile(\"/frontend/favicon.svg\"));\napp.get(\"/favicon.ico\", () => serveFile(\"/frontend/favicon.svg\"));\napp.get(\"/frontend/**/*\", (c) => serveFile(c.req.path));\napp.get(\"/source\", (c) => c.redirect(parseVal().links.self.val));\n\nfunction loginPage(): string {\n return `\nD2C France — Dashboard\n\n\n\n

📊

D2C France — E-commerce Dashboard

Haier · Candy · Hoover — you are signed out.

\nLog in\n

`;\n}\n\nfunction deniedPage(user: any): string {\n return `Access denied\n\n\n\n

🔒

Access restricted

Signed in as ${user?.username ?? \"?\"}.

This account is not on the allowed list. Ask the dashboard owner to add your email.

\n\n

`;\n}\n\nexport default oauthMiddleware(app.fetch);\n",[170],"dba780fe-e2f1-482a-98c3-33d35d9ef31a","history.ts","01a033f6-8ab0-7225-822f-8796b824b5e6",["D",1787578190526],"// Daily per-SKU snapshots (price, cumulative sales, stock) in SQLite,\n// powering the trend charts. One row per (date, sku); re-runs overwrite.\nimport { sqlite } from \"https://esm.town/v/std/sqlite/main.ts\";\n\nconst T = \"snapshots\";\n\nexport async function ensureTable() {\n await sqlite.execute(`CREATE TABLE IF NOT EXISTS ${T} (\n date TEXT NOT NULL,\n sku INTEGER NOT NULL,\n brand TEXT,\n grp TEXT,\n price REAL,\n sales INTEGER,\n stock INTEGER,\n PRIMARY KEY (date, sku)\n )`);\n}\n\n/** Record today's snapshot from the catalog blob (+ overlay already merged upstream). */\nexport async function recordSnapshot(products: any[]) {\n await ensureTable();\n const date = new Date().toISOString().slice(0, 10);\n // Batch inserts (100 rows per statement)\n for (let i = 0; i < products.length; i += 100) {\n const chunk = products.slice(i, i + 100);\n const values = chunk.map(() => \"(?,?,?,?,?,?,?)\").join(\",\");\n const args: any[] = [];\n for (const p of chunk) {\n args.push(date, p.code, p.brand ?? null, p.grp ?? null,\n p.price ?? null, p.sales ?? null, p.s48 ?? null);\n }\n await sqlite.execute({\n sql: `INSERT OR REPLACE INTO ${T} (date, sku, brand, grp, price, sales, stock) VALUES ${values}`,\n args,\n });\n }\n // Retention: keep 180 days\n await sqlite.execute({\n sql: `DELETE FROM ${T} WHERE date < ?`,\n args: [new Date(Date.now() - 180 * 864e5).toISOString().slice(0, 10)],\n });\n}\n\n/** Aggregated daily series by category group (sales cumulative + revenue estimate). */\nexport async function getTrends(days = 30) {\n await ensureTable();\n const since = new Date(Date.now() - days * 864e5).toISOString().slice(0, 10);\n const res = await sqlite.execute({\n sql: `SELECT date, grp,\n SUM(sales) AS sales,\n SUM(CASE WHEN price IS NOT NULL THEN price * sales ELSE 0 END) AS revenue\n FROM ${T} WHERE date >= ? GROUP BY date, grp ORDER BY date`,\n args: [since],\n });\n return res.rows.map((r: any) => ({ date: r[0], grp: r[1], sales: r[2], revenue: r[3] }));\n}\n\n/** Per-SKU price/sales history (for future product detail view). */\nexport async function getSkuHistory(sku: number, days = 90) {\n await ensureTable();\n const since = new Date(Date.now() - days * 864e5).toISOString().slice(0, 10);\n const res = await sqlite.execute({\n sql: `SELECT date, price, sales, stock FROM ${T} WHERE sku = ? AND date >= ? ORDER BY date`,\n args: [sku, since],\n });\n return res.rows.map((r: any) => ({ date: r[0], price: r[1], sales: r[2], stock: r[3] }));\n}\n",[177],"19222cd3-b5fb-4432-9142-b0c08f731463","feeds.ts","01a033f6-8ab0-7225-822f-833d8e782c88",["D",1787578190526],"// Résolution des URLs d'export Matrixify.\n//\n// Ces URLs ne demandent aucune authentification : quiconque les possède peut\n// télécharger le carnet de commandes ou le catalogue. Elles ne doivent donc jamais\n// figurer dans le code, qui est lisible publiquement.\n// Elles vivent dans les variables d'environnement du val, chiffrées et invisibles\n// depuis la page publique du projet :\n//\n// FEED_ORDERS_{HOOVER|HAIER|CANDY} export « Orders »\n// FEED_CATALOG_{HOOVER|HAIER|CANDY} export « Products » (prix, stock, statut)\n//\n// Pour régénérer une URL compromise : la recréer côté Matrixify, puis mettre à jour\n// la variable — aucune modification de code n'est nécessaire.\n\nexport const BRANDS = [\"HOOVER\", \"HAIER\", \"CANDY\"] as const;\nexport type Brand = typeof BRANDS[number];\n\nexport interface Feed {\n brand: Brand;\n url: string | null;\n /** Nom de la variable attendue, pour un message d'erreur exploitable. */\n varName: string;\n}\n\n/** Les trois flux d'un type donné, dans l'ordre des marques. */\nexport function feeds(kind: \"ORDERS\" | \"CATALOG\"): Feed[] {\n return BRANDS.map((brand) => {\n const varName = `FEED_${kind}_${brand}`;\n const url = (Deno.env.get(varName) ?? \"\").trim();\n return { brand, varName, url: url || null };\n });\n}\n",[183],"b2ebfda8-e8c6-4121-86e7-66664a149e71","cron.ts","01a033f6-8ab0-7225-822f-7ee06874beb0",["D",1787578190526],"interval","// Synchronisation des flux Matrixify (produits + commandes) + instantané quotidien.\n//\n// Planification : toutes les 3 heures (`0 */3 * * *`). Val Town exécute les crons en\n// UTC : une heure fixe dériverait de deux heures entre l'hiver et l'été à Paris, et\n// c'est exactement ce qui avait fait croire à un import « de 7h » alors qu'il tournait\n// à 9h. Une cadence régulière supprime le problème et garantit de récupérer chaque\n// export Matrixify (programmé toutes les 6 h) au plus tard 3 heures après sa création.\n//\n// Fraîcheur des données : les exports Matrixify sont filtrés sur « Created At –\n// Last: 2160 hours » (90 jours). L'unité compte — en JOURS, Matrixify ne retient que\n// les journées calendaires révolues et les commandes du jour même n'apparaissent\n// jamais ; en HEURES, la fenêtre s'arrête à l'heure ronde précédente et la journée en\n// cours est bien couverte. Ne pas repasser ce filtre en jours (corrigé le 31/07/2026).\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { syncCms } from \"./cms_sync.ts\";\nimport { syncOrders } from \"./orders_sync.ts\";\nimport { recordSnapshot } from \"./history.ts\";\nimport { verifyChanges } from \"./changes.ts\";\n\nexport default async function () {\n const cms = await syncCms();\n console.log(\"CMS sync:\", JSON.stringify(cms));\n\n // Confirm queued changes actually landed in Shopify\n const check = await verifyChanges();\n console.log(\"Change verification:\", JSON.stringify(check));\n\n const orders = await syncOrders();\n console.log(\"Orders sync:\", JSON.stringify(orders));\n\n // Daily snapshot (catalog with fresh overlay prices) → trends\n try {\n const catalog = await blob.getJSON(\"catalog\");\n const overlay = await blob.getJSON(\"cms_overlay\");\n const products = (catalog?.products ?? []).map((p: any) => {\n const o = overlay?.items?.[String(p.code)];\n return o?.price ? { ...p, price: o.price } : p;\n });\n if (products.length) {\n await recordSnapshot(products);\n console.log(`Snapshot: ${products.length} products recorded`);\n }\n } catch (e) {\n console.error(\"Snapshot failed:\", e);\n }\n}\n",[189],"db16f306-3d85-402d-8a22-08ca41b8f51a","cms_sync.ts","01a033f6-8ab0-7225-822f-7b8858a0a7c0",["D",1787578190526],"// Fetches the daily Matrixify CSV exports (stable URLs) and stores a per-SKU\n// overlay (price / site stock / status / image) in blob \"cms_overlay\".\n// The overlay is merged into the catalog by /api/data at read time.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { feeds } from \"./feeds.ts\";\n\n// Les URLs viennent des variables d'environnement — voir feeds.ts.\n\n/** Minimal CSV parser handling quoted fields, commas and newlines in quotes. */\nfunction parseCSV(text: string): string[][] {\n const rows: string[][] = [];\n let row: string[] = [], field = \"\", inQ = false;\n for (let i = 0; i < text.length; i++) {\n const c = text[i];\n if (inQ) {\n if (c === '\"') {\n if (text[i + 1] === '\"') { field += '\"'; i++; }\n else inQ = false;\n } else field += c;\n } else if (c === '\"') inQ = true;\n else if (c === \",\") { row.push(field); field = \"\"; }\n else if (c === \"\\n\" || c === \"\\r\") {\n if (c === \"\\r\" && text[i + 1] === \"\\n\") i++;\n row.push(field); field = \"\";\n if (row.length > 1 || row[0] !== \"\") rows.push(row);\n row = [];\n } else field += c;\n }\n if (field !== \"\" || row.length) { row.push(field); rows.push(row); }\n return rows;\n}\n\nfunction statusMap(raw: string): string {\n const s = (raw ?? \"\").trim().toLowerCase();\n if (s === \"active\") return \"Active\";\n if ([\"draft\", \"archived\", \"inactive\"].includes(s)) return \"Inactive\";\n return \"Not listed\";\n}\n\nexport interface OverlayItem {\n brand: string; price?: number | null; sstock?: number | null;\n ss?: string; img?: string | null; title?: string | null; handle?: string | null;\n}\n\nexport async function syncCms(): Promise<{ ok: boolean; feeds: Record; total: number }> {\n const items: Record = {};\n const feedResults: Record = {};\n\n for (const { brand, url, varName } of feeds(\"CATALOG\")) {\n if (!url) { feedResults[brand] = `variable ${varName} non renseignée`; continue; }\n try {\n const res = await fetch(url, { redirect: \"follow\" });\n if (!res.ok) { feedResults[brand] = `HTTP ${res.status}`; continue; }\n const text = await res.text();\n const rows = parseCSV(text);\n if (rows.length < 2) { feedResults[brand] = \"empty file\"; continue; }\n\n const header = rows[0].map((h) => h.trim().toLowerCase());\n const idx = (...names: string[]) => {\n for (const n of names) { const i = header.indexOf(n); if (i !== -1) return i; }\n return -1;\n };\n const iSku = idx(\"variant sku\");\n const iPrice = idx(\"variant price\");\n const iQty = idx(\"variant inventory qty\");\n const iStatus = idx(\"status\");\n const iImg = idx(\"image src\");\n const iTitle = idx(\"title\");\n const iHandle = idx(\"handle\");\n if (iSku === -1) { feedResults[brand] = \"no 'Variant SKU' column\"; continue; }\n\n let count = 0;\n for (let r = 1; r < rows.length; r++) {\n const row = rows[r];\n const sku = (row[iSku] ?? \"\").trim();\n if (!sku || !/^\\d+$/.test(sku)) continue;\n const existing = items[sku] ?? { brand } as OverlayItem;\n const price = iPrice !== -1 ? parseFloat(row[iPrice]) : NaN;\n if (!isNaN(price) && price > 0 && existing.price == null) existing.price = price;\n const qty = iQty !== -1 ? parseInt(row[iQty]) : NaN;\n if (!isNaN(qty) && existing.sstock == null) existing.sstock = qty;\n if (iStatus !== -1 && row[iStatus]?.trim() && !existing.ss) existing.ss = statusMap(row[iStatus]);\n const img = iImg !== -1 ? (row[iImg] ?? \"\").trim() : \"\";\n if (img.startsWith(\"http\") && !existing.img) existing.img = img;\n if (iTitle !== -1 && row[iTitle]?.trim() && !existing.title) existing.title = row[iTitle].trim();\n if (iHandle !== -1 && row[iHandle]?.trim() && !existing.handle) existing.handle = row[iHandle].trim();\n if (!items[sku]) { items[sku] = existing; count++; }\n }\n feedResults[brand] = count;\n } catch (e) {\n feedResults[brand] = String(e);\n }\n }\n\n const total = Object.keys(items).length;\n if (total > 0) {\n await blob.setJSON(\"cms_overlay\", { updated_at: new Date().toISOString(), feeds: feedResults, items });\n }\n return { ok: total > 0, feeds: feedResults, total };\n}\n",[196],"96b0d374-2e8e-44a2-9f9d-1194457d1c3a","changes.ts","01a033f6-8ab0-7225-822f-77d48f2855b8",["D",1787578190526],"// Pending product changes (price / status) with audit trail, and the\n// Matrixify import feeds that deliver them to Shopify.\n// Storage: blob \"changes_log\" = { entries: ChangeEntry[] } (newest first, capped).\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nexport interface ChangeEntry {\n id: string;\n at: string; // ISO timestamp (created)\n user: string; // pseudo Val Town (voir auditName dans main.tsx) — jamais\n // une adresse complète pour les entrées récentes ; les\n // anciennes sont masquées à l'affichage par displayUser()\n code: number; // SKU\n brand: string;\n handle?: string | null; // Shopify handle (needed for Matrixify import)\n name?: string | null;\n field: \"price\" | \"status\";\n oldValue: string | number | null;\n newValue: string | number;\n status: \"pending\" | \"sent\" | \"applied\" | \"cancelled\";\n sentAt?: string | null;\n appliedAt?: string | null; // confirmed live in Shopify by the CMS sync\n applyAt?: string | null; // ISO — do not deliver to Matrixify before this time\n}\n\n/** Sent changes keep being served this long, so a Matrixify retry or a\n * double-fetch still receives them. All operations are idempotent. */\nconst GRACE_HOURS = 12;\n\nconst MAX_ENTRIES = 2000;\n\nasync function load(): Promise {\n const data = await blob.getJSON(\"changes_log\");\n return data?.entries ?? [];\n}\n\nasync function save(entries: ChangeEntry[]) {\n await blob.setJSON(\"changes_log\", { entries: entries.slice(0, MAX_ENTRIES) });\n}\n\nexport async function addChanges(\n user: string,\n changes: { code: number; field: \"price\" | \"status\"; newValue: string | number; oldValue?: string | number | null; applyAt?: string | null }[],\n): Promise<{ ok: boolean; added: number; errors: string[] }> {\n const errors: string[] = [];\n const entries = await load();\n\n // Resolve brand/handle/name from overlay + catalog\n const overlay = await blob.getJSON(\"cms_overlay\");\n const catalog = await blob.getJSON(\"catalog\");\n const catByCode: Record = {};\n for (const p of catalog?.products ?? []) catByCode[String(p.code)] = p;\n\n let added = 0;\n for (const ch of changes) {\n const codeStr = String(ch.code);\n const o = overlay?.items?.[codeStr];\n const p = catByCode[codeStr];\n if (!o?.handle) {\n errors.push(`${codeStr}: Shopify handle unknown (add 'Handle' column to Matrixify exports)`);\n continue;\n }\n if (ch.field === \"price\") {\n const v = Number(ch.newValue);\n if (!isFinite(v) || v <= 0) { errors.push(`${codeStr}: invalid price`); continue; }\n }\n if (ch.field === \"status\" && ![\"Active\", \"Draft\"].includes(String(ch.newValue))) {\n errors.push(`${codeStr}: status must be Active or Draft`); continue;\n }\n // Supersede any pending change on the same SKU+field\n for (const e of entries) {\n if (e.code === ch.code && e.field === ch.field && e.status === \"pending\") e.status = \"cancelled\";\n }\n entries.unshift({\n id: crypto.randomUUID().slice(0, 8),\n at: new Date().toISOString(),\n user,\n code: ch.code,\n brand: o.brand ?? p?.brand ?? \"?\",\n handle: o.handle,\n name: p?.name ?? o.title ?? null,\n field: ch.field,\n oldValue: ch.oldValue ?? (ch.field === \"price\" ? (o.price ?? p?.price ?? null) : (o.ss ?? p?.ss ?? null)),\n newValue: ch.newValue,\n status: \"pending\",\n applyAt: ch.applyAt ?? null,\n });\n added++;\n }\n await save(entries);\n return { ok: errors.length === 0, added, errors };\n}\n\nexport async function listChanges(limit = 200): Promise {\n return (await load()).slice(0, limit);\n}\n\n/** Confirm against reality: compare sent changes with the live CMS overlay.\n * Called after each CMS sync. Marks entries \"applied\" once Shopify agrees. */\nexport async function verifyChanges(): Promise<{ checked: number; applied: number }> {\n const entries = await load();\n const overlay = await blob.getJSON(\"cms_overlay\");\n const items = overlay?.items ?? {};\n const syncedAt = overlay?.updated_at ?? \"\";\n let checked = 0, applied = 0;\n\n for (const e of entries) {\n if (e.status !== \"sent\" || !e.sentAt) continue;\n if (syncedAt <= e.sentAt) continue; // site data older than the change\n const o = items[String(e.code)];\n if (!o) continue;\n checked++;\n const ok = e.field === \"price\"\n ? (o.price != null && Math.abs(Number(o.price) - Number(e.newValue)) < 0.02)\n : (String(o.ss ?? \"\") === (String(e.newValue) === \"Draft\" ? \"Inactive\" : \"Active\"));\n if (ok) { e.status = \"applied\"; e.appliedAt = syncedAt; applied++; }\n }\n if (applied > 0) await save(entries);\n return { checked, applied };\n}\n\nexport async function cancelChange(id: string, user: string): Promise {\n const entries = await load();\n const e = entries.find((x) => x.id === id && x.status === \"pending\");\n if (!e) return false;\n e.status = \"cancelled\";\n e.user = `${e.user} (cancelled by ${user})`;\n await save(entries);\n return true;\n}\n\n/** Build the Matrixify import CSV for one brand and mark those changes as sent.\n * Grouped per product (handle): one row per product with the fields to update. */\nexport async function buildFeed(brand: string, consume = true): Promise {\n const entries = await load();\n const now = Date.now();\n const nowISO = new Date(now).toISOString();\n const graceISO = new Date(now - GRACE_HOURS * 3600_000).toISOString();\n const sameBrand = (e: ChangeEntry) => e.brand.toUpperCase() === brand.toUpperCase();\n const due = (e: ChangeEntry) => !e.applyAt || e.applyAt <= nowISO;\n\n // Serve pending changes AND recently-sent ones (idempotent replay window),\n // so a retry or a preview fetch by Matrixify cannot silently swallow them.\n const pending = entries.filter((e) => sameBrand(e) && due(e) && (\n e.status === \"pending\" ||\n (e.status === \"sent\" && (e.sentAt ?? \"\") >= graceISO)\n ));\n\n // Group by handle\n const byHandle: Record = {};\n for (const e of pending) {\n const g = byHandle[e.handle!] ?? { sku: String(e.code) };\n if (e.field === \"price\") g.price = Number(e.newValue);\n if (e.field === \"status\") g.status = String(e.newValue);\n byHandle[e.handle!] = g;\n }\n\n const esc = (v: string | number | undefined | null) => {\n const s = v == null ? \"\" : String(v);\n return /[\",\\n]/.test(s) ? `\"${s.replace(/\"/g, '\"\"')}\"` : s;\n };\n\n // Matrixify rule: identifying a variant (Variant SKU + Variant Command) forces a\n // price — \"Price can't be blank\". So the variant columns are only filled when a\n // price change is actually queued; a status-only row carries product columns alone.\n // An empty Status is simply ignored by Matrixify, so price-only rows are safe.\n const header = \"Handle,Command,Variant SKU,Variant Command,Variant Price,Status\";\n const lines = [header];\n for (const [handle, g] of Object.entries(byHandle)) {\n const hasPrice = g.price != null;\n lines.push([\n esc(handle), \"UPDATE\",\n hasPrice ? esc(g.sku) : \"\",\n hasPrice ? \"MERGE\" : \"\",\n hasPrice ? String(g.price) : \"\",\n g.status ?? \"\",\n ].join(\",\"));\n }\n\n if (consume && pending.length > 0) {\n for (const e of pending) {\n if (e.status === \"pending\") { e.status = \"sent\"; e.sentAt = nowISO; }\n }\n await save(entries);\n }\n return lines.join(\"\\n\") + \"\\n\";\n}\n",[202],"86baf103-0ef8-4037-8b78-5fcc70fcd5f8","campaigns.ts","01a033f6-8ab0-7225-822f-70c781ed5351",["D",1787578190526],"// Campagnes de forecast — mémorisation des arbitrages, une campagne par fichier DP.\n//\n// POURQUOI UNE CAMPAGNE PLUTÔT QU'UN SIMPLE ENREGISTREMENT\n// L'équipe demand planning fait évoluer son fichier régulièrement, et rien ne garantit\n// qu'elle retienne la prévision proposée. Si les corrections saisies au mois M se\n// réappliquaient silencieusement au fichier du mois M+1, l'outil afficherait comme\n// acquis des chiffres qui ont pu être refusés — exactement le contraire de ce qu'on\n// attend d'un outil de décision.\n//\n// Chaque dépôt de fichier ouvre donc une campagne identifiée par ce fichier et son\n// premier mois à prévoir. Une nouvelle campagne démarre VIDE. Reprendre les saisies de\n// la précédente reste possible, mais c'est un geste explicite, et la reprise se fait par\n// MOIS CALENDAIRE et non par position : le mois +1 d'août n'est pas le mois +1 de\n// septembre.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nexport interface Campaign {\n id: string;\n source_file: string;\n /** Mois prévus par cette campagne, dans l'ordre. */\n months: string[];\n /** Corrections manuelles, clé `${code}|${index de mois}`. */\n overrides: Record;\n /** Objectif de CA par mois, aligné sur `months`. null = pas d'objectif. */\n revTarget: (number | null)[];\n /** Réglages du moteur au moment de la saisie. */\n params: Record;\n created_at: string;\n updated_at: string;\n updated_by: string | null;\n /** Dernier export vers le demand planning — trace de ce qui a réellement été envoyé. */\n exported_at: string | null;\n}\n\nconst KEY = \"forecast_campaigns\";\nconst MAX = 12; // un an d'historique mensuel\n\ntype Store = Record;\n\nasync function load(): Promise {\n const s = (await blob.getJSON(KEY)) as Store | null;\n return s && typeof s === \"object\" ? s : {};\n}\n\nasync function save(store: Store) {\n // On ne garde que les campagnes les plus récentes : le blob doit rester petit, il est\n // relu à chaque ouverture de l'onglet.\n const kept = Object.values(store)\n .sort((a, b) => (b.created_at ?? \"\").localeCompare(a.created_at ?? \"\"))\n .slice(0, MAX);\n await blob.setJSON(KEY, Object.fromEntries(kept.map((c) => [c.id, c])));\n}\n\nexport async function listCampaigns(): Promise<\n Pick[]\n> {\n const store = await load();\n return Object.values(store)\n .map(({ overrides: _o, revTarget: _r, params: _p, ...meta }) => meta)\n .sort((a, b) => (b.created_at ?? \"\").localeCompare(a.created_at ?? \"\"));\n}\n\nexport async function getCampaign(id: string): Promise {\n return (await load())[id] ?? null;\n}\n\nexport async function saveCampaign(\n user: string,\n input: {\n id: string; source_file: string; months: string[];\n overrides?: Record;\n revTarget?: (number | null)[];\n params?: Record;\n exported?: boolean;\n },\n): Promise {\n const store = await load();\n const now = new Date().toISOString();\n const prev = store[input.id];\n\n const c: Campaign = {\n id: input.id,\n source_file: input.source_file,\n months: input.months,\n overrides: input.overrides ?? prev?.overrides ?? {},\n revTarget: input.revTarget ?? prev?.revTarget ?? [],\n params: input.params ?? prev?.params ?? {},\n created_at: prev?.created_at ?? now,\n updated_at: now,\n updated_by: user,\n exported_at: input.exported ? now : (prev?.exported_at ?? null),\n };\n store[input.id] = c;\n await save(store);\n return c;\n}\n\n/** Reprend les saisies d'une campagne dans une autre, en réalignant sur les mois.\n *\n * Le réalignement par mois calendaire est le cœur de l'affaire : une correction posée\n * sur « novembre » doit rester sur novembre, même si novembre est passé du 4ᵉ au 3ᵉ mois\n * de l'horizon. Les mois qui ne font plus partie de l'horizon sont abandonnés, et on\n * renvoie leur nombre pour pouvoir le dire à l'utilisateur plutôt que de le taire. */\nexport function remapOverrides(\n overrides: Record,\n fromMonths: string[],\n toMonths: string[],\n): { overrides: Record; carried: number; dropped: number } {\n const out: Record = {};\n let carried = 0, dropped = 0;\n for (const [key, value] of Object.entries(overrides)) {\n const [code, idxRaw] = key.split(\"|\");\n const month = fromMonths[Number(idxRaw)];\n const target = month ? toMonths.indexOf(month) : -1;\n if (target === -1) { dropped++; continue; }\n out[`${code}|${target}`] = value;\n carried++;\n }\n return { overrides: out, carried, dropped };\n}\n\n/** Idem pour les objectifs de CA. */\nexport function remapTargets(\n targets: (number | null)[],\n fromMonths: string[],\n toMonths: string[],\n): (number | null)[] {\n return toMonths.map((m) => {\n const i = fromMonths.indexOf(m);\n return i === -1 ? null : (targets[i] ?? null);\n });\n}\n",[208],"4f3da8ae-a76a-466b-a475-f0beff87e6af","scripts","01a033f6-8ab0-7225-822f-6ff49492cb0f",["D",1787578190526],"directory",[214],"1c838a06-56d1-4f7b-8bba-25019d4d4ef7","frontend","01a033f6-8ab0-7225-822e-fd3ba7c49a32",["D",1787578190526],[220],"pulls","branches",9,"forks","envVars",{"_232":234,"_235":64},"connections",{"_232":64,"_233":64},"count","unfilledCount",10,"emptyCount",{"_44":219,"_32":220,"_30":221,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":603,"_13":64,"_34":217,"_36":-5,"_24":25,"_161":-5,"_289":604,"_27":605},{"_44":596,"_32":597,"_30":598,"_156":221,"_42":43,"_41":157,"_158":60,"_39":599,"_13":64,"_34":35,"_36":600,"_24":25,"_161":-5,"_289":601,"_27":602},{"_44":589,"_32":590,"_30":591,"_156":221,"_42":43,"_41":157,"_158":60,"_39":592,"_13":64,"_34":35,"_36":593,"_24":25,"_161":-5,"_289":594,"_27":595},{"_44":582,"_32":583,"_30":584,"_156":221,"_42":43,"_41":157,"_158":60,"_39":585,"_13":64,"_34":35,"_36":586,"_24":25,"_161":-5,"_289":587,"_27":588},{"_44":578,"_32":33,"_30":31,"_156":221,"_42":43,"_41":14,"_158":60,"_39":579,"_13":157,"_34":35,"_36":38,"_24":25,"_161":-5,"_289":580,"_27":581},{"_44":571,"_32":572,"_30":573,"_156":221,"_42":43,"_41":157,"_158":60,"_39":574,"_13":64,"_34":35,"_36":575,"_24":25,"_161":-5,"_289":576,"_27":577},{"_44":564,"_32":565,"_30":566,"_156":221,"_42":43,"_41":157,"_158":60,"_39":567,"_13":64,"_34":35,"_36":568,"_24":25,"_161":-5,"_289":569,"_27":570},{"_44":557,"_32":558,"_30":559,"_156":221,"_42":43,"_41":157,"_158":60,"_39":560,"_13":64,"_34":35,"_36":561,"_24":25,"_161":-5,"_289":562,"_27":563},{"_44":550,"_32":551,"_30":552,"_156":221,"_42":43,"_41":157,"_158":60,"_39":553,"_13":64,"_34":20,"_36":554,"_24":25,"_161":-5,"_289":555,"_27":556},{"_44":543,"_32":544,"_30":545,"_156":221,"_42":43,"_41":157,"_158":60,"_39":546,"_13":64,"_34":35,"_36":547,"_24":25,"_161":-5,"_289":548,"_27":549},{"_44":536,"_32":537,"_30":538,"_156":221,"_42":43,"_41":157,"_158":60,"_39":539,"_13":64,"_34":35,"_36":540,"_24":25,"_161":-5,"_289":541,"_27":542},{"_44":529,"_32":530,"_30":531,"_156":221,"_42":43,"_41":157,"_158":60,"_39":532,"_13":64,"_34":20,"_36":533,"_24":25,"_161":-5,"_289":534,"_27":535},{"_44":522,"_32":523,"_30":524,"_156":221,"_42":43,"_41":157,"_158":60,"_39":525,"_13":64,"_34":35,"_36":526,"_24":25,"_161":-5,"_289":527,"_27":528},{"_44":515,"_32":516,"_30":517,"_156":221,"_42":43,"_41":157,"_158":60,"_39":518,"_13":64,"_34":35,"_36":519,"_24":25,"_161":-5,"_289":520,"_27":521},{"_44":508,"_32":509,"_30":510,"_156":221,"_42":43,"_41":157,"_158":60,"_39":511,"_13":64,"_34":35,"_36":512,"_24":25,"_161":-5,"_289":513,"_27":514},{"_44":213,"_32":214,"_30":215,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":505,"_13":64,"_34":217,"_36":-5,"_24":25,"_161":-5,"_289":506,"_27":507},{"_44":498,"_32":499,"_30":500,"_156":215,"_42":43,"_41":157,"_158":60,"_39":501,"_13":64,"_34":35,"_36":502,"_24":25,"_161":-5,"_289":503,"_27":504},{"_44":491,"_32":492,"_30":493,"_156":215,"_42":43,"_41":157,"_158":60,"_39":494,"_13":64,"_34":35,"_36":495,"_24":25,"_161":-5,"_289":496,"_27":497},{"_44":484,"_32":485,"_30":486,"_156":215,"_42":43,"_41":157,"_158":60,"_39":487,"_13":64,"_34":35,"_36":488,"_24":25,"_161":-5,"_289":489,"_27":490},{"_44":477,"_32":478,"_30":479,"_156":215,"_42":43,"_41":157,"_158":60,"_39":480,"_13":64,"_34":35,"_36":481,"_24":25,"_161":-5,"_289":482,"_27":483},{"_44":470,"_32":471,"_30":472,"_156":215,"_42":43,"_41":157,"_158":60,"_39":473,"_13":64,"_34":35,"_36":474,"_24":25,"_161":-5,"_289":475,"_27":476},{"_44":463,"_32":464,"_30":465,"_156":215,"_42":43,"_41":157,"_158":25,"_39":466,"_13":64,"_34":35,"_36":467,"_24":25,"_161":-5,"_289":468,"_27":469},{"_44":456,"_32":457,"_30":458,"_156":215,"_42":43,"_41":157,"_158":60,"_39":459,"_13":64,"_34":35,"_36":460,"_24":25,"_161":-5,"_289":461,"_27":462},{"_44":449,"_32":450,"_30":451,"_156":215,"_42":43,"_41":157,"_158":60,"_39":452,"_13":64,"_34":35,"_36":453,"_24":25,"_161":-5,"_289":454,"_27":455},{"_44":442,"_32":443,"_30":444,"_156":215,"_42":43,"_41":157,"_158":60,"_39":445,"_13":64,"_34":35,"_36":446,"_24":25,"_161":-5,"_289":447,"_27":448},{"_44":435,"_32":436,"_30":437,"_156":215,"_42":43,"_41":157,"_158":60,"_39":438,"_13":64,"_34":35,"_36":439,"_24":25,"_161":-5,"_289":440,"_27":441},{"_44":428,"_32":429,"_30":430,"_156":215,"_42":43,"_41":157,"_158":60,"_39":431,"_13":64,"_34":35,"_36":432,"_24":25,"_161":-5,"_289":433,"_27":434},{"_44":421,"_32":422,"_30":423,"_156":215,"_42":43,"_41":157,"_158":60,"_39":424,"_13":64,"_34":35,"_36":425,"_24":25,"_161":-5,"_289":426,"_27":427},{"_44":414,"_32":415,"_30":416,"_156":215,"_42":43,"_41":157,"_158":60,"_39":417,"_13":64,"_34":35,"_36":418,"_24":25,"_161":-5,"_289":419,"_27":420},{"_44":407,"_32":408,"_30":409,"_156":215,"_42":43,"_41":157,"_158":60,"_39":410,"_13":64,"_34":35,"_36":411,"_24":25,"_161":-5,"_289":412,"_27":413},{"_44":400,"_32":401,"_30":402,"_156":215,"_42":43,"_41":157,"_158":60,"_39":403,"_13":64,"_34":35,"_36":404,"_24":25,"_161":-5,"_289":405,"_27":406},{"_44":393,"_32":394,"_30":395,"_156":215,"_42":43,"_41":157,"_158":60,"_39":396,"_13":64,"_34":35,"_36":397,"_24":25,"_161":-5,"_289":398,"_27":399},{"_44":386,"_32":387,"_30":388,"_156":215,"_42":43,"_41":157,"_158":60,"_39":389,"_13":64,"_34":35,"_36":390,"_24":25,"_161":-5,"_289":391,"_27":392},{"_44":379,"_32":380,"_30":381,"_156":215,"_42":43,"_41":157,"_158":60,"_39":382,"_13":64,"_34":35,"_36":383,"_24":25,"_161":-5,"_289":384,"_27":385},{"_44":372,"_32":373,"_30":374,"_156":215,"_42":43,"_41":157,"_158":60,"_39":375,"_13":64,"_34":35,"_36":376,"_24":25,"_161":-5,"_289":377,"_27":378},{"_44":365,"_32":366,"_30":367,"_156":215,"_42":43,"_41":157,"_158":60,"_39":368,"_13":64,"_34":35,"_36":369,"_24":25,"_161":-5,"_289":370,"_27":371},{"_44":358,"_32":359,"_30":360,"_156":215,"_42":43,"_41":157,"_158":60,"_39":361,"_13":64,"_34":35,"_36":362,"_24":25,"_161":-5,"_289":363,"_27":364},{"_44":351,"_32":352,"_30":353,"_156":215,"_42":43,"_41":157,"_158":60,"_39":354,"_13":64,"_34":35,"_36":355,"_24":25,"_161":-5,"_289":356,"_27":357},{"_44":344,"_32":345,"_30":346,"_156":215,"_42":43,"_41":157,"_158":60,"_39":347,"_13":64,"_34":35,"_36":348,"_24":25,"_161":-5,"_289":349,"_27":350},{"_44":337,"_32":338,"_30":339,"_156":215,"_42":43,"_41":157,"_158":60,"_39":340,"_13":64,"_34":35,"_36":341,"_24":25,"_161":-5,"_289":342,"_27":343},{"_44":330,"_32":331,"_30":332,"_156":215,"_42":43,"_41":157,"_158":60,"_39":333,"_13":64,"_34":35,"_36":334,"_24":25,"_161":-5,"_289":335,"_27":336},{"_44":323,"_32":324,"_30":325,"_156":215,"_42":43,"_41":157,"_158":60,"_39":326,"_13":64,"_34":35,"_36":327,"_24":25,"_161":-5,"_289":328,"_27":329},{"_44":316,"_32":317,"_30":318,"_156":215,"_42":43,"_41":157,"_158":60,"_39":319,"_13":64,"_34":35,"_36":320,"_24":25,"_161":-5,"_289":321,"_27":322},{"_44":207,"_32":208,"_30":209,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":313,"_13":64,"_34":35,"_36":211,"_24":25,"_161":-5,"_289":314,"_27":315},{"_44":201,"_32":202,"_30":203,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":310,"_13":64,"_34":35,"_36":205,"_24":25,"_161":-5,"_289":311,"_27":312},{"_44":195,"_32":196,"_30":197,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":307,"_13":64,"_34":35,"_36":199,"_24":25,"_161":-5,"_289":308,"_27":309},{"_44":188,"_32":189,"_30":190,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":304,"_13":64,"_34":192,"_36":193,"_24":25,"_161":-5,"_289":305,"_27":306},{"_44":182,"_32":183,"_30":184,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":301,"_13":64,"_34":35,"_36":186,"_24":25,"_161":-5,"_289":302,"_27":303},{"_44":176,"_32":177,"_30":178,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":298,"_13":64,"_34":35,"_36":180,"_24":25,"_161":-5,"_289":299,"_27":300},{"_44":169,"_32":170,"_30":171,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":295,"_13":64,"_34":173,"_36":174,"_24":25,"_161":-5,"_289":296,"_27":297},{"_44":163,"_32":164,"_30":165,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":292,"_13":64,"_34":35,"_36":167,"_24":25,"_161":-5,"_289":293,"_27":294},{"_44":153,"_32":154,"_30":155,"_156":-5,"_42":43,"_41":157,"_158":60,"_39":288,"_13":64,"_34":35,"_36":160,"_24":25,"_161":-5,"_289":290,"_27":291},["D",1787578190526],"children",[],[154],["D",1787578190526],[],[164],["D",1787578190526],[],[170],["D",1787578190526],[],[177],["D",1787578190526],[],[183],["D",1787578190526],[],[189],["D",1787578190526],[],[196],["D",1787578190526],[],[202],["D",1787578190526],[],[208],"dc1bd6c5-1a95-4dec-8297-4739b202da11","where_setdata.ts","01a033f6-8ab0-7225-822f-687f47cb1960",["D",1787578190526],"const BASE = \"https://nicolasduchene--019fe5b7086a744799319232a1c4ddf3.web.val.run/frontend/\";\nconst js = await (await fetch(BASE + \"App.tsx\")).text();\nconst lines = js.split(\"\\n\");\nlines.forEach((l, i) => {\n if (/\\bsetData\\b/.test(l)) {\n console.log(`L${i + 1}: ${l.trim().slice(0, 240)}`);\n }\n});\n",[],[214,317],"8c622577-320f-4235-9433-bd0d9a358381","verify_refund_dating.ts","01a033f6-8ab0-7225-822f-676a046822fa",["D",1787578190526],"// Contrôle de la règle d'imputation : un avoir doit tomber dans le mois du REMBOURSEMENT,\n// jamais dans celui de la commande. On cherche les cas où les deux diffèrent.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst o = (await blob.getJSON(\"orders_lines\")) as any;\nconst lines: any[] = o.lines;\nconst orphans: any[] = o.orphan_refunds ?? [];\n\nconst month = (d: string) => d.slice(0, 7);\n\n// ── Avoirs détaillés dont la commande et le remboursement sont dans des mois différents\nconst cross = lines.filter((l) => l.k === \"ref\" && l.rf && month(l.rf) !== month(l.d));\nconsole.log(\"═══ AVOIRS DÉTAILLÉS À CHEVAL SUR DEUX MOIS ═══\");\nconsole.log(cross.length, \"lignes\");\nfor (const l of cross.slice(0, 8)) {\n console.log(` ${l.o} ${l.b} | commandé ${l.d} (${month(l.d)}) → remboursé ${l.rf} (${month(l.rf)})`);\n}\n\n// ── Idem pour les avoirs sans détail\nconst crossO = orphans.filter((x) => month(x.rf) !== month(x.d));\nconsole.log();\nconsole.log(\"═══ AVOIRS SANS DÉTAIL À CHEVAL SUR DEUX MOIS ═══\");\nconsole.log(crossO.length, \"commandes\");\nfor (const x of crossO.slice(0, 10)) {\n console.log(` ${x.o} ${x.b} | commandé ${x.d} (${month(x.d)}) → remboursé ${x.rf} (${month(x.rf)})`);\n}\n\n// ── Cas qui intéresse directement : commandé en juillet, remboursé en août\nconst julAug = [\n ...lines.filter((l) => l.k === \"ref\" && l.rf && l.d.startsWith(\"2026-07\") && l.rf.startsWith(\"2026-08\"))\n .map((l) => ({ o: l.o, b: l.b, d: l.d, rf: l.rf, kind: \"détaillé\" })),\n ...orphans.filter((x) => x.d.startsWith(\"2026-07\") && x.rf.startsWith(\"2026-08\"))\n .map((x) => ({ o: x.o, b: x.b, d: x.d, rf: x.rf, kind: \"sans détail\" })),\n];\nconsole.log();\nconsole.log(\"═══ COMMANDÉ EN JUILLET, REMBOURSÉ EN AOÛT ═══\");\nconsole.log(julAug.length, \"cas — ils doivent être déduits d'AOÛT, pas de juillet\");\nfor (const x of julAug.slice(0, 15)) {\n console.log(` ${x.o} ${x.b} (${x.kind}) | cmd ${x.d} → remb. ${x.rf}`);\n}\n\n// ── Simulation de la fenêtre : ces avoirs comptent-ils en juillet ou en août ?\nconst inWindow = (rf: string, from: string, to: string) => rf >= from && rf <= to;\nconst nJul = julAug.filter((x) => inWindow(x.rf, \"2026-07-01\", \"2026-07-31\")).length;\nconst nAug = julAug.filter((x) => inWindow(x.rf, \"2026-08-01\", \"2026-08-31\")).length;\nconsole.log();\nconsole.log(\"Comptés dans la fenêtre juillet :\", nJul, \"(doit être 0)\");\nconsole.log(\"Comptés dans la fenêtre août :\", nAug, `(doit être ${julAug.length})`);\nconsole.log(nJul === 0 && nAug === julAug.length ? \"✔ Règle d'imputation correcte\" : \"✘ ANOMALIE\");\n",[],[214,324],"42490585-3127-4ad3-a6a0-17dfcaa85d26","verify_orphans.ts","01a033f6-8ab0-7225-822f-625882db5171",["D",1787578190526],"// Contrôle après modification : les avoirs sans détail sont-ils bien captés, et les\n// lignes produit sont-elles intactes ?\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst o = (await blob.getJSON(\"orders_lines\")) as any;\nconst lines: any[] = o.lines;\nconst orphans: any[] = o.orphan_refunds ?? [];\n\nconsole.log(\"blob :\", o.updated_at, \"| lignes:\", lines.length, \"| avoirs orphelins:\", orphans.length);\nconsole.log();\n\n// 1. Les lignes produit n'ont pas bougé (aucune ligne parasite sans quantité)\nconst bad = lines.filter((l) => !(l.q > 0));\nconsole.log(\"Lignes produit sans quantité positive (doit être 0) :\", bad.length);\nconst kinds: Record = {};\nfor (const l of lines) kinds[l.k] = (kinds[l.k] ?? 0) + 1;\nconsole.log(\"Répartition :\", JSON.stringify(kinds));\nconsole.log();\n\n// 2. Les orphelins, par mois\nconst byMonth: Record = {};\nfor (const x of orphans) {\n const m = x.rf.slice(0, 7);\n const e = (byMonth[m] ??= { n: 0, known: 0, val: 0 });\n e.n++;\n if (x.amt != null) { e.known++; e.val += x.amt; }\n}\nconsole.log(\"Avoirs sans détail par mois de remboursement :\");\nfor (const m of Object.keys(byMonth).sort()) {\n const e = byMonth[m];\n console.log(\" \", m, \":\", e.n, \"avoirs |\", e.known, \"chiffrés |\", Math.round(e.val), \"€ TTC\");\n}\nconsole.log();\n\n// 3. Aucun doublon commande + date\nconst seen = new Set();\nlet dup = 0;\nfor (const x of orphans) {\n const k = `${x.b}|${x.o}|${x.rf}`;\n if (seen.has(k)) dup++;\n seen.add(k);\n}\nconsole.log(\"Doublons commande+date (doit être 0) :\", dup);\nconsole.log();\n\n// 4. Contrôle croisé : un orphelin ne doit JAMAIS coexister avec une ligne \"ref\"\n// sur la même commande — sinon l'avoir serait compté deux fois.\nconst refOrders = new Set(lines.filter((l) => l.k === \"ref\").map((l) => `${l.b}|${l.o}`));\nconst overlap = orphans.filter((x) => refOrders.has(`${x.b}|${x.o}`));\nconsole.log(\"Commandes ayant À LA FOIS un avoir détaillé et un orphelin :\", overlap.length);\nfor (const x of overlap.slice(0, 10)) console.log(\" \", x.o, x.b, x.rf);\nconsole.log();\n\n// 5. Juillet : ce que ça représente face à l'écart BI constaté (51 653 €)\nconst jul = orphans.filter((x) => x.rf >= \"2026-07-01\" && x.rf <= \"2026-07-31\");\nconsole.log(\"Juillet 2026 :\", jul.length, \"avoirs sans détail\");\nconsole.log(\" dont chiffrés :\", jul.filter((x) => x.amt != null).length);\nconsole.log(\" écart BI à couvrir : 51 653 € HT\");\nconsole.log(\" marques :\", JSON.stringify(\n jul.reduce((a: any, x) => { a[x.b] = (a[x.b] ?? 0) + 1; return a; }, {})));\n",[],[214,331],"0293adae-9426-4b88-987f-1889c5a7bfd4","run_sync.ts","01a033f6-8ab0-7225-822f-5e15199ffb9b",["D",1787578190526],"// Déclenche une synchro des commandes et affiche le compte rendu.\nimport { syncOrders } from \"../orders_sync.ts\";\nconsole.log(JSON.stringify(await syncOrders(), null, 2));\n",[],[214,338],"d49dc050-4ab9-4e75-8cee-07ae4754482c","recon_july_v2.ts","01a033f6-8ab0-7225-822f-5a6463c528c5",["D",1787578190526],"// Rapprochement juillet 2026 APRÈS captation des avoirs sans détail.\n// Référence BI : 1 840 292,18 € (NSV, services et port compris).\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst FROM = \"2026-07-01\", TO = \"2026-07-31\";\nconst BI = 1840292.18;\nconst r2 = (n: number) => Math.round(n * 100) / 100;\nconst eur = (n: number) => n.toLocaleString(\"fr-FR\", { minimumFractionDigits: 2, maximumFractionDigits: 2 }) + \" €\";\n\nconst o = (await blob.getJSON(\"orders_lines\")) as any;\nconst catalog = (await blob.getJSON(\"catalog\")) as any;\nconst lines: any[] = o.lines;\nconst orphans: any[] = o.orphan_refunds ?? [];\nconst deee: Record = {};\nfor (const p of catalog.products as any[]) deee[String(p.code)] = p.deee ?? 0;\n\nconst amtOf = (l: any) => (l.n != null ? l.n : l.p * l.q) / 1.2 - (deee[String(l.sku ?? \"\")] ?? 0) * l.q;\nconst inJul = (l: any) => {\n const d = l.k === \"ref\" ? l.rf : (l.f || l.d);\n return d && d >= FROM && d <= TO;\n};\n\nlet ful = 0, ref = 0;\nfor (const l of lines) {\n if (l.k === \"ord\" || !inJul(l)) continue;\n if (l.k === \"ful\") ful += amtOf(l); else ref += amtOf(l);\n}\nconst seen = new Set();\nlet ship = 0;\nfor (const l of lines) {\n if (l.k !== \"ful\") continue;\n const d = l.f || l.d;\n if (!d || d < FROM || d > TO) continue;\n if (l.ship && !seen.has(l.o)) { seen.add(l.o); ship += l.ship / 1.2; }\n}\nconst orphJul = orphans.filter((x) => x.rf >= FROM && x.rf <= TO);\nconst orphVal = orphJul.reduce((s, x) => s + (x.amt ?? 0) / 1.2, 0);\n\nconst avant = ful - ref + ship;\nconst apres = avant - orphVal;\n\nconsole.log(\"═══ JUILLET 2026 ═══\");\nconsole.log(\"Expéditions \", eur(ful));\nconsole.log(\"Avoirs détaillés \", eur(-ref));\nconsole.log(\"Livraison \", eur(ship));\nconsole.log(\" → total avant correc.\", eur(avant));\nconsole.log();\nconsole.log(\"Avoirs sans détail \", eur(-orphVal), `(${orphJul.length} commandes)`);\nconsole.log(\" → TOTAL DASHBOARD \", eur(apres));\nconsole.log();\nconsole.log(\"BI \", eur(BI));\nconsole.log(\"Écart avant correction \", eur(avant - BI));\nconsole.log(\"Écart après correction \", eur(apres - BI), `(${r2((apres - BI) / BI * 100)} %)`);\nconsole.log();\n\n// Détail des avoirs sans détail, par marque\nconst byBrand: Record = {};\nfor (const x of orphJul) {\n const e = (byBrand[x.b] ??= { n: 0, v: 0 });\n e.n++; e.v += (x.amt ?? 0) / 1.2;\n}\nconsole.log(\"Avoirs sans détail par marque :\");\nfor (const [b, e] of Object.entries(byBrand)) console.log(\" \", b, e.n, \"avoirs |\", eur(e.v));\nconsole.log();\nconsole.log(\"Rappel : l'écart résiduel attendu = éco-participation non déductible sur ces\");\nconsole.log(\"avoirs (montant transaction TTC, éco-part comprise) + décalage expédition/facturation.\");\n",[],[214,345],"aa268d1f-f5c2-4a04-afd4-56a8e9ea818f","recon_july_sku.ts","01a033f6-8ab0-7225-822f-5680ae0c6843",["D",1787578190526],"// Rapprochement SKU par SKU — dashboard (Shopify expédié) vs BI (facturé) sur juillet 2026.\n// Lecture seule. Le mapping de référence : le BI nomme les services \"AES00063\", Shopify\n// les nomme \"FR_AES00063\" — on retire le préfixe FR_ pour rapprocher les deux.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { BI_JULY } from \"./bi_july_data.ts\";\n\nconst FROM = \"2026-07-01\", TO = \"2026-07-31\";\nconst r2 = (n: number) => Math.round(n * 100) / 100;\nconst biKey = (sku: string) => sku.startsWith(\"FR_\") ? sku.slice(3) : sku;\n\nconst orders = (await blob.getJSON(\"orders_lines\")) as any;\nconst catalog = (await blob.getJSON(\"catalog\")) as any;\nconst lines = orders.lines as any[];\nconst deee: Record = {};\nfor (const p of catalog.products as any[]) deee[String(p.code)] = p.deee ?? 0;\n\nconst inJul = (l: any) => {\n const d = l.k === \"ref\" ? l.rf : (l.f || l.d);\n return d && d >= FROM && d <= TO;\n};\n\n// ── Côté dashboard, par référence ───────────────────────────────────────────\nconst sh: Record = {};\nfor (const l of lines) {\n if (l.k === \"ord\" || !inJul(l)) continue;\n const sku = String(l.sku ?? \"\");\n const amt = (l.n != null ? l.n : l.p * l.q) / 1.2 - (deee[sku] ?? 0) * l.q;\n const sign = l.k === \"ref\" ? -1 : 1;\n const e = (sh[biKey(sku)] ??= { v: 0, q: 0, ful: 0, ref: 0 });\n e.v += sign * amt; e.q += sign * l.q;\n if (l.k === \"ful\") e.ful += amt; else e.ref += amt;\n}\n\nconst keys = new Set([...Object.keys(sh), ...Object.keys(BI_JULY)]);\nconst rows: { sku: string; s: number; b: number; d: number; sq: number; bq: number }[] = [];\nfor (const k of keys) {\n const s = sh[k]?.v ?? 0, b = BI_JULY[k]?.[0] ?? 0;\n rows.push({ sku: k, s, b, d: s - b, sq: sh[k]?.q ?? 0, bq: BI_JULY[k]?.[1] ?? 0 });\n}\n\nconst isSvc = (k: string) => k.startsWith(\"AES\");\nconst sum = (f: (r: typeof rows[0]) => number, p = (_: typeof rows[0]) => true) =>\n r2(rows.filter(p).reduce((a, r) => a + f(r), 0));\n\nconsole.log(\"═══ TOTAUX ═══\");\nconsole.log(\"Shopify marchandises (hors services) :\", sum((r) => r.s, (r) => !isSvc(r.sku)));\nconsole.log(\"BI marchandises (hors services) :\", sum((r) => r.b, (r) => !isSvc(r.sku)));\nconsole.log(\" écart :\", sum((r) => r.d, (r) => !isSvc(r.sku)));\nconsole.log();\nconsole.log(\"Shopify services AES :\", sum((r) => r.s, (r) => isSvc(r.sku) && r.sku !== \"AESFR001\"));\nconsole.log(\"BI services AES :\", sum((r) => r.b, (r) => isSvc(r.sku) && r.sku !== \"AESFR001\"));\nconsole.log(\" écart :\", sum((r) => r.d, (r) => isSvc(r.sku) && r.sku !== \"AESFR001\"));\nconsole.log();\nconsole.log(\"BI AESFR001 (port facturé) :\", BI_JULY[\"AESFR001\"]?.[0], \"| Shopify:\", r2(sh[\"AESFR001\"]?.v ?? 0));\n\n// ── Références présentes d'un seul côté ─────────────────────────────────────\nconst onlySh = rows.filter((r) => r.b === 0 && Math.abs(r.s) > 0.01).sort((a, b) => b.s - a.s);\nconst onlyBi = rows.filter((r) => (sh[r.sku] === undefined) && Math.abs(r.b) > 0.01).sort((a, b) => b.b - a.b);\nconsole.log();\nconsole.log(\"── Uniquement dans Shopify :\", onlySh.length, \"réfs |\", r2(onlySh.reduce((a, r) => a + r.s, 0)), \"€\");\nfor (const r of onlySh.slice(0, 15)) console.log(\" \", r.sku, r2(r.s), \"€\", r.sq, \"u\");\nconsole.log(\"── Uniquement dans le BI :\", onlyBi.length, \"réfs |\", r2(onlyBi.reduce((a, r) => a + r.b, 0)), \"€\");\nfor (const r of onlyBi.slice(0, 15)) console.log(\" \", r.sku, r2(r.b), \"€\", r.bq, \"u\");\n\n// ── Plus gros écarts sur les références communes ────────────────────────────\nconst both = rows.filter((r) => r.b !== 0 && sh[r.sku] !== undefined);\nconsole.log();\nconsole.log(\"── 25 plus gros écarts (Shopify > BI) ──\");\nfor (const r of [...both].sort((a, b) => b.d - a.d).slice(0, 25)) {\n console.log(\" \", r.sku, \"sh\", r2(r.s), \"| bi\", r2(r.b), \"| Δ\", r2(r.d), \"| qté sh\", r.sq, \"bi\", r.bq);\n}\nconsole.log(\"── 25 plus gros écarts (BI > Shopify) ──\");\nfor (const r of [...both].sort((a, b) => a.d - b.d).slice(0, 25)) {\n console.log(\" \", r.sku, \"sh\", r2(r.s), \"| bi\", r2(r.b), \"| Δ\", r2(r.d), \"| qté sh\", r.sq, \"bi\", r.bq);\n}\n\n// ── Quantités ───────────────────────────────────────────────────────────────\nconsole.log();\nconsole.log(\"Qté Shopify (hors services):\", sum((r) => r.sq, (r) => !isSvc(r.sku)));\nconsole.log(\"Qté BI (hors services):\", sum((r) => r.bq, (r) => !isSvc(r.sku)));\n",[],[214,352],"42027704-53dc-4209-8e78-5b1139889071","recon_july_causes.ts","01a033f6-8ab0-7225-822f-51434979ee33",["D",1787578190526],"// Décomposition de l'écart juillet — on cherche la NATURE des différences, pas leur montant.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { BI_JULY } from \"./bi_july_data.ts\";\n\nconst r2 = (n: number) => Math.round(n * 100) / 100;\nconst biKey = (s: string) => s.startsWith(\"FR_\") ? s.slice(3) : s;\nconst isSvc = (k: string) => k.startsWith(\"AES\");\n\nconst orders = (await blob.getJSON(\"orders_lines\")) as any;\nconst catalog = (await blob.getJSON(\"catalog\")) as any;\nconst lines = orders.lines as any[];\nconst deee: Record = {};\nfor (const p of catalog.products as any[]) deee[String(p.code)] = p.deee ?? 0;\nconst amtOf = (l: any) => (l.n != null ? l.n : l.p * l.q) / 1.2 - (deee[String(l.sku ?? \"\")] ?? 0) * l.q;\n\n// ── 1. Les avoirs, mois par mois, côté Shopify ──────────────────────────────\nconsole.log(\"═══ 1. AVOIRS SHOPIFY PAR MOIS (date de remboursement) ═══\");\nconst refByMonth: Record = {};\nfor (const l of lines) {\n if (l.k !== \"ref\" || !l.rf) continue;\n const m = l.rf.slice(0, 7);\n const e = (refByMonth[m] ??= { v: 0, q: 0, n: 0 });\n e.v += amtOf(l); e.q += l.q; e.n++;\n}\nfor (const m of Object.keys(refByMonth).sort()) {\n const e = refByMonth[m];\n console.log(\" \", m, r2(e.v), \"€ |\", e.q, \"u |\", e.n, \"lignes\");\n}\n\n// ── 2. Effet volume vs effet prix sur les marchandises ──────────────────────\nconsole.log();\nconsole.log(\"═══ 2. MARCHANDISES : effet volume vs effet prix ═══\");\nconst sh: Record = {};\nfor (const l of lines) {\n if (l.k === \"ord\") continue;\n const d = l.k === \"ref\" ? l.rf : (l.f || l.d);\n if (!d || d < \"2026-07-01\" || d > \"2026-07-31\") continue;\n const sign = l.k === \"ref\" ? -1 : 1;\n const e = (sh[biKey(String(l.sku ?? \"\"))] ??= { v: 0, q: 0 });\n e.v += sign * amtOf(l); e.q += sign * l.q;\n}\nlet volEffect = 0, priceEffect = 0, noBase = 0;\nfor (const k of new Set([...Object.keys(sh), ...Object.keys(BI_JULY)])) {\n if (isSvc(k)) continue;\n const s = sh[k]?.v ?? 0, sq = sh[k]?.q ?? 0;\n const b = BI_JULY[k]?.[0] ?? 0, bq = BI_JULY[k]?.[1] ?? 0;\n if (bq !== 0) {\n const pu = b / bq; // prix unitaire BI\n volEffect += (sq - bq) * pu;\n priceEffect += s - sq * pu;\n } else { noBase += s - b; }\n}\nconsole.log(\" effet volume (Δ quantités × PU BI) :\", r2(volEffect), \"€\");\nconsole.log(\" effet prix (même quantité) :\", r2(priceEffect), \"€\");\nconsole.log(\" réfs sans base BI :\", r2(noBase), \"€\");\nconsole.log(\" total marchandises :\", r2(volEffect + priceEffect + noBase), \"€\");\n\n// ── 3. Sensibilité à la fenêtre : le BI facture-t-il avec un décalage ? ──────\nconsole.log();\nconsole.log(\"═══ 3. MARCHANDISES SHOPIFY SELON LA FENÊTRE (hors services) ═══\");\nconst windowSum = (from: string, to: string) => {\n let v = 0, q = 0;\n for (const l of lines) {\n if (l.k === \"ord\") continue;\n const sku = biKey(String(l.sku ?? \"\"));\n if (isSvc(sku)) continue;\n const d = l.k === \"ref\" ? l.rf : (l.f || l.d);\n if (!d || d < from || d > to) continue;\n const sign = l.k === \"ref\" ? -1 : 1;\n v += sign * amtOf(l); q += sign * l.q;\n }\n return [r2(v), q] as const;\n};\nfor (const [f, t] of [\n [\"2026-06-28\", \"2026-07-28\"], [\"2026-06-29\", \"2026-07-29\"], [\"2026-06-30\", \"2026-07-30\"],\n [\"2026-07-01\", \"2026-07-31\"], [\"2026-07-02\", \"2026-08-01\"], [\"2026-07-03\", \"2026-08-02\"],\n] as const) {\n const [v, q] = windowSum(f, t);\n console.log(\" \", f, \"→\", t, \":\", v, \"€ |\", q, \"u | Δ vs BI\", r2(v - 1744937.86));\n}\n\n// ── 4. Expéditions de juillet dont la commande est antérieure ───────────────\nconsole.log();\nconsole.log(\"═══ 4. DÉCALAGE COMMANDE → EXPÉDITION (lignes de juillet) ═══\");\nconst lagBuckets: Record = {};\nfor (const l of lines) {\n if (l.k !== \"ful\") continue;\n const d = l.f || l.d;\n if (!d || d < \"2026-07-01\" || d > \"2026-07-31\") continue;\n const sku = biKey(String(l.sku ?? \"\"));\n if (isSvc(sku)) continue;\n const lag = Math.round((Date.parse(d) - Date.parse(l.d)) / 864e5);\n const b = lag <= 0 ? \"0 j\" : lag <= 2 ? \"1-2 j\" : lag <= 7 ? \"3-7 j\" : lag <= 30 ? \"8-30 j\" : \">30 j\";\n const e = (lagBuckets[b] ??= { v: 0, n: 0 });\n e.v += amtOf(l); e.n++;\n}\nfor (const k of [\"0 j\", \"1-2 j\", \"3-7 j\", \"8-30 j\", \">30 j\"]) {\n if (lagBuckets[k]) console.log(\" \", k, \":\", r2(lagBuckets[k].v), \"€ |\", lagBuckets[k].n, \"lignes\");\n}\n",[],[214,359],"75d2de89-5173-4d09-a7b2-de1628cbba4e","recon_july.ts","01a033f6-8ab0-7225-822f-4cbafd764a92",["D",1787578190526],"// Script d'analyse jetable — rapprochement juillet 2026 Shopify (dashboard) vs BI.\n// Ne fait que lire les blobs et imprimer des agrégats ; n'écrit rien.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst FROM = \"2026-07-01\", TO = \"2026-07-31\";\nconst r2 = (n: number) => Math.round(n * 100) / 100;\n\nconst orders = (await blob.getJSON(\"orders_lines\")) as any;\nconst catalog = (await blob.getJSON(\"catalog\")) as any;\nconst lines = orders.lines as any[];\nconst products = catalog.products as any[];\n\nconst deee: Record = {};\nconst known = new Set();\nconst brandOf: Record = {};\nfor (const p of products) {\n deee[String(p.code)] = p.deee ?? 0;\n known.add(String(p.code));\n brandOf[String(p.code)] = p.brand;\n}\n\nconst inJul = (l: any) => {\n const d = l.k === \"ref\" ? l.rf : (l.f || l.d);\n return d && d >= FROM && d <= TO;\n};\n\n// ── Marchandises expédiées, définition dashboard ────────────────────────────\nlet gFul = 0, gRef = 0, qFul = 0, qRef = 0;\nlet nFul = 0, nRef = 0;\n// Ventilation\nlet unknownSku = 0, unknownVal = 0, unknownQ = 0;\nlet noSku = 0, noSkuVal = 0;\nconst byBrand: Record = {};\nconst unknownList: Record = {};\n\nfor (const l of lines) {\n if (l.k === \"ord\") continue;\n if (!inJul(l)) continue;\n const sku = String(l.sku ?? \"\");\n const gross = (l.n != null ? l.n : l.p * l.q) / 1.2;\n const dd = (deee[sku] ?? 0) * l.q;\n const amt = gross - dd;\n if (l.k === \"ful\") { gFul += amt; qFul += l.q; nFul++; } else { gRef += amt; qRef += l.q; nRef++; }\n const sign = l.k === \"ref\" ? -1 : 1;\n byBrand[l.b] = (byBrand[l.b] ?? 0) + sign * amt;\n if (!sku) { noSku++; noSkuVal += sign * amt; }\n else if (!known.has(sku)) {\n unknownSku++; unknownVal += sign * amt; unknownQ += sign * l.q;\n const e = (unknownList[sku] ??= { v: 0, q: 0 });\n e.v += sign * amt; e.q += sign * l.q;\n }\n}\n\n// ── Livraison facturée ──────────────────────────────────────────────────────\nconst seen = new Set();\nlet ship = 0;\nfor (const l of lines) {\n if (l.k !== \"ful\") continue;\n const d = l.f || l.d;\n if (!d || d < FROM || d > TO) continue;\n if (l.ship && !seen.has(l.o)) { seen.add(l.o); ship += l.ship / 1.2; }\n}\n\nconst goods = gFul - gRef;\n\nconsole.log(\"═══ JUILLET 2026 — côté dashboard (Shopify) ═══\");\nconsole.log(\"Expéditions :\", r2(gFul), \"€ HT hors DEEE |\", qFul, \"u |\", nFul, \"lignes\");\nconsole.log(\"Avoirs :\", r2(-gRef), \"€ |\", qRef, \"u |\", nRef, \"lignes\");\nconsole.log(\"Marchandises net :\", r2(goods), \"€\");\nconsole.log(\"Livraison :\", r2(ship), \"€ |\", seen.size, \"commandes\");\nconsole.log(\"TOTAL (goods+ship):\", r2(goods + ship), \"€\");\nconsole.log();\nconsole.log(\"Par marque (marchandises) :\", Object.fromEntries(Object.entries(byBrand).map(([k, v]) => [k, r2(v)])));\nconsole.log();\nconsole.log(\"── SKU absents du catalogue ──\");\nconsole.log(\"lignes:\", unknownSku, \"| valeur:\", r2(unknownVal), \"€ | qté:\", unknownQ);\nconst top = Object.entries(unknownList).sort((a, b) => b[1].v - a[1].v).slice(0, 40);\nfor (const [sku, e] of top) console.log(\" \", sku, r2(e.v), \"€\", e.q, \"u\");\nconsole.log(\"lignes sans SKU:\", noSku, \"| valeur:\", r2(noSkuVal), \"€\");\nconsole.log();\nconsole.log(\"DEEE total déduit:\", r2(lines.filter((l) => l.k !== \"ord\" && inJul(l))\n .reduce((s, l) => s + (l.k === \"ref\" ? -1 : 1) * (deee[String(l.sku ?? \"\")] ?? 0) * l.q, 0)), \"€\");\n",[],[214,366],"12aab309-b728-401d-9739-986466d74392","probe_refunds.ts","01a033f6-8ab0-7225-822f-4b8f5956ec82",["D",1787578190526],"// Que contient réellement le blob pour les commandes marquées « Remboursée » dans Shopify ?\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst orders = (await blob.getJSON(\"orders_lines\")) as any;\nconst lines = orders.lines as any[];\nconsole.log(\"blob mis à jour le\", orders.updated_at, \"| flux:\", JSON.stringify(orders.feeds));\nconsole.log();\n\nconst TARGETS = [\n \"FR015591\", \"FR015396\", \"FR015174\", \"FR015163\", \"FR015156\", \"FR015135\",\n \"FR014933\", \"FR014926\", \"FR014906\", \"FR014789\", \"FR014783\", \"FR014707\",\n \"FR014662\", \"FR014636\", \"FR014412\", \"FR014069\", \"FR014023\", \"FR013941\",\n];\n\nfor (const name of TARGETS) {\n const ls = lines.filter((l) => l.o === name);\n if (ls.length === 0) { console.log(name, \"→ ABSENT du blob\"); continue; }\n const byKind: Record = {};\n for (const l of ls) byKind[l.k] = (byKind[l.k] ?? 0) + 1;\n const rf = [...new Set(ls.map((l) => l.rf).filter(Boolean))];\n console.log(\n name, \"|\", ls[0].b, \"| cmd\", ls[0].d,\n \"| lignes:\", JSON.stringify(byKind),\n \"| rf:\", rf.length ? rf.join(\",\") : \"AUCUNE\",\n );\n}\n\n// ── Vue d'ensemble : commandes qui portent une date de remboursement (rf) mais\n// AUCUNE ligne de type \"ref\". C'est la signature d'un avoir non capturé.\nconsole.log();\nconsole.log(\"═══ COMMANDES AVEC rf MAIS SANS LIGNE 'ref' ═══\");\nconst byOrder: Record; b: string; d: string; val: number }> = {};\nfor (const l of lines) {\n const e = (byOrder[l.o] ??= { rf: null, kinds: new Set(), b: l.b, d: l.d, val: 0 });\n e.kinds.add(l.k);\n if (l.rf && !e.rf) e.rf = l.rf;\n if (l.k === \"ful\") e.val += (l.n != null ? l.n : l.p * l.q) / 1.2;\n}\nconst orphans = Object.entries(byOrder)\n .filter(([, e]) => e.rf && !e.kinds.has(\"ref\"))\n .map(([o, e]) => ({ o, ...e }));\nconsole.log(\"total:\", orphans.length, \"commandes |\", Math.round(orphans.reduce((s, x) => s + x.val, 0)), \"€ expédié HT\");\n\nconst byMonth: Record = {};\nfor (const x of orphans) {\n const m = x.rf!.slice(0, 7);\n const e = (byMonth[m] ??= { n: 0, v: 0 });\n e.n++; e.v += x.val;\n}\nfor (const m of Object.keys(byMonth).sort()) {\n console.log(\" \", m, \":\", byMonth[m].n, \"commandes |\", Math.round(byMonth[m].v), \"€\");\n}\nconsole.log();\nconsole.log(\"── exemples juillet ──\");\nfor (const x of orphans.filter((x) => x.rf!.startsWith(\"2026-07\")).slice(0, 20)) {\n console.log(\" \", x.o, x.b, \"cmd\", x.d, \"remb.\", x.rf, Math.round(x.val), \"€\");\n}\n",[],[214,373],"a3d9301f-77c8-412d-a68c-835ae16e308a","probe_refund_rows.ts","01a033f6-8ab0-7225-822f-474d2e5a8a83",["D",1787578190526],"// Combien de lignes « Refund Line » sont vides (sans SKU ni quantité) sur les trois flux ?\nimport { feeds } from \"../feeds.ts\";\n\nfunction parseCSV(text: string): string[][] {\n if (text.charCodeAt(0) === 0xFEFF) text = text.slice(1);\n const rows: string[][] = [];\n let row: string[] = [], field = \"\", inQ = false;\n for (let i = 0; i < text.length; i++) {\n const c = text[i];\n if (inQ) {\n if (c === '\"') { if (text[i + 1] === '\"') { field += '\"'; i++; } else inQ = false; }\n else field += c;\n } else if (c === '\"') inQ = true;\n else if (c === \",\") { row.push(field); field = \"\"; }\n else if (c === \"\\n\" || c === \"\\r\") {\n if (c === \"\\r\" && text[i + 1] === \"\\n\") i++;\n row.push(field); field = \"\";\n if (row.length > 1 || row[0] !== \"\") rows.push(row);\n row = [];\n } else field += c;\n }\n if (field !== \"\" || row.length) { row.push(field); rows.push(row); }\n return rows;\n}\n\nfor (const { brand, url } of feeds(\"ORDERS\")) {\n if (!url) continue;\n const rows = parseCSV(await (await fetch(url, { redirect: \"follow\" })).text());\n const low = rows[0].map((h) => h.trim().toLowerCase());\n const iType = low.indexOf(\"line: type\");\n const iSku = low.indexOf(\"line: sku\");\n const iQty = low.indexOf(\"line: quantity\");\n const iRef = low.indexOf(\"refund: created at\");\n const iSrc = low.indexOf(\"source\");\n\n let full = 0, empty = 0;\n const emptyByMonth: Record = {};\n const emptyBySrc: Record = {};\n const fullBySrc: Record = {};\n for (let r = 1; r < rows.length; r++) {\n if ((rows[r][iType] ?? \"\").trim() !== \"Refund Line\") continue;\n const sku = (rows[r][iSku] ?? \"\").trim();\n const qty = (rows[r][iQty] ?? \"\").trim();\n const src = (rows[r][iSrc] ?? \"\").trim() || \"(vide)\";\n const isWeb = src === \"web\";\n if (sku && qty) { full++; fullBySrc[isWeb ? \"web\" : \"marketplace\"] = (fullBySrc[isWeb ? \"web\" : \"marketplace\"] ?? 0) + 1; }\n else {\n empty++;\n emptyBySrc[isWeb ? \"web\" : \"marketplace\"] = (emptyBySrc[isWeb ? \"web\" : \"marketplace\"] ?? 0) + 1;\n const m = (rows[r][iRef] ?? \"\").slice(0, 7);\n if (m) emptyByMonth[m] = (emptyByMonth[m] ?? 0) + 1;\n }\n }\n console.log(\"═══\", brand, \"═══\");\n console.log(\" Refund Line exploitables (SKU+qté) :\", full, JSON.stringify(fullBySrc));\n console.log(\" Refund Line VIDES (date seule) :\", empty, JSON.stringify(emptyBySrc));\n const ms = Object.keys(emptyByMonth).sort();\n console.log(\" vides par mois :\", ms.map((m) => `${m}:${emptyByMonth[m]}`).join(\" \"));\n console.log();\n}\n",[],[214,380],"f23cd655-7807-44a8-ad58-1b9efbd513d2","probe_raw_text.ts","01a033f6-8ab0-7225-822f-409f67cde712",["D",1787578190526],"// Vérification sans parseur : on affiche les lignes BRUTES du CSV, telles quelles.\n// Objectif : exclure un bug d'analyse de ma part avant d'affirmer que les avoirs\n// Online Store n'ont pas de détail article.\nimport { feeds } from \"../feeds.ts\";\n\nconst TARGET = [\"FR015591\", \"FR015396\", \"FR014906\", \"FR014662\"];\n\nfor (const { brand, url } of feeds(\"ORDERS\")) {\n if (brand !== \"HAIER\" || !url) continue;\n const text = await (await fetch(url, { redirect: \"follow\" })).text();\n const raw = text.split(/\\r?\\n/);\n console.log(\"En-tête brut :\");\n console.log(raw[0]);\n console.log();\n for (const name of TARGET) {\n // Les champs sont entre guillemets dans l'export : on cherche \"FR015591\".\n const hits = raw.filter((l) => l.includes(`\"${name}\"`));\n console.log(`=== ${name} — ${hits.length} lignes brutes ===`);\n for (const l of hits) console.log(\" \", l);\n console.log();\n }\n}\n",[],[214,387],"4d45d088-ae51-4ea6-aef7-973f29ea34f1","probe_raw_export.ts","01a033f6-8ab0-7225-822f-3f117b003845",["D",1787578190526],"// Que contient l'export Matrixify BRUT pour une commande remboursée ?\n// Objectif : distinguer « Matrixify n'exporte pas les lignes d'avoir » de\n// « Shopify n'a pas de lignes d'avoir à exporter » (remboursement monétaire).\nimport { feeds } from \"../feeds.ts\";\n\nconst TARGET = [\"FR015591\", \"FR015396\", \"FR015174\", \"FR014906\", \"FR014662\", \"FR013941\"];\n\nfunction parseCSV(text: string): string[][] {\n if (text.charCodeAt(0) === 0xFEFF) text = text.slice(1);\n const rows: string[][] = [];\n let row: string[] = [], field = \"\", inQ = false;\n for (let i = 0; i < text.length; i++) {\n const c = text[i];\n if (inQ) {\n if (c === '\"') { if (text[i + 1] === '\"') { field += '\"'; i++; } else inQ = false; }\n else field += c;\n } else if (c === '\"') inQ = true;\n else if (c === \",\") { row.push(field); field = \"\"; }\n else if (c === \"\\n\" || c === \"\\r\") {\n if (c === \"\\r\" && text[i + 1] === \"\\n\") i++;\n row.push(field); field = \"\";\n if (row.length > 1 || row[0] !== \"\") rows.push(row);\n row = [];\n } else field += c;\n }\n if (field !== \"\" || row.length) { row.push(field); rows.push(row); }\n return rows;\n}\n\nfor (const { brand, url } of feeds(\"ORDERS\")) {\n if (brand !== \"HAIER\" || !url) continue;\n const res = await fetch(url, { redirect: \"follow\" });\n const rows = parseCSV(await res.text());\n const header = rows[0].map((h) => h.trim());\n const low = header.map((h) => h.toLowerCase());\n console.log(\"=== HAIER — colonnes de l'export ===\");\n console.log(header.join(\" | \"));\n console.log();\n\n const iName = low.indexOf(\"name\");\n const iType = low.indexOf(\"line: type\");\n const iRef = low.indexOf(\"refund: created at\");\n\n // Répartition globale des types de ligne\n const types: Record = {};\n for (let r = 1; r < rows.length; r++) {\n const t = (rows[r][iType] ?? \"\").trim() || \"(vide)\";\n types[t] = (types[t] ?? 0) + 1;\n }\n console.log(\"=== Répartition 'Line: Type' sur tout l'export ===\");\n for (const [t, n] of Object.entries(types).sort((a, b) => b[1] - a[1])) console.log(\" \", t, \":\", n);\n console.log();\n\n for (const name of TARGET) {\n const rs = rows.filter((r, i) => i > 0 && (r[iName] ?? \"\").trim() === name);\n if (!rs.length) continue;\n console.log(\"=== \" + name + \" — \" + rs.length + \" lignes brutes ===\");\n for (const r of rs) {\n const cells = header.map((h, i) => [h, (r[i] ?? \"\").trim()])\n .filter(([, v]) => v !== \"\")\n .map(([h, v]) => `${h}=${v}`);\n console.log(\" \", cells.join(\" · \"));\n }\n console.log();\n }\n}\n",[],[214,394],"6b358a50-b071-476e-8ba4-9f00d3b95d98","probe_orphan_nature.ts","01a033f6-8ab0-7225-822f-3b706cbe8336",["D",1787578190526],"// Ces avoirs sans détail sont-ils des retours complets ou des gestes commerciaux ?\n// On compare le montant remboursé à la valeur expédiée de la même commande.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst o = (await blob.getJSON(\"orders_lines\")) as any;\nconst lines: any[] = o.lines;\nconst orphans: any[] = (o.orphan_refunds ?? []).filter((x: any) =>\n x.rf >= \"2026-07-01\" && x.rf <= \"2026-07-31\" && x.amt != null);\n\n// Valeur TTC expédiée par commande\nconst shippedTtc: Record = {};\nfor (const l of lines) {\n if (l.k !== \"ful\") continue;\n const k = `${l.b}|${l.o}`;\n shippedTtc[k] = (shippedTtc[k] ?? 0) + (l.n != null ? l.n : l.p * l.q);\n}\n\nconst buckets = { total: 0, partiel: 0, sansExpedition: 0 };\nconst vals = { total: 0, partiel: 0, sansExpedition: 0 };\nconst exemples: string[] = [];\nfor (const x of orphans) {\n const ship = shippedTtc[`${x.b}|${x.o}`] ?? 0;\n if (ship === 0) { buckets.sansExpedition++; vals.sansExpedition += x.amt; continue; }\n const ratio = x.amt / ship;\n if (ratio >= 0.95) { buckets.total++; vals.total += x.amt; }\n else {\n buckets.partiel++; vals.partiel += x.amt;\n if (exemples.length < 12) {\n exemples.push(` ${x.o} ${x.b} | expédié ${ship.toFixed(2)} € | remboursé ${x.amt.toFixed(2)} € (${(ratio * 100).toFixed(0)} %)`);\n }\n }\n}\n\nconsole.log(\"═══ NATURE DES AVOIRS SANS DÉTAIL — JUILLET ═══\");\nconsole.log(orphans.length, \"avoirs chiffrés\\n\");\nconsole.log(\"Remboursement TOTAL (≥95 % de l'expédié) :\", buckets.total, \"|\", Math.round(vals.total), \"€ TTC\");\nconsole.log(\"Remboursement PARTIEL (<95 %) :\", buckets.partiel, \"|\", Math.round(vals.partiel), \"€ TTC\");\nconsole.log(\"Aucune expédition trouvée :\", buckets.sansExpedition, \"|\", Math.round(vals.sansExpedition), \"€ TTC\");\nconsole.log();\nconsole.log(\"Exemples de partiels :\");\nfor (const e of exemples) console.log(e);\n\nconst moy = orphans.reduce((s, x) => s + x.amt, 0) / orphans.length;\nconsole.log();\nconsole.log(\"Montant moyen remboursé :\", moy.toFixed(2), \"€ TTC\");\n",[],[214,401],"28a91a78-7c25-45ea-8e19-173d5ffef6e1","probe_headers.ts","01a033f6-8ab0-7225-822f-37e025921fe3",["D",1787578190526],"// Les nouvelles colonnes sont-elles présentes dans les trois exports ?\nimport { feeds } from \"../feeds.ts\";\n\nconst WANTED = [\"transaction: kind\", \"transaction: amount\", \"refund: id\", \"refund: restock\"];\n\nfor (const { brand, url } of feeds(\"ORDERS\")) {\n if (!url) { console.log(brand, \"— pas d'URL\"); continue; }\n const res = await fetch(url, { redirect: \"follow\" });\n const text = await res.text();\n const first = text.split(/\\r?\\n/)[0].replace(/^/, \"\");\n const cols = first.split(\",\").map((c) => c.trim().replace(/^\"|\"$/g, \"\"));\n const low = cols.map((c) => c.toLowerCase());\n console.log(\"═══\", brand, \"—\", cols.length, \"colonnes | fichier\", (text.length / 1e6).toFixed(1), \"Mo\");\n console.log(\" \", cols.join(\" | \"));\n for (const w of WANTED) {\n console.log(` ${low.includes(w) ? \"✔\" : \"✘\"} ${w}`);\n }\n console.log();\n}\n",[],[214,408],"89d80135-88d5-425d-9a86-23317e2ffbc0","find_syntax_error.ts","01a033f6-8ab0-7225-822f-31c01bdd56cf",["D",1787578190526],"// Localise l'erreur de syntaxe dans le JS transpilé servi au navigateur.\n// `new Function` ne donne pas de position : on ajoute les lignes une à une et on\n// repère la première dont l'ajout rend le tout non parsable — approche grossière\n// mais suffisante pour pointer la zone fautive.\nconst src = await (await fetch(\"https://d2c-france-dashboard.val.run/frontend/App.tsx\")).text();\nconst body = src\n .replace(/^\\s*import[\\s\\S]*?from\\s+[\"'][^\"']+[\"'];?/gm, \"\")\n .replace(/^\\s*import\\s+[\"'][^\"']+[\"'];?/gm, \"\")\n .replace(/^\\s*export\\s+default\\s+/gm, \"const __d = \")\n .replace(/^\\s*export\\s+/gm, \"\");\n\nconst lines = body.split(\"\\n\");\nconsole.log(\"lignes:\", lines.length);\n\n// Un préfixe peut être invalide simplement parce qu'il est incomplet. On cherche\n// donc la dernière ligne à partir de laquelle PLUS AUCUN suffixe ne redevient valide.\nconst parses = (s: string) => { try { new Function(s); return true; } catch { return false; } };\n\nlet firstBad = -1;\nfor (let i = 1; i <= lines.length; i++) {\n const prefix = lines.slice(0, i).join(\"\\n\");\n if (parses(prefix)) firstBad = i; // dernier préfixe complet et valide\n}\nconsole.log(\"dernier préfixe valide : ligne\", firstBad);\nconsole.log(\"\\n── contexte ──\");\nfor (let i = Math.max(0, firstBad - 6); i < Math.min(lines.length, firstBad + 14); i++) {\n const mark = i === firstBad ? \" ⟵ ICI\" : \"\";\n console.log(`${String(i + 1).padStart(5)} | ${lines[i].slice(0, 150)}${mark}`);\n}\n",[],[214,415],"6da1f1f4-e2a7-4d94-88cc-de4a075fd233","dump_products_js.ts","01a033f6-8ab0-7225-822f-2eb72e4935b3",["D",1787578190526],"// Affiche le JS transpilé autour des zones que j'ai modifiées dans Products,\n// pour voir ce que le transpileur a réellement produit.\nconst src = await (await fetch(\"https://d2c-france-dashboard.val.run/frontend/App.tsx\")).text();\nconst lines = src.split(\"\\n\");\n\nconst marks = [\"tscroll\", \"stick\", \"bg-white\", \"Columns\"];\nconst hits = new Set();\nlines.forEach((l, i) => {\n if (marks.some((m) => l.includes(m))) {\n for (let d = -2; d <= 2; d++) hits.add(i + d);\n }\n});\nlet last = -99;\nfor (const i of [...hits].sort((a, b) => a - b)) {\n if (i < 0 || i >= lines.length) continue;\n if (i - last > 1) console.log(\" …\");\n console.log(`${String(i + 1).padStart(5)} | ${lines[i].slice(0, 230)}`);\n last = i;\n}\n",[],[214,422],"62663918-5e21-4a1b-9205-944928234680","diag_units.ts","01a033f6-8ab0-7225-822f-28c5946cf922",["D",1787578190526],"// Décomposition fine de l'écart d'unités entre Overview et Business.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst CUT = \"2026-03-01\", TO = \"2026-08-09\";\nconst r0 = (n: number) => Math.round(n).toLocaleString(\"fr-FR\");\n\nconst cat = (await blob.getJSON(\"catalog\")) as any;\nconst ord = (await blob.getJSON(\"orders_lines\")) as any;\nconst products: any[] = cat.products;\nconst lines: any[] = ord.lines;\nconst known = new Set(products.map((p) => String(p.code)));\nconst isSvc = (s: string) => /AES/i.test(s);\n\nconst ovUnits = products.reduce((s, p) => s + (p.sales || 0), 0);\nconst preU = products.reduce((s, p) => s + (p.salesPre || 0), 0);\n\nlet shopGoods = 0, shopSvc = 0;\nfor (const l of lines) {\n if (l.k === \"ord\") continue;\n const d = l.k === \"ref\" ? l.rf : (l.f || l.d);\n if (!d || d < CUT || d > TO) continue;\n const sign = l.k === \"ref\" ? -1 : 1;\n if (isSvc(String(l.sku ?? \"\"))) shopSvc += sign * l.q;\n else shopGoods += sign * l.q;\n}\n\nconsole.log(\"catalogue `sales` (Overview) :\", r0(ovUnits));\nconsole.log(\"catalogue `salesPre` (avant bascule) :\", r0(preU));\nconsole.log(\"Shopify marchandises depuis la bascule :\", r0(shopGoods));\nconsole.log(\"Shopify SERVICES depuis la bascule :\", r0(shopSvc));\nconsole.log();\nconsole.log(\"Business = salesPre + Shopify(marchandises+services) =\",\n r0(preU + shopGoods + shopSvc));\nconsole.log(\"Business hors services =\", r0(preU + shopGoods));\nconsole.log(\"Overview =\", r0(ovUnits));\nconsole.log(\"écart hors services =\", r0(preU + shopGoods - ovUnits));\nconsole.log();\nconsole.log(\"generated_at catalogue :\", cat.generated_at);\nconsole.log(\"orders updated_at :\", ord.updated_at);\nconsole.log(\"nsv_cutover :\", cat.nsv_cutover ?? \"(absent → défaut 2026-03-01)\");\nconsole.log(\"bi_asof (si présent) :\", cat.bi_asof ?? \"—\");\nconsole.log(\"lignes commandes :\", r0(lines.length));\n\n// Taille du payload servi à /api/business — c'est lui qui est téléchargé par CHAQUE\n// onglet qui en a besoin (Overview, Scorecards, Business, Logistic).\nconst bytes = new TextEncoder().encode(JSON.stringify(ord)).length;\nconsole.log(\"poids /api/business :\", (bytes / 1e6).toFixed(1), \"Mo\");\n\n// Le champ catalogue `sales` vient du BI, dont l'extraction est plus ancienne que\n// Shopify. On regarde les unités marchandises Shopify mois par mois pour voir\n// jusqu'où le BI semble s'arrêter.\nconst byMonth: Record = {};\nfor (const l of lines) {\n if (l.k === \"ord\") continue;\n const d = l.k === \"ref\" ? l.rf : (l.f || l.d);\n if (!d || d < CUT) continue;\n if (isSvc(String(l.sku ?? \"\"))) continue;\n const m = d.slice(0, 7);\n byMonth[m] = (byMonth[m] ?? 0) + (l.k === \"ref\" ? -l.q : l.q);\n}\nconsole.log(\"\\nUnités marchandises Shopify par mois :\");\nlet cum = preU;\nfor (const m of Object.keys(byMonth).sort()) {\n cum += byMonth[m];\n console.log(` ${m} : ${String(byMonth[m]).padStart(6)} cumul ${r0(cum)}`);\n}\nconsole.log(` (Overview affiche ${r0(ovUnits)} → le BI semble s'arrêter là)`);\n",[],[214,429],"313e8758-ee9f-44eb-818a-4ee55690a3a1","diag_overview_vs_business.ts","01a033f6-8ab0-7225-822f-2773ecc4cab3",["D",1787578190526],"// Pourquoi Overview et Business ne disent-ils pas la même chose sur la même période ?\n// Lecture seule : on reproduit exactement les deux calculs du frontend.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst FROM = \"2026-01-01\", TO = \"2026-08-09\";\nconst CUT = \"2026-03-01\";\nconst r0 = (n: number) => Math.round(n).toLocaleString(\"fr-FR\");\nconst M = (n: number) => (n / 1e6).toFixed(2) + \" M€\";\n\nconst cat = (await blob.getJSON(\"catalog\")) as any;\nconst ord = (await blob.getJSON(\"orders_lines\")) as any;\nconst products: any[] = cat.products;\nconst lines: any[] = ord.lines;\nconst orphans: any[] = ord.orphan_refunds ?? [];\n\nconst deee: Record = {};\nconst known = new Set();\nfor (const p of products) { deee[String(p.code)] = p.deee ?? 0; known.add(String(p.code)); }\nconst isSvc = (sku: string) => /AES/i.test(sku);\n\n// ── OVERVIEW ────────────────────────────────────────────────────────────────\n// tot.sales = somme du champ catalogue `sales` (donnée pipeline, pas Shopify)\nconst ovUnits = products.reduce((s, p) => s + (p.sales || 0), 0);\nconst hasNsv = products.some((p) => (p.rev ?? 0) > 0);\nconst ovRevFallback = products.reduce((s, p) => s + (p.rev ?? 0), 0);\nconst ovRevPre = products.reduce((s, p) => s + (p.revPre ?? 0), 0);\nconst ovSalesPre = products.reduce((s, p) => s + (p.salesPre ?? 0), 0);\n\n// shippedRevenueBySku(lines, products, CUT, TO) — marchandises Shopify depuis la bascule\nlet shippedGoods = 0;\nfor (const l of lines) {\n if (l.k !== \"ful\" && l.k !== \"ref\") continue;\n const d = l.k === \"ref\" ? l.rf : (l.f || l.d);\n if (!d || d < CUT || d > TO) continue;\n const sku = String(l.sku ?? \"\");\n const amt = (l.n != null ? l.n : l.p * l.q) / 1.2 - (deee[sku] ?? 0) * l.q;\n shippedGoods += (l.k === \"ref\" ? -amt : amt);\n}\nconst seenShip = new Set();\nlet shipDeliv = 0;\nfor (const l of lines) {\n if (l.k !== \"ful\") continue;\n const d = l.f || l.d;\n if (!d || d < CUT || d > TO) continue;\n if (l.ship && !seenShip.has(l.o)) { seenShip.add(l.o); shipDeliv += l.ship / 1.2; }\n}\nconst ovRevOrders = ovRevPre + shippedGoods + shipDeliv;\n\nconsole.log(\"═══ OVERVIEW ═══\");\nconsole.log(\" Sales YTD (u) :\", r0(ovUnits), \" ← champ catalogue `sales`\");\nconsole.log(\" Revenue si commandes OK:\", M(ovRevOrders));\nconsole.log(\" Revenue de repli (p.rev):\", M(ovRevFallback), hasNsv ? \"\" : \"(pas de NSV)\");\nconsole.log(\" → l'écran affiche « Revenue est. » donc c'est le REPLI qui s'affiche\");\n\n// ── BUSINESS ────────────────────────────────────────────────────────────────\nconst biIncluded = FROM < CUT;\nconst shopFrom = biIncluded ? CUT : FROM;\nlet goods = 0, units = 0, refVal = 0, refUnits = 0;\nlet svcUnits = 0, svcVal = 0, unknownUnits = 0;\nfor (const l of lines) {\n if (l.k === \"ord\") continue;\n const d = l.k === \"ref\" ? l.rf : (l.f || l.d);\n if (!d || d < shopFrom || d > TO) continue;\n const sku = String(l.sku ?? \"\");\n const amt = (l.n != null ? l.n : l.p * l.q) / 1.2 - (deee[sku] ?? 0) * l.q;\n if (l.k === \"ful\") { goods += amt; units += l.q; }\n else { refVal += amt; refUnits += l.q; }\n if (isSvc(sku)) { svcUnits += (l.k === \"ref\" ? -l.q : l.q); svcVal += (l.k === \"ref\" ? -amt : amt); }\n else if (!known.has(sku)) unknownUnits += (l.k === \"ref\" ? -l.q : l.q);\n}\nconst seen2 = new Set();\nlet deliv = 0;\nfor (const l of lines) {\n if (l.k !== \"ful\") continue;\n const d = l.f || l.d;\n if (!d || d < shopFrom || d > TO) continue;\n if (l.ship && !seen2.has(l.o)) { seen2.add(l.o); deliv += l.ship / 1.2; }\n}\nconst orphVal = orphans.filter((o) => o.rf >= shopFrom && o.rf <= TO)\n .reduce((s, o) => s + (o.amt ?? 0) / 1.2, 0);\nconst preBi = biIncluded ? ovRevPre : 0;\nconst preU = biIncluded ? ovSalesPre : 0;\nconst busCa = goods - refVal - orphVal + deliv + preBi;\nconst busUnits = units - refUnits + preU;\n\nconsole.log(\"\\n═══ BUSINESS ═══\");\nconsole.log(\" Revenue :\", M(busCa));\nconsole.log(\" Units sold :\", r0(busUnits));\nconsole.log(\" dont part BI :\", r0(preU), \"u /\", M(preBi));\nconsole.log(\" dont Shopify net :\", r0(units - refUnits), \"u\");\nconsole.log(\" dont SERVICES (AES) :\", r0(svcUnits), \"u /\", M(svcVal));\nconsole.log(\" dont SKU inconnus :\", r0(unknownUnits), \"u\");\nconsole.log(\" Livraison :\", M(deliv));\nconsole.log(\" Avoirs sans détail :\", M(-orphVal));\n\nconsole.log(\"\\n═══ ÉCARTS ═══\");\nconsole.log(\" CA : Business\", M(busCa), \"vs Overview\", M(ovRevFallback),\n \"→\", M(busCa - ovRevFallback));\nconsole.log(\" Unités: Business\", r0(busUnits), \"vs Overview\", r0(ovUnits),\n \"→\", r0(busUnits - ovUnits));\nconsole.log(\"\\n Si Overview utilisait les commandes :\", M(ovRevOrders),\n \"→ écart résiduel\", M(busCa - ovRevOrders));\nconsole.log(\" (cet écart résiduel = les avoirs sans détail, déduits par Business seul)\");\n\nconsole.log(\"\\n═══ APRÈS CORRECTION — ce que les écrans doivent afficher ═══\");\nconsole.log(\" Overview Revenue YTD :\", M(ovRevOrders), \" (plus de repli silencieux)\");\nconsole.log(\" Business Revenue :\", M(busCa));\nconsole.log(\" → écart attendu :\", M(busCa - ovRevOrders), \"= avoirs sans détail\");\nconsole.log();\nconsole.log(\" Business Units (appareils):\", r0(busUnits - svcUnits));\nconsole.log(\" Business Services :\", r0(svcUnits), \"/\", M(svcVal), \"de CA\");\nconsole.log(\" Overview Sales YTD (u) :\", r0(ovUnits), \" (base BI, plus ancienne)\");\nconsole.log(\" → écart appareils :\", r0(busUnits - svcUnits - ovUnits),\n \"= ventes postérieures à l'extraction BI\");\n",[],[214,436],"8946704c-a828-466d-aeee-8c1c3e63f433","check_frontend.ts","01a033f6-8ab0-7225-822f-21b8deba1891",["D",1787578190526],"// Contrôle de syntaxe des modules frontend tels qu'ils sont RÉELLEMENT servis.\n// Une page blanche vient d'une erreur au chargement d'un module : on récupère donc\n// le JS transpilé et on tente de le parser, exactement comme le ferait le navigateur.\nconst FILES = [\n \"App.tsx\", \"Simulator.tsx\", \"Parameters.tsx\", \"Forecast.tsx\",\n \"helpers.ts\", \"Icons.tsx\", \"pricing.ts\", \"forecasting.ts\", \"index.tsx\",\n];\nconst BASE = \"https://d2c-france-dashboard.val.run/frontend/\";\n\nfor (const f of FILES) {\n const res = await fetch(BASE + f);\n const src = await res.text();\n let verdict = \"OK\";\n try {\n // new Function ne valide pas les import/export ; on les neutralise d'abord.\n const stripped = src\n .replace(/^\\s*import[\\s\\S]*?from\\s+[\"'][^\"']+[\"'];?/gm, \"\")\n .replace(/^\\s*import\\s+[\"'][^\"']+[\"'];?/gm, \"\")\n .replace(/^\\s*export\\s+default\\s+/gm, \"const __d = \")\n .replace(/^\\s*export\\s+/gm, \"\");\n new Function(stripped);\n } catch (e) {\n verdict = \"ERREUR DE SYNTAXE — \" + (e as Error).message;\n }\n console.log(`${res.status} ${f.padEnd(18)} ${(src.length / 1024).toFixed(0).padStart(4)} Ko ${verdict}`);\n}\n",[],[214,443],"06b22935-e015-4e03-af85-ff6815fdc7af","check_forecast_detail.ts","01a033f6-8ab0-7225-822f-1d94dedd5048",["D",1787578190526],"// Rejoue ce que fait le panneau de détail de Forecast, pour CHAQUE référence.\n// Objectif : garantir qu'aucun SKU ne peut plus provoquer d'écran blanc, et pas\n// seulement celui sur lequel on a cliqué par hasard.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst dp = (await blob.getJSON(\"forecast_dp\")) as any;\nif (!dp) {\n console.log(\"Pas de fichier Demand Planning — rien à vérifier.\");\n} else {\n const HORIZON = 6;\n\n // Reconstitution de `split`, à l'identique du composant.\n let last = -1;\n for (let i = 0; i < dp.months.length; i++) {\n if (dp.skus.some((s: any) => (s.hist[i] ?? 0) > 0)) last = i;\n }\n const target: string[] = dp.months.slice(last + 1, last + 1 + HORIZON);\n const targetIdx = target.map((_, i) => last + 1 + i);\n console.log(\"mois d'historique :\", dp.months.length, \"| dernier réel :\", dp.months[last]);\n console.log(\"mois à prévoir :\", target.join(\", \"));\n console.log(\"références :\", dp.skus.length);\n\n // Le panneau lit, pour une référence NON nouvelle :\n // r.s.avail[targetIdx[i]] et r.final[i] ← r.final vient de `filtered`\n // On vérifie que ces deux accès sont sûrs pour toutes les références.\n let noAvail = 0, shortAvail = 0, ok = 0;\n const problems: string[] = [];\n for (const s of dp.skus) {\n if (!Array.isArray(s.avail)) { noAvail++; problems.push(`${s.code}: avail absent`); continue; }\n for (const idx of targetIdx) {\n if (idx >= s.avail.length) { shortAvail++; break; }\n }\n // `final` est toujours de longueur HORIZON puisqu'il dérive de `proposal`,\n // lui-même construit sur HORIZON. On le vérifie tout de même côté longueurs.\n if (!Array.isArray(s.hist) || s.hist.length === 0) {\n problems.push(`${s.code}: hist vide → le graphique d'historique serait vide`);\n }\n ok++;\n }\n console.log(\"\\nRéférences sans tableau `avail` :\", noAvail);\n console.log(\"Références dont `avail` est plus court:\", shortAvail,\n \"(accès hors borne → undefined, neutralisé par ?? 0)\");\n console.log(\"Références exploitables :\", ok);\n if (problems.length) {\n console.log(\"\\nPoints d'attention (max 10) :\");\n for (const p of problems.slice(0, 10)) console.log(\" \", p);\n } else {\n console.log(\"\\n✔ Aucun accès dangereux détecté dans le panneau de détail.\");\n }\n}\n",[],[214,450],"b73d905e-afc4-4457-9922-8e7f2fe5ad75","check_dangling_refs.ts","01a033f6-8ab0-7225-822f-1abc40b80a2a",["D",1787578190526],"// Une variable supprimée mais encore référencée passe TOUS les contrôles de syntaxe\n// et d'import : l'erreur ne survient qu'au rendu du composant concerné, donc sur un\n// onglet précis, souvent après la livraison. On la traque ici dans le JS transpilé.\nconst BASE = \"https://nicolasduchene--019fe6af333e71cca0781e0c1873f17b.web.val.run/frontend/\";\nconst js = await (await fetch(BASE + \"App.tsx\")).text();\n\n// Le panneau de détail de Forecast doit puiser dans `filtered` (qui porte `final`)\n// et jamais dans `rows`. C'est cette confusion qui provoquait l'écran blanc au clic\n// sur un SKU ; un contrôle figé évite qu'elle revienne à la faveur d'une refonte.\n{\n const fc = await (await fetch(BASE + \"Forecast.tsx\")).text();\n const bad = /rows\\.find\$\\s*\\(?\\s*x\\s*\$?\\s*=>\\s*x\\.s\\.code === detail/.test(fc);\n const good = /filtered\\.find\$\\s*\\(?\\s*x\\s*\$?\\s*=>\\s*x\\.s\\.code === detail/.test(fc);\n console.log(`${good && !bad ? \"✔\" : \"✘\"} Forecast : détail lu depuis ${good ? \"filtered\" : \"rows\"}`);\n if (!good || bad) console.log(\" ✘ le panneau de détail retomberait en écran blanc\");\n}\n\n// Identifiants retirés lors de la mutualisation du chargement des commandes.\n// `setData` n'est PAS dans cette liste : App() conserve un état `data` pour le\n// catalogue (/api/data), qui n'a rien à voir avec les commandes. Le compter comme\n// orphelin déclencherait une fausse alerte à chaque exécution.\nconst REMOVED = [\"setOrders\"];\nlet bad = 0;\nfor (const id of REMOVED) {\n const re = new RegExp(`\\\\b${id}\\\\b`, \"g\");\n const n = (js.match(re) ?? []).length;\n console.log(`${n === 0 ? \"✔\" : \"✘\"} ${id.padEnd(12)} ${n} référence(s) restante(s)`);\n if (n > 0) bad++;\n}\n\n// Il ne doit plus rester qu'UN seul appel à /api/business dans tout le bundle :\n// celui du magasin mutualisé. Chaque appel supplémentaire vaut 7 Mo téléchargés.\nconst nCalls = (js.match(/[\"'`]\\/api\\/business[\"'`]/g) ?? []).length;\nconsole.log(`${nCalls === 0 ? \"✔\" : \"✘\"} appels /api/business dans App.tsx : ${nCalls} (attendu 0, tout passe par ordersStore)`);\nif (nCalls > 0) bad++;\n\n// Toute variable doit être définie avant usage dans sa fonction. On vérifie les\n// nouvelles pour s'assurer qu'elles sont bien déclarées ET utilisées.\nconst EXPECTED = [\"useOrders\", \"svcUnits\", \"svcRevenue\", \"deviceUnitsFul\", \"ordersMeta\"];\nfor (const id of EXPECTED) {\n const n = (js.match(new RegExp(`\\\\b${id}\\\\b`, \"g\")) ?? []).length;\n console.log(`${n >= 2 ? \"✔\" : \"✘\"} ${id.padEnd(12)} ${n} occurrence(s) (déclaration + usage)`);\n if (n < 2) bad++;\n}\n\n// `refundUnits` n'alimente plus le compteur d'unités : soit il sert encore, soit il\n// ne sert plus — dans les deux cas ce n'est pas une erreur, on l'affiche pour info.\nconst ru = (js.match(/\\brefundUnits\\b/g) ?? []).length;\nconsole.log(`ℹ refundUnits ${ru} occurrence(s)`);\n\nconsole.log(bad === 0 ? \"\\n✔ Aucune référence orpheline.\" : `\\n✘ ${bad} problème(s).`);\n",[],[214,457],"d57d5146-d63c-4852-a8d6-2c924ab8138c","check_branch.ts","01a033f6-8ab0-7225-822f-15ce38b9b78b",["D",1787578190526],"// Contrôle avant fusion, exécuté sur l'URL de la BRANCHE et non sur la production.\n// Deux niveaux : le module se parse-t-il, puis s'importe-t-il réellement ?\n// C'est ce second niveau qui attrape les exports manquants et les imports cassés.\n// ⚠️ LIMITE CONNUE de ces contrôles : ils valident la syntaxe et le chargement des\n// modules, PAS le rendu. Une violation de la règle des hooks — un `useMemo` placé\n// après un `return` conditionnel — produit du code parfaitement valide qui ne casse\n// qu'à l'exécution, et seulement une fois les données arrivées. C'est ainsi que\n// l'onglet Amazon est tombé en erreur #310 malgré des contrôles au vert.\n// Le filet reste ErrorBoundary : la section fautive s'isole au lieu de tout emporter.\n// Règle à respecter à l'écriture : TOUS les hooks avant TOUT return conditionnel.\nconst BASE = \"https://nicolasduchene--01a033b8544477bab117977f84cc969b.web.val.run/frontend/\";\nconst MODS = [\"helpers.ts\", \"Icons.tsx\", \"pricing.ts\", \"forecasting.ts\", \"ordersStore.ts\",\n \"ErrorBoundary.tsx\", \"Amazon.tsx\",\n \"Simulator.tsx\", \"Parameters.tsx\", \"Forecast.tsx\", \"App.tsx\", \"index.tsx\"];\n\nlet ko = 0;\nfor (const m of MODS) {\n const res = await fetch(BASE + m);\n const src = await res.text();\n let syntax = \"OK\";\n try {\n new Function(src\n .replace(/^\\s*import[\\s\\S]*?from\\s+[\"'][^\"']+[\"'];?/gm, \"\")\n .replace(/^\\s*import\\s+[\"'][^\"']+[\"'];?/gm, \"\")\n .replace(/^\\s*export\\s+default\\s+/gm, \"const __d = \")\n .replace(/^\\s*export\\s+/gm, \"\"));\n } catch (e) {\n syntax = \"SYNTAXE ✘ \" + (e as Error).message.slice(0, 80);\n ko++;\n }\n let load = \"—\";\n // index.tsx monte l'application sur `document` dès son évaluation : il ne PEUT pas\n // s'importer hors navigateur. On se limite donc au contrôle de syntaxe pour lui.\n if (syntax === \"OK\" && m !== \"index.tsx\") {\n try {\n const mod = await import(BASE + m);\n load = `${Object.keys(mod).length} exports`;\n } catch (e) {\n load = \"IMPORT ✘ \" + (e as Error).message.slice(0, 90);\n ko++;\n }\n }\n console.log(`${res.status} ${m.padEnd(17)} ${(src.length / 1024).toFixed(0).padStart(4)} Ko ${syntax.padEnd(12)} ${load}`);\n}\nconsole.log(ko === 0 ? \"\\n✔ Branche saine — fusion possible.\" : `\\n✘ ${ko} problème(s) — NE PAS FUSIONNER.`);\n",[],[214,464],"89d76e5a-4f2d-46cf-a5dc-cf8faf230282","check_amazon_nsv.ts","01a033f6-8ab0-7225-822f-138afcce847b",["D",1787578190526],"// Valeurs attendues du NSV, calculées hors interface, pour comparer à l'écran.\n// NSV = sell-in − commission contractuelle − SOA. Le SOA étant un simulateur non\n// enregistré, il vaut 0 ici : le NSV affiché doit donc correspondre à « après\n// commission » tant qu'aucune simulation n'est saisie.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst d = (await blob.getJSON(\"amazon\")) as any;\nconst ps: any[] = d.products;\nconst eur = (x: number) => (x / 1e6).toFixed(3) + \" M€\";\nconst pct = (x: number | null) => x == null ? \"—\" : (x * 100).toFixed(1) + \" %\";\n\nfor (const period of Object.keys(d.sellout_periods ?? {}).sort()) {\n const sold = ps.filter((p) => p.sellout?.[period]);\n let sellIn = 0, comm = 0, nsv = 0, noContract = 0;\n const byCat: Record = {};\n for (const p of sold) {\n const cogs = p.sellout[period].shippedCogs ?? 0;\n sellIn += cogs;\n const k = p.grp ? d.contracts?.[p.grp]?.comm : null;\n if (k == null || !cogs) { noContract += cogs; continue; }\n const c = cogs * k;\n comm += c; nsv += cogs - c;\n const e = (byCat[p.grp] ??= { sellIn: 0, comm: 0, nsv: 0 });\n e.sellIn += cogs; e.comm += c; e.nsv += cogs - c;\n }\n const covered = sellIn - noContract;\n console.log(`\\n═══ ${period} ═══`);\n console.log(` Sell-in brut ${eur(sellIn)}`);\n console.log(` dont sans contrat ${eur(noContract)} ← exclu du NSV`);\n console.log(` Sell-in couvert ${eur(covered)}`);\n console.log(` Commission ${eur(comm)}`);\n console.log(` NSV (avant SOA) ${eur(nsv)}`);\n console.log(` Taux NSV ${pct(covered > 0 ? nsv / covered : null)}`);\n console.log(\" ── par catégorie ──\");\n for (const [c, v] of Object.entries(byCat).sort((a, b) => b[1].sellIn - a[1].sellIn)) {\n const rate = d.contracts?.[c]?.comm;\n console.log(` ${c.padEnd(12)} sell-in ${eur(v.sellIn)} comm ${pct(rate)}`\n + ` → NSV ${eur(v.nsv)}`);\n }\n}\n\n// Contrôle : le taux NSV doit être exactement 1 − commission pondérée.\nconsole.log(\"\\nContrôle de cohérence : NSV = sell-in × (1 − taux de commission).\");\nconsole.log(\"Les catégories ayant des taux différents (SDA 22 %, MDA/MW 19 %, TV 17 %),\");\nconsole.log(\"le taux global dépend du mix — il doit se situer entre 78 % et 83 %.\");\n",[],[214,471],"e664a119-9dc5-4fa9-b883-536588647e64","check_amazon.ts","01a033f6-8ab0-7225-822f-0ea1db1726aa",["D",1787578190526],"// Contrôle du blob Amazon : reproduit EXACTEMENT les agrégats de l'onglet, pour\n// vérifier que ce qui s'affiche est cohérent avec ce qui a été envoyé.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst d = (await blob.getJSON(\"amazon\")) as any;\nif (!d) {\n console.log(\"❌ Aucun blob `amazon`.\");\n} else {\n const ps: any[] = d.products;\n const eur = (x: number) => (x / 1e6).toFixed(2) + \" M€\";\n const n0 = (x: number) => Math.round(x).toLocaleString(\"fr-FR\");\n\n console.log(\"généré le :\", d.generated_at);\n console.log(\"catalogue :\", d.source_file);\n console.log(\"périodes :\", Object.keys(d.sellout_periods ?? {}).join(\", \"));\n console.log(\"références :\", ps.length);\n console.log(\"hors catalogue :\", ps.filter((p) => p.offCatalogue).length);\n console.log(\"avec sell-out :\", ps.filter((p) => p.sellout).length);\n\n for (const period of Object.keys(d.sellout_periods ?? {}).sort()) {\n const sold = ps.filter((p) => p.sellout?.[period]);\n let sellIn = 0, outRev = 0, outU = 0, shipU = 0, unf = 0, ret = 0, ord = 0, conf = 0;\n let npmW = 0, npmB = 0, mnW = 0, mnB = 0;\n for (const p of sold) {\n const s = p.sellout[period];\n const c = s.shippedCogs ?? 0;\n sellIn += c; outRev += s.orderedRev ?? 0; outU += s.orderedUnits ?? 0;\n shipU += s.shippedUnits ?? 0; unf += s.unfilled ?? 0; ret += s.returns ?? 0;\n ord += s.orderedUnits ?? 0; conf += s.confirmedUnits ?? 0;\n if (s.netPpm != null && c > 0) { npmW += s.netPpm * c; npmB += c; }\n if (p.mnPct != null && c > 0) { mnW += p.mnPct * c; mnB += c; }\n }\n console.log(`\\n═══ ${period} — ${sold.length} références vendues ═══`);\n console.log(` Sell-in (Amazon nous paie) : ${eur(sellIn)}`);\n console.log(` Sell-out (prix public) : ${eur(outRev)} — ${n0(outU)} u`);\n console.log(` Notre marge nette pondérée : ${(mnB > 0 ? mnW / mnB * 100 : 0).toFixed(1)} %`);\n console.log(` Net PPM Amazon pondéré : ${(npmB > 0 ? npmW / npmB : 0).toFixed(1)} %`);\n console.log(` Taux de confirmation : ${(ord > 0 ? conf / ord * 100 : 0).toFixed(1)} %`);\n console.log(` Retours : ${(shipU > 0 ? ret / shipU * 100 : 0).toFixed(1)} % (${n0(ret)} u)`);\n console.log(` Demande non servie : ${n0(unf)} u`);\n\n // Les quatre listes d'opportunités, avec les mêmes seuils que l'onglet\n const lowConf = sold.filter((p) => (p.sellout[period].confirmPct ?? 100) < 60\n && (p.sellout[period].orderedUnits ?? 0) >= 50);\n const unfilled = sold.filter((p) => (p.sellout[period].unfilled ?? 0) > 0);\n const lowMrg = sold.filter((p) => p.mnPct != null && p.mnPct < 0.10\n && (p.sellout[period].shippedCogs ?? 0) > 10000);\n const rrpTrap = sold.filter((p) => p.rrp && p.po && p.rrp / p.po < 1.2);\n console.log(` Opportunités : confirmation ${lowConf.length} · rupture ${unfilled.length}`\n + ` · marge faible ${lowMrg.length} · RRP bas ${rrpTrap.length}`);\n }\n\n // Contrôles de qualité — ce qui apparaîtra vide à l'écran\n console.log(\"\\n═══ QUALITÉ DES DONNÉES ═══\");\n const noPo = ps.filter((p) => !p.po).length;\n const noCost = ps.filter((p) => !p.scti).length;\n const noRrp = ps.filter((p) => !p.rrp).length;\n const noMn = ps.filter((p) => p.mnPct == null).length;\n console.log(` sans PO ${noPo} · sans coût standard ${noCost} · sans RRP ${noRrp} · sans marge ${noMn}`);\n const grps: Record = {};\n for (const p of ps) grps[p.grp ?? \"(sans catégorie)\"] = (grps[p.grp ?? \"(sans catégorie)\"] ?? 0) + 1;\n console.log(\" par catégorie :\", JSON.stringify(grps));\n}\n",[],[214,478],"f798c536-2d5d-4888-8c11-ff9cff558dd5","boot_check.ts","01a033f6-8ab0-7225-822f-0a2ce10779fe",["D",1787578190526],"// Charge vraiment les modules frontend, comme le ferait le navigateur.\n// Plus fort qu'un contrôle de syntaxe : détecte aussi les imports cassés et les\n// erreurs qui surviennent à l'évaluation du module (export manquant, référence\n// à une icône inexistante, etc.) — les causes classiques d'une page blanche.\nconst BASE = \"https://d2c-france-dashboard.val.run/frontend/\";\nconst MODS = [\"helpers.ts\", \"Icons.tsx\", \"pricing.ts\", \"forecasting.ts\",\n \"Simulator.tsx\", \"Parameters.tsx\", \"Forecast.tsx\", \"App.tsx\"];\n\nlet ko = 0;\nfor (const m of MODS) {\n try {\n const mod = await import(BASE + m);\n const keys = Object.keys(mod);\n const dflt = \"default\" in mod ? typeof mod.default : \"—\";\n console.log(`✔ ${m.padEnd(17)} ${String(keys.length).padStart(3)} exports · default: ${dflt}`);\n } catch (e) {\n ko++;\n console.log(`✘ ${m.padEnd(17)} ${(e as Error).message.slice(0, 200)}`);\n }\n}\n\n// Les icônes réellement utilisées par App doivent exister : une icône manquante\n// se traduit par « type is invalid » au rendu, donc par un écran blanc.\nconst Ico = await import(BASE + \"Icons.tsx\");\nconst NEEDED = [\"IconMenu\", \"IconClose\", \"IconAlert\", \"IconReturn\", \"IconEuro\",\n \"IconColumns\", \"IconSave\", \"IconPrint\", \"IconDownload\", \"IconLock\", \"IconLogout\"];\nconst missing = NEEDED.filter((n) => typeof (Ico as any)[n] !== \"function\");\nconsole.log(missing.length === 0\n ? `\\n✔ Les ${NEEDED.length} icônes attendues sont exportées.`\n : `\\n✘ Icônes manquantes : ${missing.join(\", \")}`);\n\nconsole.log(ko === 0 && missing.length === 0\n ? \"\\n✔ Tous les modules se chargent — la page doit s'afficher.\"\n : \"\\n✘ Chargement en échec.\");\n",[],[214,485],"8c56e253-29b3-4508-a4b0-c01cb457c149","bi_july_data.ts","01a033f6-8ab0-7225-822f-05f5f915b18b",["D",1787578190526],"// Agrégat BI juillet 2026 par produit — [NSV EUR, ACT Qty]. Donnée figée, issue de\n// l'export \"Billing Document Date\" de juillet. Sert uniquement au rapprochement.\nexport const BI_JULY: Record = {\"31010268\":[653.15,1],\"31010279\":[-403.81,-1],\"31010282\":[808.95,3],\"31010285\":[606.73,2],\"31010294\":[5260.61,19],\"31010312\":[273.15,1],\"31010538\":[320.48,1],\"31010662\":[369.64,1],\"31011106\":[361.4,1],\"31011142\":[0.0,0],\"31011242\":[1161.79,2],\"31011594\":[649.16,2],\"31011595\":[6479.26,15],\"31018659\":[7337.45,18],\"31018660\":[1309.2,3],\"31018810\":[370.48,1],\"31018812\":[428.81,1],\"31018818\":[3833.56,14],\"31018821\":[14019.56,57],\"31018853\":[1977.39,5],\"31018882\":[5663.1,22],\"31019058\":[10259.84,36],\"31019061\":[5008.19,15],\"31019064\":[718.46,2],\"31019074\":[1084.2,3],\"31019077\":[9165.16,28],\"31019192\":[2799.67,8],\"31019208\":[353.81,1],\"31019437\":[20728.83,58],\"31019438\":[2455.72,5],\"31019586\":[1082.62,2],\"31019589\":[9678.55,26],\"31019591\":[5765.36,17],\"31019592\":[24984.59,66],\"31019691\":[1561.36,6],\"31019798\":[9627.34,22],\"31019971\":[528.46,2],\"31020057\":[863.02,3],\"31020096\":[557.62,2],\"31020099\":[17423.88,75],\"31020147\":[624.65,1],\"31020158\":[527.98,1],\"31020162\":[-261.37,-1],\"31020170\":[16872.26,78],\"31020177\":[6077.25,23],\"31020178\":[10880.36,43],\"31020185\":[1089.43,4],\"31020191\":[319.65,1],\"31020201\":[3420.41,13],\"31020203\":[1818.17,6],\"31020207\":[922.27,3],\"31020208\":[30415.44,126],\"31020242\":[18758.98,80],\"31020243\":[5946.42,22],\"31020257\":[15342.85,55],\"31020288\":[-586.77,-2],\"31020398\":[13986.33,56],\"31020405\":[-287.14,-1],\"31020428\":[346.41,1],\"31020431\":[384.65,1],\"31020438\":[605.96,2],\"31020456\":[2885.01,10],\"31020480\":[1274.43,4],\"31020494\":[4141.72,11],\"31020500\":[587.15,1],\"31020501\":[694.65,1],\"31020571\":[2049.36,8],\"31020574\":[5382.19,22],\"31020577\":[1238.6,4],\"31020581\":[1038.12,3],\"31020591\":[3245.7,3],\"31020601\":[19623.04,70],\"31020605\":[2135.74,5],\"31020607\":[2165.26,4],\"31020608\":[2599.58,7],\"31020648\":[21565.12,92],\"31020650\":[2809.06,12],\"31020691\":[2430.52,8],\"31020712\":[24868.66,112],\"31020713\":[-219.65,-1],\"31020748\":[324.65,1],\"31020758\":[5948.72,16],\"31020761\":[2572.89,6],\"31020774\":[13762.56,37],\"31020818\":[486.32,1],\"31020833\":[366.32,1],\"31020841\":[2363.86,8],\"31020865\":[587.07,1],\"31020868\":[1041.45,3],\"31020889\":[242.15,1],\"31020907\":[30256.17,81],\"31020921\":[414.65,1],\"31020940\":[653.55,2],\"31020983\":[683.91,1],\"31021046\":[12324.51,44],\"31021070\":[16114.29,58],\"31021190\":[820.48,1],\"31021208\":[3957.05,12],\"31021346\":[811.88,2],\"31102144\":[2697.19,9],\"31102175\":[3141.58,12],\"31102188\":[3023.93,10],\"31102189\":[1455.74,5],\"31102190\":[234.65,1],\"31102198\":[524.86,1],\"31102233\":[1061.43,3],\"31102244\":[708.11,3],\"31102424\":[1759.53,6],\"31102514\":[8205.7,31],\"31102854\":[-269.17,-1],\"31102872\":[-520.4,-1],\"31102935\":[2226.93,3],\"31102960\":[1097.8,2],\"31102963\":[652.98,1],\"31102968\":[11417.28,26],\"31103004\":[1407.86,4],\"31103013\":[486.32,1],\"31103016\":[842.8,2],\"31103018\":[427.98,1],\"31103028\":[607.63,2],\"31103053\":[647.07,1],\"31103063\":[1941.54,5],\"31103069\":[-777.74,-2],\"31103103\":[655.13,2],\"31103108\":[461.32,1],\"31103109\":[337.15,1],\"31103135\":[737.15,1],\"31700001\":[1655.87,1],\"31801013\":[-286.32,-1],\"31801017\":[828.73,1],\"31801163\":[952.98,1],\"31801220\":[436.4,1],\"31801245\":[2853.3,8],\"31900556\":[669.65,1],\"32002327\":[1053.59,4],\"32002331\":[894.78,3],\"32002367\":[2609.19,12],\"32002374\":[270.48,1],\"32002375\":[952.27,3],\"32002385\":[278.81,1],\"32002392\":[295.48,1],\"32002427\":[2334.32,5],\"32002471\":[751.78,3],\"32002472\":[2554.27,9],\"32002477\":[1003.46,2],\"32002505\":[6507.04,25],\"32002543\":[974.28,2],\"32002571\":[1808.37,7],\"32002576\":[-195.48,-1],\"32002580\":[11716.19,53],\"32002583\":[317.15,1],\"32002595\":[-610.42,-1],\"32002600\":[620.48,1],\"32002608\":[3027.45,12],\"32002631\":[437.15,1],\"32002641\":[529.6,2],\"32002646\":[745.48,1],\"32002663\":[362.13,1],\"32901423\":[640.96,2],\"32901457\":[2143.38,7],\"32901576\":[2031.66,7],\"32901654\":[232.15,1],\"32901672\":[1399.29,2],\"32901690\":[512.07,1],\"32901695\":[279.65,1],\"32901701\":[1569.07,5],\"32901707\":[831.43,3],\"32901720\":[1856.21,6],\"32901818\":[23144.03,91],\"32901820\":[799.3,2],\"32901830\":[2151.26,4],\"32901845\":[2780.17,9],\"32901963\":[487.14,1],\"33702187\":[-126.2,-1],\"33703490\":[-30.0,0],\"33703557\":[1612.37,8],\"33703812\":[1701.19,4],\"33703813\":[1175.6,2],\"33703817\":[523.92,2],\"33703818\":[354.46,1],\"33703836\":[1338.98,5],\"33703856\":[621.43,2],\"33703881\":[295.3,1],\"33703887\":[1874.4,3],\"33703899\":[512.8,1],\"33703901\":[227.8,1],\"33704006\":[642.26,2],\"33704007\":[312.79,1],\"33704008\":[321.14,1],\"33704037\":[896.73,3],\"33704155\":[437.8,1],\"33704216\":[5213.3,24],\"33704218\":[246.11,1],\"33704338\":[217.87,1],\"33704386\":[558.93,2],\"33704389\":[191.97,1],\"33704460\":[220.3,1],\"33801781\":[471.88,4],\"33801788\":[1094.48,8],\"33801793\":[303.29,3],\"33801936\":[153.59,1],\"33801976\":[965.39,13],\"33801977\":[2225.54,27],\"33802769\":[525.88,4],\"33802882\":[309.53,2],\"33802962\":[662.64,7],\"33802992\":[3239.52,23],\"33802995\":[176.08,1],\"33803110\":[-495.67,-1],\"33803131\":[1269.81,9],\"33803209\":[146.09,1],\"33803210\":[718.76,8],\"33803262\":[548.07,5],\"33803265\":[11289.19,104],\"33803291\":[71.09,1],\"33803295\":[1274.69,2],\"33803347\":[58974.46,462],\"33803348\":[246.09,1],\"33803484\":[942.58,6],\"34003941\":[2862.17,7],\"34004089\":[-296.47,-1],\"34004090\":[774.75,1],\"34004092\":[649.83,1],\"34004214\":[290.25,1],\"34004216\":[5941.52,16],\"34004283\":[39587.26,107],\"34004284\":[-470.67,-1],\"34004290\":[1736.06,3],\"34004291\":[6395.07,18],\"34004491\":[4879.34,9],\"34004840\":[1353.16,5],\"34004944\":[16670.34,23],\"34004945\":[3297.24,3],\"34004965\":[6893.75,5],\"34004991\":[542.3,1],\"34004992\":[2248.65,3],\"34005004\":[1057.33,1],\"34005042\":[5974.3,10],\"34005044\":[874.0,1],\"34005147\":[11719.23,6],\"34005155\":[4260.97,6],\"34005176\":[1532.99,2],\"34005177\":[2979.57,4],\"34005233\":[2266.15,3],\"34005268\":[2674.31,4],\"34005275\":[32385.77,43],\"34005277\":[6148.8,7],\"34005279\":[4857.48,5],\"34005285\":[1391.41,1],\"34005287\":[8169.63,11],\"34005295\":[3927.24,2],\"34005298\":[3799.32,4],\"34005412\":[6575.94,13],\"34005437\":[30683.1,51],\"34005467\":[2499.15,5],\"34005469\":[9769.88,13],\"34005516\":[9146.01,44],\"34005521\":[208.6,1],\"34005536\":[1232.31,5],\"34005539\":[592.93,2],\"34005567\":[-228.92,-1],\"34005570\":[1830.65,10],\"34005571\":[12327.08,55],\"34005608\":[7722.79,47],\"34005637\":[513.05,1],\"34005638\":[2019.2,4],\"34005644\":[4109.14,8],\"34005651\":[1126.56,6],\"34005657\":[18501.33,69],\"34005699\":[4958.64,8],\"34005702\":[5417.4,8],\"34005704\":[6101.31,7],\"34005711\":[3186.97,3],\"34005713\":[5786.46,6],\"34005716\":[71552.59,176],\"34005726\":[2553.65,5],\"34005729\":[1941.2,4],\"34005731\":[4831.15,5],\"34005746\":[12357.17,37],\"34005752\":[978.98,3],\"34005757\":[429.79,1],\"34005762\":[-284.39,-1],\"34005763\":[438.05,1],\"34005773\":[583.18,1],\"34005798\":[2757.32,5],\"34005803\":[4497.26,10],\"34005805\":[976.16,3],\"34005806\":[989.2,3],\"34005808\":[987.18,2],\"34005819\":[18087.89,60],\"34005823\":[18056.36,27],\"34005824\":[19488.23,30],\"34005825\":[13198.6,33],\"34005851\":[6522.76,9],\"34005853\":[35906.88,69],\"34005854\":[-1430.4,-3],\"34005876\":[34563.28,148],\"34005877\":[654.82,2],\"34005882\":[3597.88,26],\"34005915\":[1131.33,2],\"34005917\":[6314.88,11],\"34005925\":[4330.58,10],\"34005926\":[1769.19,4],\"34005933\":[7945.54,25],\"34005937\":[14075.53,12],\"34005939\":[874.08,1],\"34005940\":[2114.66,2],\"34005941\":[1510.39,2],\"34005944\":[6345.7,13],\"34005963\":[521.46,1],\"34005965\":[462.3,1],\"34005967\":[9247.55,10],\"34005971\":[2949.66,2],\"34006007\":[7262.64,20],\"34006013\":[304.44,1],\"34006014\":[1251.08,4],\"34006049\":[-652.34,-1],\"34006058\":[5251.13,8],\"34006060\":[-637.27,-1],\"34006105\":[757.94,2],\"34006119\":[-541.43,-1],\"34006122\":[287.03,1],\"34006185\":[751.26,2],\"34006188\":[8968.26,10],\"34006189\":[6189.15,5],\"34006220\":[974.82,1],\"34006225\":[1308.16,1],\"34901202\":[221.1,1],\"34901207\":[-362.76,-1],\"34901246\":[592.92,2],\"34901253\":[288.13,1],\"34901431\":[11741.81,38],\"34901570\":[1178.28,3],\"34901578\":[2286.99,3],\"34901586\":[27812.0,87],\"34901588\":[1377.15,3],\"34901590\":[9539.62,28],\"34901601\":[6606.68,29],\"34901611\":[965.51,4],\"34901616\":[216.49,1],\"34901636\":[2423.88,10],\"34901638\":[5601.66,16],\"34901661\":[3074.49,3],\"34901673\":[5301.84,24],\"34901689\":[15353.97,68],\"34901706\":[459.05,1],\"34901731\":[1316.69,4],\"34901756\":[6762.28,15],\"34901770\":[3160.24,5],\"34901772\":[15114.6,29],\"35600392\":[77.94,9],\"35600536\":[43.3,5],\"35601053\":[5.83,1],\"35601069\":[58.61,8],\"35601288\":[54.04,6],\"35601289\":[114.16,20],\"35601650\":[189.89,23],\"35601663\":[29.98,3],\"35601706\":[767.81,52],\"35601724\":[199.87,24],\"35601729\":[459.98,14],\"35601735\":[149.42,17],\"35601760\":[24.98,3],\"35601767\":[53.94,6],\"35601865\":[163.69,19],\"35601868\":[34.4,5],\"35601927\":[22.43,1],\"35602001\":[5.17,1],\"35602039\":[53.33,1],\"35602061\":[23.98,2],\"35602146\":[25.98,3],\"35602165\":[34.39,5],\"35602206\":[32.25,3],\"35602207\":[748.27,25],\"35602240\":[33.32,2],\"35602251\":[163.38,10],\"35602278\":[31.16,4],\"35602316\":[34.84,2],\"35602561\":[30.0,2],\"35602563\":[39.58,1],\"35602564\":[279.59,5],\"35602652\":[13.17,1],\"35602655\":[7.79,1],\"35602662\":[24.92,1],\"35602673\":[166.44,3],\"35602689\":[143.84,16],\"35602707\":[129.75,3],\"35602708\":[40.59,6],\"35602717\":[99.93,12],\"35602718\":[63.22,11],\"35602732\":[18.56,3],\"35602734\":[112.58,13],\"35602735\":[89.12,12],\"35602744\":[36.0,3],\"35602745\":[62.09,5],\"35602746\":[165.01,29],\"35602748\":[8.25,1],\"35602749\":[22.43,1],\"35602779\":[11.25,1],\"35602783\":[4.05,1],\"35602788\":[9.83,1],\"35602790\":[10.05,1],\"35602798\":[5.25,1],\"35602853\":[11.08,2],\"35602854\":[4.65,1],\"35602855\":[4.65,1],\"35602866\":[8.33,1],\"35602868\":[300.5,4],\"35602879\":[195.88,15],\"35602923\":[23.33,3],\"35602924\":[7.42,1],\"35602925\":[5.85,1],\"35602944\":[-17.7,0],\"36200098\":[1051.38,2],\"36200139\":[699.86,1],\"36200145\":[2349.44,4],\"36200146\":[3444.87,7],\"36200149\":[1249.43,4],\"36200150\":[1534.44,6],\"36200157\":[820.69,1],\"36200164\":[345.69,1],\"36200165\":[972.21,3],\"36200166\":[1012.95,4],\"36200167\":[3364.99,22],\"36800001\":[9261.37,76],\"36800003\":[8476.68,76],\"36800004\":[1633.18,26],\"36800006\":[739.35,5],\"36800018\":[6708.21,42],\"36800019\":[3557.84,26],\"36900720\":[120.27,1],\"36900721\":[158.1,1],\"36900901\":[245.27,1],\"36900961\":[149.3,3],\"36900962\":[202.9,4],\"36901039\":[134.69,2],\"36901680\":[119.69,2],\"36901715\":[530.26,10],\"36901717\":[629.06,6],\"36901719\":[1000.25,11],\"36901730\":[124.79,3],\"36901732\":[125.51,2],\"36901737\":[284.28,3],\"36901741\":[212.29,3],\"36901745\":[154.19,2],\"36901746\":[488.52,3],\"36901748\":[415.79,2],\"36901931\":[667.2,2],\"36901973\":[169.35,1],\"37001565\":[2694.73,27],\"37001568\":[672.09,2],\"37001607\":[2149.0,5],\"37001613\":[8175.28,70],\"37001682\":[-24.99,0],\"37001688\":[3056.05,14],\"37001691\":[1950.42,14],\"37001695\":[1047.68,7],\"37001696\":[175.28,1],\"37001697\":[2411.69,15],\"37001714\":[24787.04,52],\"37001717\":[1585.66,5],\"37001718\":[2512.01,7],\"37001719\":[523.62,3],\"37600001\":[185.44,4],\"37700002\":[66.41,1],\"37800001\":[57.16,1],\"37900591\":[557.3,1],\"38000254\":[349.63,6],\"38000977\":[86.82,1],\"38000979\":[125.32,2],\"38000980\":[63.49,1],\"38000983\":[136.15,2],\"38000992\":[165.48,3],\"38000994\":[660.29,7],\"38001000\":[62.66,1],\"38001029\":[314.48,6],\"38001036\":[727.14,19],\"38001042\":[63.58,1],\"38001043\":[1167.65,20],\"38001069\":[1467.68,17],\"38383016\":[205.59,2],\"38400002\":[206.68,3],\"38600006\":[264.77,3],\"38600008\":[66.23,1],\"38700004\":[77.25,2],\"38800011\":[73.84,1],\"38900778\":[637.16,2],\"38900779\":[637.16,2],\"38900780\":[1586.73,4],\"38900781\":[831.24,2],\"38900784\":[3760.25,32],\"38900792\":[728.66,2],\"38900825\":[1959.91,12],\"38900826\":[164.83,1],\"38900833\":[337.16,2],\"38900838\":[1537.03,7],\"39001481\":[391.68,6],\"39001563\":[339.83,4],\"39002211\":[1255.56,11],\"39002217\":[1682.21,24],\"39002259\":[646.71,8],\"39002262\":[320.47,2],\"39002263\":[507.3,7],\"39002267\":[657.57,9],\"39002268\":[10494.79,158],\"39002271\":[875.04,8],\"39002287\":[130.42,2],\"39002299\":[98.54,1],\"39002306\":[1419.6,17],\"39002310\":[1359.66,19],\"39002311\":[1069.41,15],\"39002315\":[181.05,1],\"39002321\":[1018.08,14],\"39002322\":[759.4,12],\"39002324\":[64.38,1],\"39002325\":[288.83,5],\"39002326\":[269.0,5],\"39002327\":[220.09,4],\"39002328\":[113.76,2],\"39002329\":[17703.76,376],\"39002330\":[2668.25,52],\"39002347\":[316.07,4],\"39002348\":[496.27,7],\"39002355\":[484.67,4],\"39300209\":[114.4,2],\"39300771\":[214.19,3],\"39400987\":[696.98,11],\"39401009\":[217.46,1],\"39401010\":[316.89,3],\"39401011\":[373.12,1],\"39401015\":[3633.38,20],\"39401017\":[806.8,5],\"39401032\":[583.73,6],\"39401034\":[15.04,1],\"39401036\":[1364.6,17],\"39401054\":[7845.69,127],\"39401056\":[3163.36,36],\"39401083\":[1001.22,14],\"39401085\":[3857.32,82],\"39401091\":[10541.88,90],\"39401092\":[328.26,2],\"39401095\":[824.13,6],\"39401096\":[403.15,5],\"39401097\":[3199.07,34],\"39401098\":[3500.89,37],\"39401109\":[1672.53,12],\"39401112\":[1064.81,7],\"39401114\":[434.44,3],\"39401115\":[673.68,5],\"9178278\":[65.56,9],\"AES00001\":[328.32,0],\"AES00002\":[30.0,0],\"AES00003\":[543.31,0],\"AES00004\":[393.31,0],\"AES00005\":[2121.54,0],\"AES00007\":[259.98,0],\"AES00011\":[129.99,0],\"AES00013\":[371.64,0],\"AES00015\":[67.47,0],\"AES00016\":[476.64,0],\"AES00018\":[5210.14,0],\"AES00019\":[33.33,0],\"AES00020\":[1526.58,0],\"AES00022\":[3676.47,0],\"AES00023\":[1133.19,0],\"AES00046\":[861.35,0],\"AES00047\":[508.2,0],\"AES00048\":[626.38,0],\"AES00049\":[189.94,0],\"AES00050\":[428.12,0],\"AES00051\":[457.32,0],\"AES00053\":[49.98,0],\"AES00054\":[1308.04,0],\"AES00055\":[1632.13,0],\"AES00056\":[15.0,0],\"AES00058\":[1551.24,0],\"AES00059\":[694.81,0],\"AES00060\":[1392.03,0],\"AES00061\":[1996.73,0],\"AES00062\":[2992.24,0],\"AES00063\":[5632.14,0],\"AES00064\":[198.66,0],\"AES00065\":[123.48,0],\"AES00066\":[1283.94,0],\"AES00067\":[39.58,0],\"AES00068\":[374.71,0],\"AES00069\":[60.41,0],\"AES00070\":[244.93,0],\"AES00071\":[36.0,0],\"AES00072\":[8.33,0],\"AES00074\":[4604.93,0],\"AES00075\":[2511.91,0],\"AES00076\":[16.66,0],\"AES00078\":[4456.62,0],\"AES00079\":[3507.17,0],\"AES00082\":[573.14,0],\"AES00083\":[889.7,0],\"AES00090\":[425.99,0],\"AES00091\":[111.64,0],\"AESFR001\":[39248.96,0]};\n",[],[214,492],"5c86fd49-0e18-43a2-8cad-494aff722037","analyse_retours.ts","01a033f6-8ab0-7225-822f-0363ea73e507",["D",1787578190526],"// Analyse ponctuelle — avoirs (remboursements) depuis le 1er janvier 2026, par catégorie.\n//\n// Deux précautions de lecture :\n// 1. Les SKU de service (FR_AES*, AESFR*) sont des prestations et l'éco-participation,\n// pas des appareils : ils sont exclus du décompte d'unités.\n// 2. Janvier–février relève de la migration Shopify (commandes de décembre 2025\n// annulées en masse). On sépare donc la période antérieure à la bascule du\n// 1er mars de la période exploitable qui suit.\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst YTD = \"2026-01-01\";\nconst CUTOVER = \"2026-03-01\";\n\nconst orders = (await blob.getJSON(\"orders_lines\")) as any;\nconst catalog = (await blob.getJSON(\"catalog\")) as any;\nconst lines: any[] = orders?.lines ?? [];\n\nconst grpOf: Record = {};\nconst nameOf: Record = {};\nfor (const p of catalog?.products ?? []) {\n grpOf[String(p.code)] = p.grp ?? \"Other\";\n nameOf[String(p.code)] = p.name ?? \"\";\n}\nconst isService = (sku: string) => /^FR_AES|^AESFR/i.test(sku);\n\ninterface Agg { u: number; eur: number; ord: Set }\nconst mk = (): Agg => ({ u: 0, eur: 0, ord: new Set() });\nconst ht = (l: any) => (l.n != null ? l.n : l.p * l.q) / 1.2;\n\nconst shipped: Record = {}; // expédié après la bascule\nconst refPost: Record = {}; // avoirs après la bascule\nconst refPre: Record = {}; // avoirs avant la bascule (migration)\nconst bySku: Record = {};\n\nconst add = (m: Record, g: string, l: any) => {\n const a = (m[g] ??= mk());\n a.u += l.q; a.eur += ht(l); a.ord.add(`${l.b}|${l.o}`);\n};\n\nfor (const l of lines) {\n const sku = String(l.sku ?? \"\");\n if (!sku || isService(sku)) continue;\n const g = grpOf[sku] ?? \"Non rattaché au catalogue\";\n\n if (l.k === \"ful\" && (l.f || l.d) >= CUTOVER) add(shipped, g, l);\n if (l.k === \"ref\" && (l.rf ?? \"\") >= YTD) {\n const post = (l.rf ?? \"\") >= CUTOVER;\n add(post ? refPost : refPre, g, l);\n if (post) {\n const s = (bySku[sku] ??= { u: 0, eur: 0 });\n s.u += l.q; s.eur += ht(l);\n }\n }\n}\n\nconst r = (n: number) => Math.round(n);\nconst pct = (a: number, b: number) => b > 0 ? (a / b * 100).toFixed(2) + \" %\" : \"—\";\nconst sum = (m: Record, f: (a: Agg) => number) =>\n Object.values(m).reduce((s, a) => s + f(a), 0);\n\nconst groups = [...new Set([...Object.keys(shipped), ...Object.keys(refPost), ...Object.keys(refPre)])];\nconst table = groups.map((g) => ({\n categorie: g,\n expedieUnites: r(shipped[g]?.u ?? 0),\n avoirsUnites: r(refPost[g]?.u ?? 0),\n tauxAvoir: pct(refPost[g]?.u ?? 0, shipped[g]?.u ?? 0),\n avoirsEUR: r(refPost[g]?.eur ?? 0),\n avoirsMigrationUnites: r(refPre[g]?.u ?? 0),\n avoirsMigrationEUR: r(refPre[g]?.eur ?? 0),\n})).sort((a, b) => b.avoirsUnites - a.avoirsUnites);\n\nconsole.log(JSON.stringify({\n majDonnees: orders?.updated_at ?? null,\n perimetre: \"appareils uniquement (SKU de service exclus)\",\n apresBascule: {\n periode: `${CUTOVER} → aujourd'hui`,\n expedieUnites: r(sum(shipped, (a) => a.u)),\n expedieEUR: r(sum(shipped, (a) => a.eur)),\n avoirsUnites: r(sum(refPost, (a) => a.u)),\n avoirsEUR: r(sum(refPost, (a) => a.eur)),\n commandesConcernees: new Set(Object.values(refPost).flatMap((a) => [...a.ord])).size,\n tauxAvoirUnites: pct(sum(refPost, (a) => a.u), sum(shipped, (a) => a.u)),\n },\n avantBascule_migration: {\n periode: `${YTD} → ${CUTOVER}`,\n avoirsUnites: r(sum(refPre, (a) => a.u)),\n avoirsEUR: r(sum(refPre, (a) => a.eur)),\n },\n parCategorie: table,\n skuAvoirsApresBascule: Object.entries(bySku)\n .sort((a, b) => b[1].eur - a[1].eur)\n .map(([sku, v]) => ({ sku, nom: nameOf[sku] ?? \"inconnu\", groupe: grpOf[sku] ?? \"?\", unites: v.u, eur: r(v.eur) })),\n}, null, 2));\n",[],[214,499],["D",1787578190526],[252,253,254,255,256,257,258,259,260,261,262,263,264,265,266,267,268,269,270,271,272,273,274,275,276,277,278],[214],"6b01ff10-9fd7-4828-a0ee-185d6b68191d","pricing.ts","01a033f6-8ab0-7225-822e-f820979ba725",["D",1787578190526],"// Pricing model — a faithful port of the Python pipeline (push_dashboard.py) and of the\n// \"Parameters\" sheet of the Excel workbook.\n//\n// Les valeurs ci-dessous ne sont plus que le repli : la source de vérité est le blob\n// `params` (voir params.ts), édité depuis l'onglet Parameters et lu par le pipeline\n// Python. `applyParams()` recharge PARAMS à l'ouverture du dashboard ; si l'appel API\n// échoue, on garde ces constantes plutôt que de calculer avec des valeurs vides.\n//\n// PARAMS reste un objet muté sur place, et non remplacé : les fonctions ci-dessous le\n// capturent par référence, et une réaffectation les laisserait sur l'ancien objet.\n\nexport const PARAMS = {\n OPEX: 0.086, // B5 — operating expenses, % of price excl. VAT\n MKTG: 0.10, // B15 — marketing, % of price excl. VAT\n VAT: 0.20, // B16\n WEEE_GROSS: 10.584, // B11 — gross WEEE take-back cost\n TAKEBACK: 0.15, // B12 — share of units actually taken back\n RET_MDA: 0.037, // B13 — return rate, large appliances (CCV)\n RET_SDA: 0.10, // B14 — return rate, small appliances (DPD)\n SHIP_SDA: 6.49, // B17 — delivery charged to the customer, incl. VAT (DPD)\n SHIP_MDA: 27.31, // B18 — delivery charged to the customer, incl. VAT (CCV)\n};\n\n/** Applique les paramètres enregistrés. Seules les clés connues et numériques sont\n * reprises : une réponse partielle ou abîmée ne doit pas introduire de NaN dans le\n * breakeven, où il se propagerait silencieusement à toutes les marges. */\nexport function applyParams(values: Partial>): void {\n for (const k of Object.keys(PARAMS) as (keyof typeof PARAMS)[]) {\n const v = values?.[k];\n if (typeof v === \"number\" && isFinite(v)) PARAMS[k] = v;\n }\n}\n\n/** DPD rate per kg bracket (index = ceil(weight)), incl. VAT. Above 30 kg → CCV. */\nexport const DPD_RATES: Record = {\n 1: 5.49, 2: 5.49, 3: 5.79, 4: 6.09, 5: 6.39, 6: 6.69, 7: 6.99, 8: 7.29, 9: 7.59, 10: 7.89,\n 11: 8.29, 12: 8.69, 13: 9.09, 14: 9.49, 15: 9.89, 16: 10.29, 17: 10.69, 18: 11.09,\n 19: 11.49, 20: 11.89, 21: 12.29, 22: 12.69, 23: 13.09, 24: 13.49, 25: 13.89, 26: 14.29,\n 27: 14.69, 28: 15.09, 29: 15.49, 30: 15.89,\n};\n\n/** CCV rate by weight bracket: [min kg, rate]. The last bracket ≤ weight wins. */\nexport const CCV_BRACKETS: [number, number][] = [\n [0, 36.66], [30, 42.63], [40, 48.60], [50, 54.08], [60, 59.84], [70, 67.76],\n [80, 75.69], [90, 83.62], [100, 91.55], [110, 99.47], [120, 107.40], [130, 115.32],\n [140, 123.25], [150, 131.17], [160, 139.10], [170, 147.03],\n];\n\n/** Categories shipped by DPD when light enough. Everything else goes CCV. */\nexport const DPD_CATEGORIES = new Set([\n \"Blender\", \"Kettle\", \"Toaster\", \"Food Chopper\", \"Hot Beverage Machine\",\n \"Hand Blender\", \"Coffee Machine\", \"Air Fryer\",\n \"Stick Vacuum\", \"Cylinder Vacuum\", \"Handheld Vacuum\", \"Wet & Dry\",\n \"Robot Vacuum\", \"Steam Cleaner\", \"Air Treatment\",\n \"Consumable\", \"Microwave\",\n]);\n\n/** Every category known to the pipeline, grouped for the simulator's dropdown. */\nexport const CATEGORIES: { grp: string; cats: string[] }[] = [\n { grp: \"Cooling\", cats: [\"Refrigerator / Fridge-Freezer\", \"Built-in Refrigerator\", \"Freezer\", \"Built-in Freezer\", \"Wine Cellar\"] },\n { grp: \"Washing\", cats: [\"Washing Machine\", \"Built-in Washing Machine\", \"Tumble Dryer\", \"Built-in Tumble Dryer\", \"Dishwasher\", \"Built-in Dishwasher\", \"Drying Cabinet\"] },\n { grp: \"Cooking\", cats: [\"Cooker\", \"Built-in Oven\", \"Hob\", \"Hood\", \"Microwave\", \"Built-in Microwave\", \"Steam Drawer / Specialties\"] },\n { grp: \"Floorcare\", cats: [\"Stick Vacuum\", \"Cylinder Vacuum\", \"Handheld Vacuum\", \"Robot Vacuum\", \"Wet & Dry\", \"Steam Cleaner\", \"Air Treatment\"] },\n { grp: \"KSDA\", cats: [\"Coffee Machine\", \"Hot Beverage Machine\", \"Air Fryer\", \"Blender\", \"Hand Blender\", \"Kettle\", \"Toaster\", \"Food Chopper\"] },\n { grp: \"TV\", cats: [\"TV\"] },\n { grp: \"Consumable\", cats: [\"Consumable\"] },\n];\n\nexport type Carrier = \"DPD\" | \"CCV\";\n\n/** DPD only for small-appliance categories up to 30 kg — same rule as the pipeline. */\nexport function carrierFor(weight: number | null, category: string): Carrier {\n if (DPD_CATEGORIES.has(category) && (weight ?? 0) <= 30) return \"DPD\";\n return \"CCV\";\n}\n\n/** Outbound logistics cost per unit. Returns null when out of the DPD grid (>30 kg). */\nexport function logisticsCost(weight: number | null, carrier: Carrier): number | null {\n if (weight == null) return null;\n if (carrier === \"DPD\") return DPD_RATES[Math.ceil(weight)] ?? null;\n let rate = CCV_BRACKETS[0][1];\n for (const [kgMin, r] of CCV_BRACKETS) if (weight >= kgMin) rate = r;\n return rate + PARAMS.WEEE_GROSS * PARAMS.TAKEBACK;\n}\n\n/** Breakeven price INCL. VAT, rounded up to the next €10 — identical to the pipeline. */\nexport function breakeven(cost: number | null, logi: number | null, carrier: Carrier): number | null {\n if (!cost || logi == null) return null;\n const { RET_MDA, RET_SDA, SHIP_MDA, SHIP_SDA, OPEX, MKTG, VAT } = PARAMS;\n const ret = carrier === \"CCV\" ? RET_MDA : RET_SDA;\n const ship = carrier === \"CCV\" ? SHIP_MDA : SHIP_SDA;\n const be = (cost + logi * (1 + ret) - ship / (1 + VAT)) / (1 - OPEX - MKTG) * (1 + VAT);\n return Math.ceil(be / 10) * 10;\n}\n\n/** Gross margin = (excl. VAT price − standard cost − eco-participation) / excl. VAT price.\n * This is the \"Margin\" shown everywhere else in the dashboard. It deliberately ignores\n * logistics, returns, OPEX and marketing — so it is NOT zero at the breakeven price. */\nexport function grossMargin(priceIncl: number | null, cost: number | null, deee: number): number | null {\n if (!priceIncl || cost == null) return null;\n const ht = priceIncl / (1 + PARAMS.VAT);\n return ht > 0 ? (ht - cost - deee) / ht : null;\n}\n\n/** Price incl. VAT that yields a given gross margin. */\nexport function priceForGrossMargin(margin: number, cost: number | null, deee: number): number | null {\n if (cost == null || margin >= 1) return null;\n return ((cost + deee) / (1 - margin)) * (1 + PARAMS.VAT);\n}\n\n/** Full contribution margin: what actually remains once every variable cost is deducted —\n * logistics (grossed up for returns), the delivery the customer pays back, OPEX and\n * marketing. Zero (by construction) at the unrounded breakeven price. */\nexport function contributionMargin(\n priceIncl: number | null, cost: number | null, deee: number,\n logi: number | null, carrier: Carrier,\n): { amount: number; rate: number } | null {\n if (!priceIncl || cost == null || logi == null) return null;\n const { RET_MDA, RET_SDA, SHIP_MDA, SHIP_SDA, OPEX, MKTG, VAT } = PARAMS;\n const ret = carrier === \"CCV\" ? RET_MDA : RET_SDA;\n const ship = carrier === \"CCV\" ? SHIP_MDA : SHIP_SDA;\n const ht = priceIncl / (1 + VAT);\n const amount = ht - cost - deee - logi * (1 + ret) + ship / (1 + VAT) - ht * (OPEX + MKTG);\n return { amount, rate: ht > 0 ? amount / ht : 0 };\n}\n\n/** True breakeven, incl. VAT, once the eco-participation is also covered.\n * The catalogue formula (above) omits the DEEE, so it sits slightly BELOW the price at\n * which contribution actually reaches zero. Unrounded, to show the real threshold. */\nexport function breakevenWithDeee(\n cost: number | null, deee: number, logi: number | null, carrier: Carrier,\n): number | null {\n if (!cost || logi == null) return null;\n const { RET_MDA, RET_SDA, SHIP_MDA, SHIP_SDA, OPEX, MKTG, VAT } = PARAMS;\n const ret = carrier === \"CCV\" ? RET_MDA : RET_SDA;\n const ship = carrier === \"CCV\" ? SHIP_MDA : SHIP_SDA;\n return (cost + deee + logi * (1 + ret) - ship / (1 + VAT)) / (1 - OPEX - MKTG) * (1 + VAT);\n}\n\n// ─────────────────────────────────────────────────────────────────────────────\n// Official eco-participation scale — ecosystem, \"Barème EEE ménagers\",\n// version 27/03/2026, amounts applicable from 1 January 2026. Amounts are EXCL. VAT.\n// Where the scale offers eco-modulated variants (HFC gas, removable battery, RFB),\n// we take the standard case — the penalty variants are noted in `note`.\n// CAVEAT: the scale brackets are on NET weight, while the catalogue often carries\n// packaged weight. Near a bracket boundary, check the net weight.\n// ─────────────────────────────────────────────────────────────────────────────\n\nexport interface DeeeBracket {\n /** Upper bound of the bracket, on the driver's unit. Last bracket has none. */\n upTo?: number;\n value: number;\n label: string;\n code: string;\n}\n\nexport interface DeeeTariff {\n /** What the amount depends on: net weight (kg), screen size (inches), or nothing. */\n driver: \"weight\" | \"screen\" | \"none\";\n /** ecosystem equipment family this catalogue category maps to. */\n family: string;\n brackets: DeeeBracket[];\n note?: string;\n}\n\n/** Catalogue category → official tariff. */\nexport const DEEE_SCALE: Record = {\n // ── Gros électroménager produisant du froid (Cat. 01) — one weight scale ──\n ...Object.fromEntries(\n [\"Refrigerator / Fridge-Freezer\", \"Built-in Refrigerator\", \"Freezer\",\n \"Built-in Freezer\", \"Wine Cellar\"].map((c) => [c, {\n driver: \"weight\" as const,\n family: \"Réfrigérateur / Congélateur / Cave à vin\",\n brackets: [\n { upTo: 40, value: 12.23, label: \"≤ 40 kg\", code: \"11023\" },\n { upTo: 80, value: 20.20, label: \"40–80 kg\", code: \"11013\" },\n { value: 25.17, label: \"> 80 kg\", code: \"11083\" },\n ],\n }]),\n ),\n\n // ── Lave-linge / lave-vaisselle / sèche-linge (Cat. 04) ──\n ...Object.fromEntries(\n [\"Washing Machine\", \"Built-in Washing Machine\", \"Dishwasher\",\n \"Built-in Dishwasher\", \"Tumble Dryer\", \"Built-in Tumble Dryer\"].map((c) => [c, {\n driver: \"none\" as const,\n family: \"Lave-linge / Lave-vaisselle / Sèche-linge\",\n brackets: [{ value: 12.85, label: \"absence de gaz HFC\", code: \"14013\" }],\n note: \"Avec gaz HFC : 14,48 € (lave-linge, séchant) ou 13,93 € (lave-vaisselle combiné).\",\n }]),\n ),\n\n // ── Autre gros électroménager hors froid ──\n \"Cooker\": {\n driver: \"none\", family: \"Cuisinière\",\n brackets: [{ value: 12.20, label: \"forfait\", code: \"14020\" }],\n },\n \"Drying Cabinet\": {\n driver: \"none\", family: \"Appareil de séchage\",\n brackets: [{ value: 12.20, label: \"forfait\", code: \"14020\" }],\n },\n \"Built-in Oven\": {\n driver: \"weight\", family: \"Four encastrable\",\n brackets: [\n { upTo: 30, value: 7.13, label: \"≤ 30 kg\", code: \"14160\" },\n { value: 12.20, label: \"> 30 kg\", code: \"14170\" },\n ],\n },\n \"Microwave\": {\n driver: \"none\", family: \"Four micro-ondes\",\n brackets: [{ value: 3.17, label: \"forfait\", code: \"15120\" }],\n },\n \"Built-in Microwave\": {\n driver: \"none\", family: \"Four micro-ondes\",\n brackets: [{ value: 3.17, label: \"forfait\", code: \"15120\" }],\n },\n ...Object.fromEntries(\n [\"Hood\", \"Hob\", \"Steam Drawer / Specialties\"].map((c) => [c, {\n driver: \"none\" as const,\n family: \"Hotte / Table de cuisson / Tiroir chauffe-plat\",\n brackets: [{ value: 3.90, label: \"forfait\", code: \"14030\" }],\n }]),\n ),\n\n // ── Électronique grand public — TV, driven by SCREEN SIZE, not weight ──\n \"TV\": {\n driver: \"screen\", family: \"TV\",\n brackets: [\n { upTo: 32, value: 5.08, label: \"≤ 32\\\"\", code: \"42093\" },\n { upTo: 50, value: 9.26, label: \"32–50\\\"\", code: \"42083\" },\n { value: 12.64, label: \"> 50\\\"\", code: \"42073\" },\n ],\n note: \"Présence de RFB : +1,92 € (7,00 / 11,18 / 14,56 €).\",\n },\n\n // ── Aspirateurs (Cat. 05) ──\n \"Cylinder Vacuum\": {\n driver: \"none\", family: \"Aspirateur traîneau\",\n brackets: [{ value: 1.45, label: \"filaire, absence de RFB\", code: \"25013\" }],\n note: \"Présence de RFB : 2,45 €. Non filaire à batterie non séparable : 2,24 €.\",\n },\n \"Stick Vacuum\": {\n driver: \"none\", family: \"Aspirateur balai\",\n brackets: [{ value: 1.87, label: \"batterie séparable\", code: \"25074\" }],\n note: \"Batterie non séparable : 2,66 €. Filaire avec RFB : 2,87 €.\",\n },\n \"Robot Vacuum\": {\n driver: \"none\", family: \"Aspirateur robot\",\n brackets: [{ value: 2.83, label: \"batterie séparable\", code: \"25080\" }],\n note: \"Batterie non séparable : 3,62 €.\",\n },\n \"Handheld Vacuum\": {\n driver: \"none\", family: \"Aspirateur à main\",\n brackets: [{ value: 0.30, label: \"forfait\", code: \"25150\" }],\n },\n \"Wet & Dry\": {\n driver: \"none\", family: \"Aspirateur balai\",\n brackets: [{ value: 1.87, label: \"filaire, absence de RFB\", code: \"25073\" }],\n note: \"Le catalogue applique encore 1,86 € (barème antérieur). Si l'appareil relève plutôt \"\n + \"de « Nettoyage sols », le montant est de 1,71 €.\",\n },\n \"Consumable\": {\n driver: \"none\", family: \"Autre petit PEM / accessoire\",\n brackets: [{ value: 0.30, label: \"forfait\", code: \"25150\" }],\n note: \"Beaucoup d'accessoires ne sont pas des EEE et ne supportent aucune éco-participation.\",\n },\n \"Steam Cleaner\": {\n driver: \"none\", family: \"Nettoyeur vapeur portatif\",\n brackets: [{ value: 0.35, label: \"portatif\", code: \"25040\" }],\n note: \"Repasseuse / centrale vapeur : 1,71 €.\",\n },\n\n // ── Petit électroménager cuisine ──\n \"Coffee Machine\": {\n driver: \"none\", family: \"Cafetière expresso\",\n brackets: [{ value: 1.14, label: \"expresso\", code: \"25100\" }],\n note: \"Cafetière à broyeur : 1,74 €. Cafetière filtre : 0,34 €.\",\n },\n \"Hot Beverage Machine\": {\n driver: \"none\", family: \"Cafetière filtre / Bouilloire / Théière\",\n brackets: [{ value: 0.34, label: \"forfait\", code: \"25033\" }],\n },\n \"Kettle\": {\n driver: \"none\", family: \"Bouilloire\",\n brackets: [{ value: 0.34, label: \"forfait\", code: \"25033\" }],\n },\n \"Toaster\": {\n driver: \"none\", family: \"Grille-pain\",\n brackets: [{ value: 0.35, label: \"forfait\", code: \"25040\" }],\n },\n \"Air Fryer\": {\n driver: \"none\", family: \"Friteuse\",\n brackets: [{ value: 0.35, label: \"forfait\", code: \"25040\" }],\n },\n ...Object.fromEntries(\n [\"Blender\", \"Hand Blender\", \"Food Chopper\"].map((c) => [c, {\n driver: \"none\" as const,\n family: \"Blender / Centrifugeuse / Mixeur\",\n brackets: [{ value: 0.33, label: \"forfait\", code: \"25140\" }],\n note: \"Robot pâtissier : 1,83 €. Robot chauffant : 2,05 €.\",\n }]),\n ),\n\n // ── Génie climatique ──\n \"Air Treatment\": {\n driver: \"weight\", family: \"Traitement de l'air\",\n brackets: [\n { upTo: 5, value: 0.91, label: \"≤ 5 kg\", code: \"15050\" },\n { value: 1.81, label: \"> 5 kg, sans fluide\", code: \"14040\" },\n ],\n note: \"Avec fluide frigorigène, l'appareil bascule sur le barème du froid (12,23 € ≤ 40 kg).\",\n },\n};\n\n/** The official eco-participation for a category, given weight (kg) or screen size (inches). */\nexport function officialDeee(\n category: string,\n weight: number | null,\n screenInches: number | null = null,\n): { value: number; bracket: DeeeBracket; tariff: DeeeTariff } | null {\n const t = DEEE_SCALE[category];\n if (!t) return null;\n const driverValue = t.driver === \"screen\" ? screenInches : t.driver === \"weight\" ? weight : null;\n if (t.driver !== \"none\" && driverValue == null) {\n // No driver value yet — fall back to the first bracket so the field is never empty\n return { value: t.brackets[0].value, bracket: t.brackets[0], tariff: t };\n }\n for (const b of t.brackets) {\n if (b.upTo == null || (driverValue ?? 0) <= b.upTo) return { value: b.value, bracket: b, tariff: t };\n }\n const last = t.brackets[t.brackets.length - 1];\n return { value: last.value, bracket: last, tariff: t };\n}\n\n/** Flags a driver value close enough to a bracket boundary to be worth checking.\n * The window is asymmetric on weight: the scale is on NET weight while catalogue weights\n * are usually PACKAGED, so a product sitting just ABOVE a boundary is the risky case —\n * measured on the catalogue, 40–47 kg fridges carry the ≤ 40 kg amount. */\nexport function nearBracketBoundary(\n category: string, weight: number | null, screenInches: number | null = null,\n): number | null {\n const t = DEEE_SCALE[category];\n if (!t || t.driver === \"none\") return null;\n const v = t.driver === \"screen\" ? screenInches : weight;\n if (v == null) return null;\n const above = t.driver === \"weight\" ? 0.25 : 0.1; // packaging overhead only applies to weight\n for (const b of t.brackets) {\n if (b.upTo == null) continue;\n if (v > b.upTo && v <= b.upTo * (1 + above)) return b.upTo;\n if (v <= b.upTo && v >= b.upTo * 0.9) return b.upTo;\n }\n return null;\n}\n\n/** One observed eco-participation value within a category, with the weights it covers. */\nexport interface DeeeBand {\n value: number;\n count: number;\n minWt: number | null;\n maxWt: number | null;\n}\n\n/** The eco-participation values actually used in a category, heaviest last.\n * On cooling appliances the amount follows a weight scale (12,23 € / 20,20 € / 25,17 €),\n * on most other categories a single value applies — this surfaces both cases. */\nexport function deeeBands(\n products: { cat?: string | null; deee?: number | null; wt?: number | null }[],\n category: string,\n): DeeeBand[] {\n const byValue = new Map();\n for (const p of products) {\n if (p.cat !== category || !p.deee || p.deee <= 0) continue;\n const k = p.deee.toFixed(2);\n const e = byValue.get(k) ?? { count: 0, wts: [] };\n e.count++;\n if (p.wt != null) e.wts.push(p.wt);\n byValue.set(k, e);\n }\n return [...byValue.entries()]\n .map(([k, e]) => ({\n value: parseFloat(k),\n count: e.count,\n minWt: e.wts.length ? Math.min(...e.wts) : null,\n maxWt: e.wts.length ? Math.max(...e.wts) : null,\n }))\n .sort((a, b) => a.value - b.value);\n}\n\n/** Best guess for a new product's eco-participation: the value used by the catalogue\n * products of the same category whose weight is closest. Falls back to the most common\n * value in the category when no weight is available. Returns null if the category has none. */\nexport function suggestDeee(\n products: { cat?: string | null; deee?: number | null; wt?: number | null }[],\n category: string,\n weight: number | null,\n neighbours = 7,\n): number | null {\n const pool = products.filter((p) => p.cat === category && p.deee && p.deee > 0);\n if (pool.length === 0) return null;\n\n const withWt = pool.filter((p) => p.wt != null);\n if (weight != null && withWt.length > 0) {\n const near = [...withWt]\n .sort((a, b) => Math.abs((a.wt ?? 0) - weight) - Math.abs((b.wt ?? 0) - weight))\n .slice(0, Math.min(neighbours, withWt.length));\n // Modal value among the nearest neighbours — robust to a stray outlier\n const tally = new Map();\n for (const p of near) {\n const k = (p.deee ?? 0).toFixed(2);\n tally.set(k, (tally.get(k) ?? 0) + 1);\n }\n let best = \"\", bestN = -1;\n for (const [k, n] of tally) if (n > bestN) { best = k; bestN = n; }\n return parseFloat(best);\n }\n\n const tally = new Map();\n for (const p of pool) {\n const k = (p.deee ?? 0).toFixed(2);\n tally.set(k, (tally.get(k) ?? 0) + 1);\n }\n let best = \"\", bestN = -1;\n for (const [k, n] of tally) if (n > bestN) { best = k; bestN = n; }\n return parseFloat(best);\n}\n\n/** Psychological rounding used by the pipeline for the recommended price: x9.99 */\nexport function roundPsy(v: number): number {\n return Math.ceil(v / 10) * 10 - 0.01;\n}\n",[],[220,509],"d67553eb-27e2-46aa-92c9-b9293e776d7e","ordersStore.ts","01a033f6-8ab0-7225-822e-f49ca7f3c806",["D",1787578190526],"// Chargement MUTUALISÉ des lignes de commandes.\n//\n// `/api/business` renvoie l'intégralité du blob des commandes — près de 7 Mo. Quatre\n// onglets en ont besoin (Overview, Scorecards, Business, Logistic) et chacun le\n// téléchargeait pour son propre compte : quatre requêtes, ~28 Mo, plusieurs secondes\n// d'attente, et surtout des écrans qui affichaient des chiffres différents selon\n// l'ordre d'arrivée des réponses.\n//\n// Ici la requête est lancée UNE fois et sa promesse est conservée. Tout composant qui\n// la demande ensuite reçoit la même donnée, immédiatement si elle est déjà là. Les\n// quatre onglets sont donc toujours d'accord entre eux, par construction.\n\nimport { useEffect, useState } from \"https://esm.sh/react@18.2.0\";\n\nexport interface OrdersPayload {\n updated_at?: string;\n retention_days?: number;\n feeds?: Record;\n lines: any[];\n orphan_refunds?: any[];\n}\n\n/** null = le flux a répondu mais n'est pas exploitable (non configuré, erreur HTTP). */\ntype Loaded = OrdersPayload | null;\n\nlet inflight: Promise | null = null;\nlet settled: Loaded | undefined = undefined;\n\n/** Lance la requête si besoin, sinon rend la promesse en cours ou le résultat déjà là. */\nexport function loadOrders(): Promise {\n if (settled !== undefined) return Promise.resolve(settled);\n if (!inflight) {\n inflight = fetch(\"/api/business\")\n .then((r) => (r.ok ? r.json() : null))\n .then((d: Loaded) => {\n // Un payload sans tableau `lines` est inutilisable : on le ramène à null pour\n // que les écrans affichent leur message de configuration plutôt qu'un vide.\n settled = d && Array.isArray((d as OrdersPayload).lines) ? d : null;\n return settled;\n })\n .catch(() => {\n settled = null;\n return settled;\n });\n }\n return inflight;\n}\n\n/** Force un rechargement au prochain appel — utile après une synchro manuelle. */\nexport function invalidateOrders() {\n inflight = null;\n settled = undefined;\n}\n\n/**\n * `undefined` tant que la donnée n'est pas arrivée, `null` si elle est absente,\n * sinon le payload. Cette distinction est importante : elle permet aux écrans de\n * montrer un état de chargement au lieu d'un chiffre de repli qui ressemble à un\n * résultat définitif — c'est exactement ce qui faisait diverger Overview et Business.\n */\nexport function useOrders(): OrdersPayload | null | undefined {\n const [state, setState] = useState(settled);\n useEffect(() => {\n let alive = true;\n loadOrders().then((d) => { if (alive) setState(d); });\n return () => { alive = false; };\n }, []);\n return state;\n}\n",[],[220,516],"9e40eeed-2090-499c-9e0d-2ddc1f0f330a","index.tsx","01a033f6-8ab0-7225-822e-f02b8b188671",["D",1787578190526],"/** @jsxImportSource https://esm.sh/react@18.2.0 */\nimport { createRoot } from \"https://esm.sh/react-dom@18.2.0/client?deps=react@18.2.0\";\nimport App from \"/frontend/App.tsx\";\n\nconst root = createRoot(document.getElementById(\"root\")!);\nroot.render();\n",[],[220,523],"fbd09a29-1eb1-4cd2-ac74-71e501731cc5","index.html","01a033f6-8ab0-7225-822e-ecfa52ed95fe",["D",1787578190526],"\n\n\n \n \n D2C France — E-commerce Dashboard\n \n \n \n \n \n\n\n

\n \n\n\n",[],[220,530],"2dc1f11b-7e5e-47bd-95a4-8b33eeb8cf7e","helpers.ts","01a033f6-8ab0-7225-822e-e9ac113c3623",["D",1787578190526],"// Shared helpers: formatting, filtering, aggregation, CSV export.\n\nexport interface Product {\n brand: string; code: number; ean?: string | null; name: string;\n cat?: string | null; grp: string; wt?: number | null; carrier?: string | null;\n lc: string; ss: string; sstock?: number | null; s48: number; s47p: number; s47d: number;\n cost?: number | null; pvc?: number | null; deee?: number | null; logi?: number | null;\n be?: number | null; mip?: number | null; rsp?: number | null; mrsp?: number | null;\n price?: number | null; mcur?: number | null; comp?: number | null; compname?: string | null;\n gap?: number | null; pos?: string | null; fc: number; cov?: number | null;\n sales: number; rev?: number | null; revPre?: number | null; salesPre?: number | null;\n spw?: number | null; st?: number | null; fneed?: string | null;\n cstat: string; cact: string; img?: string | null;\n}\n\n/** Category groups. `icon` maps to a component in Icons.tsx (see GROUP_ICONS). */\nexport const GROUPS = [\n { key: \"Cooling\", label: \"Cooling\", icon: \"fridge\" },\n { key: \"Washing\", label: \"Washing\", icon: \"washer\" },\n { key: \"Cooking\", label: \"Cooking\", icon: \"oven\" },\n { key: \"Floorcare\", label: \"Floorcare\", icon: \"vacuum\" },\n { key: \"KSDA\", label: \"KSDA\", icon: \"coffee\" },\n { key: \"TV\", label: \"TV\", icon: \"tv\" },\n { key: \"Consumable\", label: \"Consumable\", icon: \"detergent\" },\n { key: \"Other\", label: \"Other\", icon: \"other\" },\n];\n\nexport const BRAND_COLORS: Record = {\n CANDY: \"bg-sky-600\", HOOVER: \"bg-red-600\", HAIER: \"bg-blue-900\",\n};\n\nexport const fmtInt = (n: number | null | undefined) =>\n n == null ? \"—\" : Math.round(n).toLocaleString(\"en-US\");\n\nexport const fmtEur = (n: number | null | undefined, dec = 2) =>\n n == null ? \"—\" : n.toLocaleString(\"en-US\", { minimumFractionDigits: dec, maximumFractionDigits: dec }) + \" €\";\n\nexport const fmtPct = (n: number | null | undefined, dec = 1) =>\n n == null || !isFinite(n) ? \"—\" : (n * 100).toFixed(dec) + \"%\";\n\n// ── Commercial / technical alert derivation ───────────────────────────────────\nexport interface Alert { type: \"commercial\" | \"technical\"; issue: string; severity: number; detail?: string }\n\nexport function productAlerts(p: Product): Alert[] {\n const out: Alert[] = [];\n const price = p.price ?? 0;\n // Technical (catalogue actions from pipeline)\n if (p.cact && !p.cact.includes(\"No action\") && !p.cact.includes(\"Listing ready\")) {\n out.push({ type: \"technical\", issue: p.cact, severity: p.cact.includes(\"Create\") || p.cact.includes(\"Activate\") ? 1 : 2 });\n }\n // Commercial\n if (price > 0 && p.be != null && price < p.be) {\n out.push({ type: \"commercial\", issue: \"🔴 Price below breakeven\", severity: 1, detail: `${fmtEur(price)} < BE ${fmtEur(p.be)}` });\n } else if (price > 0 && p.mip != null && price < p.mip) {\n out.push({ type: \"commercial\", issue: \"🟠 Price below min internal price\", severity: 2, detail: `${fmtEur(price)} < MIP ${fmtEur(p.mip)}` });\n }\n if (p.pos === \"above\" && price > 0) {\n out.push({ type: \"commercial\", issue: \"🔴 Above market price\", severity: 2, detail: `Gap ${fmtPct(p.gap)} vs ${p.compname ?? \"competitor\"}` });\n }\n const activeLc = !/(End of Life|Phasing Out|Discontinued|Out of SCP)/i.test(p.lc);\n if (activeLc && p.fc > 0 && (p.st ?? 0) < 0.7 && p.s48 > 0 && p.sales >= 0) {\n out.push({ type: \"commercial\", issue: \"🟡 Low sell-through (<70%)\", severity: 3, detail: `${fmtPct(p.st)} — stock ${fmtInt(p.s48)} u` });\n }\n if (p.fneed && (p.fneed.includes(\"to plan\") || p.fneed.includes(\"to forecast\"))) {\n out.push({ type: \"commercial\", issue: \"📈 Forecast insufficient\", severity: 2, detail: p.fneed });\n }\n return out;\n}\n\n// ── Revenue, one definition for the whole dashboard ───────────────────────────\n// Every revenue figure is built the same way:\n// BI invoiced NSV before the Shopify cut-over (migration: Shopify is unusable)\n// + Shopify shipped from the cut-over onwards (fresher, synced nightly)\n// Shopify revenue is excl. VAT, excl. eco-participation (as the BI's NSV), net of\n// discounts, with refunds deducted on their own refund date.\n// Tabs differ only in SCOPE — whether services and delivery are counted — never in\n// the underlying definition.\n\nexport const DEFAULT_CUTOVER = \"2026-03-01\";\n\n/** Shipped revenue per SKU from the orders feed, over [from, to]. */\nexport function shippedRevenueBySku(\n lines: any[], products: Product[], from: string, to: string,\n): Record {\n const deee: Record = {};\n for (const p of products) deee[String(p.code)] = p.deee ?? 0;\n const out: Record = {};\n for (const l of lines) {\n if (l.k !== \"ful\" && l.k !== \"ref\") continue;\n const d = l.k === \"ref\" ? l.rf : (l.f || l.d);\n if (!d || d < from || d > to) continue;\n const sku = String(l.sku ?? \"\");\n const amt = (l.n != null ? l.n : l.p * l.q) / 1.2 - (deee[sku] ?? 0) * l.q;\n out[sku] = (out[sku] ?? 0) + (l.k === \"ref\" ? -amt : amt);\n }\n return out;\n}\n\n/** Delivery charged to the customer over [from, to] — counted once per order. */\nexport function shippedDelivery(lines: any[], from: string, to: string): number {\n const seen = new Set();\n let s = 0;\n for (const l of lines) {\n if (l.k !== \"ful\") continue;\n const d = l.f || l.d;\n if (!d || d < from || d > to) continue;\n if (l.ship && !seen.has(l.o)) { seen.add(l.o); s += l.ship / 1.2; }\n }\n return s;\n}\n\n/** BI revenue for the part of [from, …] that precedes the cut-over.\n * The BI is only available as a per-SKU total, not day by day, so it is added whole —\n * and only when the selected period actually starts before the cut-over. Returns 0 as\n * soon as the period starts later, or while the pipeline has not published `revPre`. */\nexport function preCutoverRevenue(\n products: Product[], from: string, cutover = DEFAULT_CUTOVER,\n): number {\n if (from >= cutover) return 0;\n return products.reduce((s, p) => s + (p.revPre ?? 0), 0);\n}\n\n/** Date à laquelle la fenêtre Shopify doit réellement commencer.\n *\n * Dès que la part BI couvre les mois antérieurs à la bascule, Shopify ne doit PAS\n * les couvrir aussi : Shopify contient un résidu sur ces mois (73 k€ en janvier–février,\n * la migration n'ayant laissé qu'une fraction des ventes), qui serait sinon additionné\n * au BI. Toujours passer par cette fonction plutôt que d'utiliser `from` directement. */\nexport function shopifyWindowStart(\n from: string, cutover = DEFAULT_CUTOVER, biIncluded = true,\n): string {\n return biIncluded && from < cutover ? cutover : from;\n}\n\n/** Quantités facturées avant la bascule — pendant de preCutoverRevenue.\n * Les deux vont toujours ensemble : afficher un CA complet en face d'unités\n * partielles fausserait le panier moyen et le prix de vente moyen. */\nexport function preCutoverUnits(\n products: Product[], from: string, cutover = DEFAULT_CUTOVER,\n): number {\n if (from >= cutover) return 0;\n return products.reduce((s, p) => s + (p.salesPre ?? 0), 0);\n}\n\n// ── Aggregation ───────────────────────────────────────────────────────────────\n/** @param revBySku optional per-SKU revenue override (composed BI + Shopify figure). */\nexport function aggregate(products: Product[], revBySku?: Record) {\n const sales = products.reduce((s, p) => s + (p.sales || 0), 0);\n const fc = products.reduce((s, p) => s + (p.fc || 0), 0);\n const active = products.filter((p) => p.ss === \"Active\").length;\n const alerts = products.reduce((s, p) => s + productAlerts(p).length, 0);\n // Revenue = invoiced NSV from the BI (excl. VAT, net of discounts) — the same figure\n // as the internal reporting. Falls back to Σ site price × units only when the orders\n // export carries no NSV column, in which case it is VAT-inclusive and pre-discount.\n const hasNsv = products.some((p) => (p.rev ?? 0) > 0);\n const revOf = (p: Product) =>\n revBySku ? (revBySku[String(p.code)] ?? 0)\n : hasNsv ? (p.rev ?? 0)\n : (p.price ?? 0) * (p.sales || 0);\n const rev = products.reduce((s, p) => s + revOf(p), 0);\n // Revenue-weighted average current margin\n let wm = 0, wrev = 0;\n for (const p of products) {\n const w = revOf(p);\n if (p.mcur != null && w > 0) { wm += p.mcur * w; wrev += w; }\n }\n const avgMargin = wrev > 0 ? wm / wrev : null;\n return {\n count: products.length, sales, fc, st: fc > 0 ? sales / fc : null,\n active, alerts, rev, avgMargin, revIsInvoiced: hasNsv || !!revBySku,\n };\n}\n\nexport const fmtEurK = (n: number | null | undefined) => {\n if (n == null || !isFinite(n)) return \"—\";\n if (Math.abs(n) >= 1e6) return (n / 1e6).toFixed(2) + \" M€\";\n if (Math.abs(n) >= 1e3) return (n / 1e3).toFixed(1) + \" k€\";\n return n.toFixed(0) + \" €\";\n};\n\n/** Complète une série journalière avec les jours absents, à zéro.\n *\n * Une agrégation par date ne produit une entrée que pour les jours où il s'est passé\n * quelque chose. Tracée telle quelle, la série accole un vendredi et un lundi comme\n * s'ils étaient consécutifs : le graphique devient faux, et les journées réellement\n * faibles ressemblent à des trous de données. On matérialise donc les jours vides.\n *\n * Le remplissage se limite à l'intervalle réellement observé : un filtre plus large\n * que les données ne doit pas générer de barres vides en tête ou en fin de série. */\nexport function fillDays(series: [string, number][], until?: string): [string, number][] {\n if (series.length < 2) return series;\n const map = new Map(series);\n const last = series[series.length - 1][0];\n // `until` permet de prolonger jusqu'à aujourd'hui même si aucune donnée n'est encore\n // remontée : une journée connue mais vide doit se voir, sinon on la croit oubliée.\n const end = until && until > last ? until : last;\n const out: [string, number][] = [];\n const d = new Date(series[0][0] + \"T12:00:00Z\");\n for (let guard = 0; guard < 1000; guard++) {\n const iso = d.toISOString().slice(0, 10);\n out.push([iso, map.get(iso) ?? 0]);\n if (iso >= end) break;\n d.setUTCDate(d.getUTCDate() + 1);\n }\n return out;\n}\n\n/** Jour de week-end — sert à expliquer visuellement les journées à zéro. */\nexport function isWeekend(iso: string): boolean {\n const g = new Date(iso + \"T12:00:00Z\").getUTCDay();\n return g === 0 || g === 6;\n}\n\n/** Nom d'utilisateur affichable, jamais une adresse e-mail complète.\n * Les entrées d'audit déjà enregistrées contiennent parfois une adresse (et un motif\n * du type « a@b.com (cancelled by c@d.com) ») : on n'en garde que la partie locale,\n * ce qui préserve la lisibilité du journal sans exposer d'adresse personnelle.\n * L'historique stocké n'est pas réécrit — le masquage se fait à l'affichage. */\nexport function displayUser(v?: string | null): string {\n if (!v) return \"—\";\n return v.replace(/([\\w.+-]+)@[\\w.-]+\\.\\w+/g, \"$1\");\n}\n\nexport function groupBy(arr: T[], key: (t: T) => string): Record {\n const out: Record = {};\n for (const item of arr) (out[key(item)] ??= []).push(item);\n return out;\n}\n\n// ── CSV export ────────────────────────────────────────────────────────────────\nexport function downloadCSV(filename: string, headers: string[], rows: (string | number | null | undefined)[][]) {\n const esc = (v: string | number | null | undefined) => {\n const s = v == null ? \"\" : String(v);\n return /[\",;\\n]/.test(s) ? `\"${s.replace(/\"/g, '\"\"')}\"` : s;\n };\n const csv = [headers.map(esc).join(\";\"), ...rows.map((r) => r.map(esc).join(\";\"))].join(\"\\n\");\n const blob = new Blob([\"\" + csv], { type: \"text/csv;charset=utf-8\" });\n const a = document.createElement(\"a\");\n a.href = URL.createObjectURL(blob);\n a.download = filename;\n a.click();\n URL.revokeObjectURL(a.href);\n}\n",[],[220,537],"26c14301-3a9e-4fa1-86fc-794081f84c43","forecasting.ts","01a033f6-8ab0-7225-822e-e7aae72f860e",["D",1787578190526],"// Moteur de prévision — méthode calée sur l'historique réel, pas choisie a priori.\n//\n// ─────────────────────────────────────────────────────────────────────────────\n// CE QUE DISENT LES DONNÉES (backtest à origine glissante sur 13 mois, 522 SKU)\n//\n// 1. La demande est massivement intermittente : sur la classification\n// Syntetos-Boylan-Croston, 60 % des références sont « intermittentes » et 27 %\n// « heurtées ». Seules 1,1 % sont lisses. Aucune méthode saisonnière par référence\n// n'est estimable sur des séries pareilles.\n//\n// 2. Les méthodes conçues pour l'intermittence ne gagnent pas ici. Mesuré :\n// Croston MASE 1,237, SBA 1,066, TSB 1,087 — toutes battues par le naïf. C'est\n// cohérent avec la compétition M5, où le naïf saisonnier devançait Croston de 11,5 %.\n//\n// 3. L'extrapolation de tendance est désastreuse : biais +126 % à 6 mois pour une\n// dérive linéaire, +333 % pour une tendance amortie. La croissance du portefeuille\n// vient des NOUVELLES références (293 des 522 sont apparues en 2026), pas des\n// existantes — celles-ci reculent de 57 % d'un semestre à l'autre.\n//\n// 4. Ce qui marche : un niveau récent pour l'horizon court, une moyenne longue pour\n// l'horizon lointain. La fenêtre d'historique est le seul réglage qui compte.\n//\n// 5. La sur-prévision à horizon lointain (+25 % mesuré) vient presque entièrement des\n// références en fin de vie que l'on continue de projeter. Les ramener à zéro corrige\n// le biais à +5,8 % sans aucun coefficient arbitraire — c'est la raison pour laquelle\n// ce module n'applique AUCUNE décroissance automatique : le cycle de vie explique\n// mieux le phénomène qu'un pourcentage appliqué à tout le monde.\n//\n// Pas de saisonnalité : elle n'existe pas franchement sur le gros électroménager, et\n// les 13 mois disponibles ne permettraient de toute façon pas de la distinguer du cycle\n// de vie produit.\n// ─────────────────────────────────────────────────────────────────────────────\n\nexport interface ForecastParams {\n /** Fenêtre d'historique retenue pour estimer le niveau, en mois. Seul vrai levier. */\n window: number;\n /** Part marketplace visée à terme. null = pas d'hypothèse, aucun effet. */\n mpShareTarget: number | null;\n /** Part marketplace observée sur les dernières semaines, point de départ. */\n mpShareCurrent: number;\n /** Ramener à zéro la prévision des références en fin de vie. */\n zeroEndOfLife: boolean;\n}\n\nexport const DEFAULT_FORECAST_PARAMS: ForecastParams = {\n window: 6,\n mpShareTarget: null,\n mpShareCurrent: 0,\n zeroEndOfLife: true,\n};\n\n// ── Cycle de vie ─────────────────────────────────────────────────────────────\nexport type LifeStage = \"active\" | \"launch\" | \"end of life\" | \"unknown\";\n\n/** Deux sources décrivent le cycle de vie et ne s'accordent pas toujours : le statut\n * du catalogue (issu du SCP, le même que dans l'onglet Products) et le champ PIPO du\n * fichier Demand Planning. On retient le plus prudent des deux — si l'une des sources\n * annonce une fin de vie, on ne va pas prévoir des volumes comme si de rien n'était. */\nexport function lifeStage(catalogueLc?: string | null, pipo?: string | null): LifeStage {\n const lc = (catalogueLc ?? \"\").toLowerCase();\n const pi = (pipo ?? \"\").toUpperCase();\n\n if (/end of life|phasing out|discontinued|out of scp/.test(lc)) return \"end of life\";\n if (pi.startsWith(\"PHASE OUT\") || pi.startsWith(\"PHASING OUT\")) return \"end of life\";\n if (/new|nouveau/.test(lc) || pi.startsWith(\"PHASE IN\")) return \"launch\";\n if (/active/.test(lc) || pi.startsWith(\"ACTIF\")) return \"active\";\n return \"unknown\";\n}\n\n/** Moyenne pondérée des `window` derniers mois, poids décroissant en 0,5^(i/halfLife).\n * halfLife très grand ⇒ moyenne simple ; halfLife = 1 ⇒ le dernier mois pèse le double\n * du précédent. */\nexport function weightedLevel(hist: number[], halfLife: number, window: number): number {\n const seg = hist.slice(-window).reverse();\n if (seg.length === 0) return 0;\n // halfLife = Infinity signifie « moyenne simple ». Passer par 0,5^(i/1000) donnerait\n // un poids de 0,9986 au troisième mois : presque un, mais assez pour que la moyenne\n // ne soit pas exactement la moyenne, ce qui rend les contrôles impossibles à écrire.\n const flat = !isFinite(halfLife);\n let num = 0, den = 0;\n for (let i = 0; i < seg.length; i++) {\n const w = flat ? 1 : Math.pow(0.5, i / halfLife);\n num += seg[i] * w;\n den += w;\n }\n return den > 0 ? num / den : 0;\n}\n\n/** Niveau composite : réactif à l'horizon 1, stable à l'horizon 6.\n * Le poids bascule linéairement, ce qui reproduit le fait mesuré que la moyenne\n * récente gagne à court terme (MASE 1,024 contre 1,172) et perd à long terme\n * (0,934 contre 0,777). */\nexport function compositeLevel(hist: number[], horizon: number, window: number): number {\n const a = Math.max(0, Math.min(1, 1 - (horizon - 1) / 5));\n return a * weightedLevel(hist, 1, window) + (1 - a) * weightedLevel(hist, Infinity, window);\n}\n\nexport interface SkuForecastInput {\n hist: number[];\n /** Nombre de mois à prévoir. */\n horizon: number;\n stage: LifeStage;\n params: ForecastParams;\n}\n\nexport interface SkuForecastOutput {\n values: number[];\n level: number;\n /** Renseigné quand la prévision a été mise à zéro plutôt que calculée. */\n zeroedReason: string | null;\n}\n\nexport function forecastSku(input: SkuForecastInput): SkuForecastOutput {\n const { hist, horizon, stage, params } = input;\n const level = compositeLevel(hist, 1, params.window);\n\n // Une référence en fin de vie n'a pas vocation à être réapprovisionnée : on ne\n // prévoit rien, même si elle vend encore. C'est une décision d'assortiment, pas une\n // prédiction — et elle est réversible d'un clic.\n if (params.zeroEndOfLife && stage === \"end of life\") {\n return { values: Array(horizon).fill(0), level, zeroedReason: \"end of life\" };\n }\n\n // Montée en charge marketplace : le canal ajoute du volume sans rien retirer au DtoC,\n // donc on applique le rapport des « totaux pour 1 de DtoC ». Passer de 28 % à 35 % de\n // part marketplace, c'est multiplier le total par 1,108 — pas par 1,25.\n let uplift = 1;\n const t = params.mpShareTarget;\n if (t != null && params.mpShareCurrent < 0.999 && t < 0.999) {\n const u = (1 - params.mpShareCurrent) / (1 - t);\n if (isFinite(u) && u > 0) uplift = u;\n }\n\n const values: number[] = [];\n for (let h = 1; h <= horizon; h++) {\n values.push(compositeLevel(hist, h, params.window) * uplift);\n }\n return { values, level, zeroedReason: null };\n}\n\n// ── Classification de la demande ─────────────────────────────────────────────\nexport type DemandClass = \"smooth\" | \"erratic\" | \"intermittent\" | \"lumpy\" | \"no history\";\n\n/** Classification Syntetos-Boylan-Croston. Les seuils 1,32 et 0,49 sont ceux de la\n * littérature et servent à savoir à quel point une prévision par référence est fiable. */\nexport function classifyDemand(hist: number[]): { kind: DemandClass; adi: number; cv2: number } {\n const nz = hist.filter((v) => v > 0);\n if (nz.length === 0) return { kind: \"no history\", adi: 0, cv2: 0 };\n const adi = hist.length / nz.length;\n const mean = nz.reduce((s, v) => s + v, 0) / nz.length;\n const varc = nz.reduce((s, v) => s + (v - mean) ** 2, 0) / nz.length;\n const cv2 = mean > 0 ? varc / (mean * mean) : 0;\n const kind: DemandClass = adi < 1.32\n ? (cv2 < 0.49 ? \"smooth\" : \"erratic\")\n : (cv2 < 0.49 ? \"intermittent\" : \"lumpy\");\n return { kind, adi, cv2 };\n}\n\nexport type Reliability = \"high\" | \"medium\" | \"low\" | \"very low\" | \"n/a\";\n\n/** Traduit le profil de demande en fiabilité attendue de la prévision.\n *\n * C'est la même information, dite dans les termes de la décision à prendre : le profil\n * décrit la forme de la série, la fiabilité dit à quel point on peut s'appuyer sur le\n * chiffre proposé. Les deux dimensions jouent dans le même sens — plus la demande est\n * espacée et plus les volumes sautent, moins une prévision par référence tient. */\nexport function reliabilityOf(kind: DemandClass): Reliability {\n switch (kind) {\n case \"smooth\": return \"high\";\n case \"erratic\": return \"medium\";\n case \"intermittent\": return \"low\";\n case \"lumpy\": return \"very low\";\n default: return \"n/a\";\n }\n}\n\n/** Phrase d'explication, pour l'infobulle. */\nexport function reliabilityReason(kind: DemandClass): string {\n switch (kind) {\n case \"smooth\": return \"Sells every month at a steady rate — the forecast is dependable.\";\n case \"erratic\": return \"Sells every month but volumes swing — expect the total to be right, the monthly split less so.\";\n case \"intermittent\": return \"Months with no sales at all — the monthly split is largely arbitrary, judge the 6-month total.\";\n case \"lumpy\": return \"Months with no sales AND very uneven volumes — the least predictable case, worth a manual figure.\";\n default: return \"No sales history to work from.\";\n }\n}\n\n// ── Nouveaux produits : recherche d'analogue ─────────────────────────────────\nexport interface AnalogueCandidate {\n code: number;\n name: string;\n brand: string;\n pl: string;\n hist: number[];\n price?: number | null;\n cost?: number | null;\n grp?: string | null;\n}\n\nexport interface AnalogueMatch {\n candidate: AnalogueCandidate;\n score: number;\n reasons: string[];\n /** Courbe de démarrage de l'analogue : ses `n` premiers mois de vie commerciale. */\n rampUp: number[];\n}\n\n/** Trouve les références les plus proches d'un nouveau produit.\n *\n * L'approche est celle du « forecast par analogue » : à défaut d'historique propre,\n * on emprunte celui d'un produit comparable, puis on l'ajuste par le rapport de prix.\n * Le score combine gamme, marque et proximité tarifaire — ce sont les attributs dont\n * nous disposons de façon fiable, et il vaut mieux trois critères solides que dix\n * approximatifs.\n *\n * Seules les références ayant réellement démarré sont éligibles : reprendre la courbe\n * d'un produit qui n'a jamais décollé reviendrait à prévoir zéro. */\nexport function findAnalogues(\n target: { pl: string; brand: string; price?: number | null; grp?: string | null },\n pool: AnalogueCandidate[],\n limit = 5,\n rampMonths = 6,\n): AnalogueMatch[] {\n const out: AnalogueMatch[] = [];\n\n for (const c of pool) {\n const total = c.hist.reduce((s, v) => s + v, 0);\n if (total < 5) continue; // démarrage trop faible pour servir de modèle\n\n let score = 0;\n const reasons: string[] = [];\n\n if (c.pl && target.pl && c.pl === target.pl) { score += 50; reasons.push(\"same range\"); }\n else if (c.grp && target.grp && c.grp === target.grp) { score += 25; reasons.push(\"same family\"); }\n\n if (c.brand === target.brand) { score += 15; reasons.push(\"same brand\"); }\n\n if (target.price && c.price) {\n const ratio = Math.min(target.price, c.price) / Math.max(target.price, c.price);\n score += ratio * 35;\n const gap = Math.abs(target.price - c.price) / target.price;\n if (gap <= 0.15) reasons.push(\"close price\");\n else if (gap <= 0.35) reasons.push(\"comparable price\");\n }\n\n if (score <= 0) continue;\n\n const first = c.hist.findIndex((v) => v > 0);\n const rampUp = first === -1 ? [] : c.hist.slice(first, first + rampMonths);\n out.push({ candidate: c, score, reasons, rampUp });\n }\n\n return out.sort((a, b) => b.score - a.score).slice(0, limit);\n}\n\n/** Prévision d'un nouveau produit à partir de ses analogues.\n *\n * Trois principes, chacun corrigeant une erreur constatée sur les données réelles.\n *\n * 1. ON N'EXTRAPOLE JAMAIS AU-DELÀ DE CE QUE L'ANALOGUE A VÉCU. La version précédente\n * prolongeait la dernière valeur observée : un analogue lancé il y a deux mois et\n * monté à 176 unités voyait ce 176 recopié sur les quatre mois suivants. D'où des\n * projections plates à 31 ou 38 par mois pour des références qui vendent une unité.\n * Désormais un analogue ne pèse que sur les mois qu'il a réellement traversés.\n *\n * 2. ON CALE SUR L'ÂGE DU PRODUIT. Une référence qui vend depuis trois mois n'est pas\n * à son mois 1 : on lit la courbe de l'analogue à partir du même âge, sinon on lui\n * promet un démarrage qu'elle a déjà raté.\n *\n * 3. ON INDEXE SUR SON PROPRE DÉMARRAGE. Si le produit a vendu 1 unité là où ses\n * analogues en faisaient 8, il tourne à 12 % de leur rythme et il n'y a aucune\n * raison qu'il les rattrape. Le ratio observé/attendu corrige la projection —\n * c'est le geste standard du demand planning, et c'est ce qui manquait le plus.\n *\n * Le rapport de prix reste appliqué par-dessus : un produit deux fois plus cher se vend\n * rarement au même rythme. L'élasticité de -0,7 est une hypothèse explicite. */\nexport function forecastNewProduct(\n matches: AnalogueMatch[],\n targetPrice: number | null | undefined,\n horizon = 6,\n opts: { age?: number; observed?: number; elasticity?: number } = {},\n): { values: number[]; basedOn: number; index: number | null; expected: number } {\n const { age = 0, observed = 0, elasticity = -0.7 } = opts;\n const usable = matches.filter((m) => m.rampUp.length > 0);\n if (usable.length === 0) {\n return { values: Array(horizon).fill(0), basedOn: 0, index: null, expected: 0 };\n }\n\n const priceScale = (m: AnalogueMatch) => {\n if (!targetPrice || !m.candidate.price) return 1;\n const s = Math.pow(targetPrice / m.candidate.price, elasticity);\n return Math.max(0.3, Math.min(3, s)); // garde-fou contre un prix aberrant\n };\n\n // ── Indexation sur le démarrage réel ──\n // Ce qu'un produit moyen de ce groupe aurait fait sur ses `age` premiers mois.\n let expected = 0, expWeight = 0;\n if (age > 0) {\n for (const m of usable) {\n if (m.rampUp.length < age) continue; // pas assez de recul pour comparer\n const s = m.rampUp.slice(0, age).reduce((a, b) => a + b, 0) * priceScale(m);\n expected += s * m.score;\n expWeight += m.score;\n }\n if (expWeight > 0) expected /= expWeight;\n }\n // Bornes larges : on corrige un ordre de grandeur, on ne recopie pas un accident.\n const index = age > 0 && expected > 0\n ? Math.max(0.15, Math.min(3, observed / expected))\n : null;\n\n const values: number[] = [];\n for (let h = 0; h < horizon; h++) {\n const idx = age + h; // même âge que le produit visé\n // MÉDIANE et non moyenne. Les courbes disponibles portent toutes la montée en charge\n // marketplace de juin-juillet 2026 : un analogue passé de 6 à 176 unités en un mois\n // n'illustre pas la maturation d'un produit, il illustre l'ouverture d'un canal.\n // Une moyenne laisse ce pic dicter la projection, une médiane l'ignore.\n const vals = usable\n .filter((m) => idx < m.rampUp.length)\n .map((m) => m.rampUp[idx] * priceScale(m))\n .sort((a, b) => a - b);\n // Aucun analogue ne couvre ce mois : on prolonge le dernier niveau calculé plutôt\n // que d'inventer une montée en charge que rien n'atteste.\n const raw = vals.length === 0\n ? (values[h - 1] ?? 0) / (index ?? 1)\n : vals.length % 2\n ? vals[(vals.length - 1) / 2]\n : (vals[vals.length / 2 - 1] + vals[vals.length / 2]) / 2;\n values.push(raw * (index ?? 1));\n }\n\n // ── Amortissement de la montée en charge ──\n // Même après médiane, les mois lointains des analogues tombent sur juin-juillet 2026,\n // c'est-à-dire sur l'ouverture des marketplaces : une référence lancée en février voit\n // son « mois 6 » multiplié par dix pour une raison qui ne doit rien à sa maturation.\n // On plafonne donc la croissance d'un mois sur l'autre. À 1,6 par mois, un lancement\n // peut encore être multiplié par onze sur six mois — on interdit le saut brutal, pas\n // la montée en charge.\n const MAX_GROWTH = 1.6;\n for (let h = 1; h < values.length; h++) {\n const ceiling = Math.max(values[h - 1], 1) * MAX_GROWTH;\n if (values[h] > ceiling) values[h] = ceiling;\n }\n\n return { values, basedOn: usable.length, index, expected };\n}\n\n// ── Qualité de la prévision en vigueur ───────────────────────────────────────\n/** Compare la prévision commerciale au réalisé sur les mois échus.\n * Le biais compte davantage que l'erreur absolue : une prévision qui se trompe au\n * hasard coûte du stock, une prévision systématiquement basse coûte des ventes. */\nexport function forecastQuality(\n hist: number[], fc: number[], monthsElapsed: number,\n): { bias: number | null; mape: number | null; actual: number; forecast: number } {\n let a = 0, f = 0, absErr = 0, n = 0;\n for (let i = 0; i < monthsElapsed && i < hist.length; i++) {\n a += hist[i];\n f += fc[i] ?? 0;\n if (hist[i] > 0) { absErr += Math.abs((fc[i] ?? 0) - hist[i]) / hist[i]; n++; }\n }\n return {\n bias: a > 0 ? (f - a) / a : null,\n mape: n > 0 ? absErr / n : null,\n actual: a,\n forecast: f,\n };\n}\n",[],[220,544],"9218607f-8994-405a-9a86-9992ed5b8aa3","favicon.svg","01a033f6-8ab0-7225-822e-e2c6f992a991",["D",1787578190526],"\n",[],[220,551],"34d6e5bc-f1c2-40e0-b61e-3c0bc0cb4437","Simulator.tsx","01a033f6-8ab0-7225-822e-dec0953acb3b",["D",1787578190526],"/** @jsxImportSource https://esm.sh/react@18.2.0 */\nimport { useMemo, useState } from \"https://esm.sh/react@18.2.0\";\nimport { fmtEur, fmtPct, GROUPS, type Product } from \"/frontend/helpers.ts\";\nimport {\n breakeven, breakevenWithDeee, carrierFor, CATEGORIES, contributionMargin, deeeBands,\n grossMargin, logisticsCost, nearBracketBoundary, officialDeee, PARAMS,\n priceForGrossMargin, roundPsy,\n} from \"/frontend/pricing.ts\";\nimport * as Ico from \"/frontend/Icons.tsx\";\n\nconst NAVY = \"#1F3864\";\nconst ACCENT = \"#4F46E5\";\n\n/** Pricing simulator for a product launch: enter category, standard cost and weight,\n * read the breakeven, the margins, and the price needed to hit a margin target. */\nexport default function Simulator(\n { products, Card, Kpi, Pill }: {\n products: Product[];\n Card: any; Kpi: any; Pill: any;\n },\n) {\n const [cat, setCat] = useState(\"Washing Machine\");\n const [cost, setCost] = useState(\"180\");\n const [weight, setWeight] = useState(\"70\");\n const [screen, setScreen] = useState(\"55\"); // TVs are tariffed on screen size\n // null = follow the official scale; a string = the user has overridden it\n const [deeeOverride, setDeeeOverride] = useState(null);\n const [target, setTarget] = useState(\"25\");\n const [comp, setComp] = useState(\"\");\n\n const nCost = parseFloat(cost.replace(\",\", \".\")) || null;\n const nWeight = parseFloat(weight.replace(\",\", \".\")) || null;\n const nTarget = (parseFloat(target.replace(\",\", \".\")) || 0) / 100;\n const nComp = parseFloat(comp.replace(\",\", \".\")) || null;\n\n // Eco-participation from the official ecosystem scale (2026), driven by net weight\n // or — for TVs — by screen size.\n const nScreen = parseFloat(screen.replace(\",\", \".\")) || null;\n const official = useMemo(\n () => officialDeee(cat, nWeight, nScreen),\n [cat, nWeight, nScreen],\n );\n const boundary = nearBracketBoundary(cat, nWeight, nScreen);\n const autoDeee = official?.value ?? null;\n const deee = deeeOverride ?? (autoDeee != null ? autoDeee.toFixed(2) : \"0\");\n const nDeee = parseFloat(deee.replace(\",\", \".\")) || 0;\n const deeeIsAuto = deeeOverride === null;\n // What the catalogue actually applies today — a cross-check on the scale mapping\n const bands = useMemo(() => deeeBands(products, cat), [products, cat]);\n const catalogueDisagrees = autoDeee != null && bands.length > 0 &&\n !bands.some((b) => Math.abs(b.value - autoDeee) < 0.005);\n\n const carrier = carrierFor(nWeight, cat);\n const logi = logisticsCost(nWeight, carrier);\n const be = breakeven(nCost, logi, carrier);\n const beReal = breakevenWithDeee(nCost, nDeee, logi, carrier);\n const targetRaw = priceForGrossMargin(nTarget, nCost, nDeee);\n const targetPrice = targetRaw != null ? roundPsy(targetRaw) : null;\n\n const beGross = grossMargin(be, nCost, nDeee);\n const beContrib = contributionMargin(be, nCost, nDeee, logi, carrier);\n const tGross = grossMargin(targetPrice, nCost, nDeee);\n const tContrib = contributionMargin(targetPrice, nCost, nDeee, logi, carrier);\n\n // Recommended price = max(breakeven, cheapest competitor), rounded to x9.99 —\n // the same rule the catalogue pipeline applies to compute RSP.\n const rsp = be != null\n ? roundPsy(Math.max(nComp && nComp > 0 ? nComp : be, be))\n : null;\n const rspGross = grossMargin(rsp, nCost, nDeee);\n const rspContrib = contributionMargin(rsp, nCost, nDeee, logi, carrier);\n\n // Comparable products already in the catalogue — sanity-check the assumptions\n const comparables = useMemo(() => {\n if (!nWeight) return [];\n return products\n .filter((p) => p.cat === cat && p.wt != null && p.cost != null)\n .map((p) => ({ p, dw: Math.abs((p.wt ?? 0) - nWeight) }))\n .sort((a, b) => a.dw - b.dw)\n .slice(0, 6)\n .map((x) => x.p);\n }, [products, cat, nWeight]);\n\n const overDpdGrid = carrier === \"DPD\" && logi == null;\n const grpOf = (c: string) => CATEGORIES.find((g) => g.cats.includes(c))?.grp ?? \"Other\";\n\n // Margin ladder: what price each margin target implies\n const ladder = [0.15, 0.20, 0.25, 0.30, 0.35, 0.40].map((m) => {\n const raw = priceForGrossMargin(m, nCost, nDeee);\n const price = raw != null ? roundPsy(raw) : null;\n return { m, price, contrib: contributionMargin(price, nCost, nDeee, logi, carrier) };\n });\n\n return (\n

\n \n

\n \n \n \n \n \n \n \n \n \n {official?.tariff.driver === \"screen\" && (\n \n \n \n )}\n Eco-participation (DEEE){\" \"}\n {deeeIsAuto && autoDeee != null\n ? barème 2026\n : }\n }\n hint={official\n ? `${official.tariff.family} · ${official.bracket.label} · code ${official.bracket.code}`\n : \"Excl. VAT, per unit — this category is not in the scale\"}>\n \n \n \n \n \n

\n \n \n \n

\n Group: {grpOf(cat)}\n Carrier: {carrier}\n {\" \"}({carrier === \"DPD\" ? \"small appliance ≤ 30 kg\" : \"large appliance\"})\n Logistics cost: {fmtEur(logi)}/unit\n Return rate: {fmtPct(carrier === \"CCV\" ? PARAMS.RET_MDA : PARAMS.RET_SDA)}\n Delivery charged: {fmtEur(carrier === \"CCV\" ? PARAMS.SHIP_MDA : PARAMS.SHIP_SDA)} incl. VAT\n

\n {official ? (\n

\n \n Official scale — {official.tariff.family}\n {official.tariff.driver !== \"none\"\n ? official.tariff.driver === \"screen\" ? \" (by screen size)\" : \" (by net weight)\"\n : \"\"} :\n \n {official.tariff.brackets.map((b) => {\n const active = Math.abs(b.value - nDeee) < 0.005;\n return (\n \n );\n })}\n

\n {official.tariff.note && (\n

{official.tariff.note}

\n )}\n {boundary != null && (\n

\n You are within 10 % of the {boundary}\n {official.tariff.driver === \"screen\" ? \"\\\"\" : \" kg\"} boundary. The scale applies to\n the net weight, while the catalogue often carries the packaged weight —\n check the net figure before committing to this bracket.\n

\n )}\n {catalogueDisagrees && (\n

\n The catalogue applies {bands.map((b) => fmtEur(b.value)).join(\" / \")} to this\n category, which does not include {fmtEur(autoDeee ?? 0)}. Worth checking which of\n the two is out of date.\n

\n )}\n

\n ) : (\n

\n {cat} is not mapped to the ecosystem scale — the field defaults to 0.\n Enter the value manually if this appliance is subject to one.\n

\n )}\n {overDpdGrid && (\n

\n Above 30 kg the DPD grid does not apply — the pipeline would route this product to\n CCV. Check the category: a small-appliance category at this weight is unusual.\n

\n )}\n \n\n {be == null ? (\n \n

\n Enter a standard cost and a weight to compute the breakeven.\n

\n \n ) : (\n <>\n

\n \n \n \n \n

\n\n \n

\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n {nComp && nComp > 0 && (\n \n )}\n \n

Scenario	Price incl. VAT	Price excl. VAT	Gross margin	Contribution €	Contribution %

\n Gross margin = (price excl. VAT − standard cost − eco-participation) / price excl. VAT —\n the same definition as the Margin column elsewhere in the dashboard.\n Contribution additionally deducts logistics grossed up for returns, OPEX {fmtPct(PARAMS.OPEX, 1)} and\n marketing {fmtPct(PARAMS.MKTG, 0)}, and adds back the delivery the customer pays.\n

\n {beReal != null && be != null && (\n

\n The catalogue breakeven formula does not include the eco-participation.\n Contribution actually reaches zero at {fmtEur(beReal)} incl. VAT\n {beReal > be ? <>, i.e. {fmtEur(beReal - be)} above the {fmtEur(be)} shown as\n breakeven — selling at breakeven loses that amount per unit. : \".\"}\n

\n )}\n \n\n

\n \n

\n \n \n \n \n \n \n \n \n {ladder.map(({ m, price, contrib }) => (\n \n \n \n \n \n \n ))}\n \n

Gross margin	Price incl. VAT	Contribution %	vs breakeven
{fmtPct(m, 0)}	{fmtEur(price)}	\n {fmtPct(contrib?.rate)}\n	\n {price != null && be != null\n ? (price >= be ? \"+\" : \"\") + fmtEur(price - be) : \"—\"}\n

\n \n\n \n {targetPrice != null && logi != null && (\n \n )}\n \n

\n\n {comparables.length > 0 && (\n \n

\n \n \n \n \n \n \n \n \n \n \n \n \n {comparables.map((p) => (\n \n \n \n \n \n \n \n \n \n \n ))}\n \n

SKU	Product	Weight	Std cost	Breakeven	Site price	Margin	Sales YTD
{p.code}	{p.name}	{p.wt?.toFixed(1)} kg	{fmtEur(p.cost)}	{fmtEur(p.be)}	{fmtEur(p.price)}	{fmtPct(p.mcur)}	{p.sales}

\n Same category, weights closest to {nWeight} kg. Use them to sanity-check the\n standard cost and to see where the market actually sits.\n

\n \n )}\n \n )}\n

\n );\n}\n\nfunction Field({ label, hint, children }: any) {\n return (\n \n {label}\n {children}\n {hint && {hint}}\n \n );\n}\n\nfunction NumInput({ value, onChange, suffix, placeholder }: any) {\n return (\n

\n onChange(e.target.value)}\n placeholder={placeholder} inputMode=\"decimal\"\n className=\"w-full px-2 py-2 border rounded-lg text-sm pr-8\" />\n {suffix}\n

\n );\n}\n\nfunction ScenarioRow({ label, price, gm, contrib, strong, muted }: any) {\n return (\n \n {label}\n {fmtEur(price)}\n {price != null ? fmtEur(price / (1 + PARAMS.VAT)) : \"—\"}\n {fmtPct(gm)}\n \n {contrib ? fmtEur(contrib.amount) : \"—\"}\n \n \n {fmtPct(contrib?.rate)}\n \n \n );\n}\n\n/** Waterfall from price excl. VAT down to contribution, with the two finance-relevant\n * subtotals: contribution to fixed costs (before OPEX and marketing) and the figure\n * left once those overheads are charged. Which one is \"the\" contribution depends on\n * whether OPEX and marketing behave as variable costs or as allocated fixed pools. */\nfunction CostBreakdown({ price, cost, deee, logi, carrier }: any) {\n const { OPEX, MKTG, VAT, RET_MDA, RET_SDA, SHIP_MDA, SHIP_SDA } = PARAMS;\n const ret = carrier === \"CCV\" ? RET_MDA : RET_SDA;\n const ship = carrier === \"CCV\" ? SHIP_MDA : SHIP_SDA;\n const ht = price / (1 + VAT);\n\n const direct: [string, number][] = [\n [\"Price excl. VAT\", ht],\n [\"Standard cost\", -cost],\n [\"Eco-participation (DEEE)\", -deee],\n [`Logistics incl. returns (${fmtPct(ret)})`, -logi * (1 + ret)],\n [\"Delivery paid by customer\", ship / (1 + VAT)],\n ];\n const overhead: [string, number][] = [\n [`OPEX (${fmtPct(OPEX, 1)})`, -ht * OPEX],\n [`Marketing (${fmtPct(MKTG, 0)})`, -ht * MKTG],\n ];\n const toFixed = direct.reduce((s, [, v]) => s + v, 0);\n const total = toFixed + overhead.reduce((s, [, v]) => s + v, 0);\n\n const Row = ({ label, v, head }: any) => (\n \n {label}\n \n {v >= 0 && !head ? \"+\" : \"\"}{fmtEur(v)}\n \n {ht > 0 && !head ? fmtPct(v / ht, 1) : \"\"}\n \n );\n const Subtotal = ({ label, v, strong }: any) => (\n \n {label}\n {fmtEur(v)}\n \n {ht > 0 ? fmtPct(v / ht, 1) : \"\"}\n \n \n );\n\n return (\n <>\n

\n \n \n {direct.map(([l, v], i) => )}\n \n {overhead.map(([l, v]) => )}\n \n \n

\n The first subtotal is the true contribution margin if OPEX and marketing are fixed pools\n allocated as a % of sales; the second is the right one if they are genuinely variable.\n

\n \n );\n}\n",[],[220,558],"370b7ca6-abcb-4df5-99c7-f14b479b62d1","Parameters.tsx","01a033f6-8ab0-7225-822e-dbec6f8602c7",["D",1787578190526],"/** @jsxImportSource https://esm.sh/react@18.2.0 */\n// Onglet Parameters — édition des paramètres économiques.\n//\n// L'onglet n'est proposé qu'aux éditeurs, mais c'est cosmétique : la vraie barrière est\n// PUT /api/params côté serveur, qui refuse un compte en lecture seule.\n//\n// Point important pour l'utilisateur, rappelé dans l'interface : modifier un paramètre\n// ne recalcule PAS instantanément les colonnes Breakeven et Logistics de l'onglet\n// Products. Celles-ci viennent du pipeline Python et ne bougeront qu'à son prochain\n// passage. Le simulateur, lui, réagit immédiatement.\nimport { useEffect, useState } from \"https://esm.sh/react@18.2.0\";\nimport { DPD_RATES, CCV_BRACKETS, DPD_CATEGORIES } from \"/frontend/pricing.ts\";\nimport { displayUser, fmtEur } from \"/frontend/helpers.ts\";\n\nconst NAVY = \"#1F3864\";\nconst ACCENT = \"#4F46E5\";\n\ninterface Def {\n key: string; label: string; help: string;\n unit: \"pct\" | \"eur\"; min: number; max: number; fallback: number; cell: string;\n}\n\ninterface Record_ {\n values: Record;\n updated_at: string | null;\n updated_by: string | null;\n history: { at: string; user: string; key: string; from: number; to: number }[];\n defs: Def[];\n}\n\n/** Affichage : une fraction est présentée en pourcentage, un montant en euros. */\nconst toDisplay = (v: number, unit: string) =>\n unit === \"pct\" ? String(Math.round(v * 1e6) / 1e4) : String(Math.round(v * 1e4) / 1e4);\n\n/** Saisie : l'inverse. On accepte la virgule décimale, réflexe naturel en français. */\nconst toStored = (s: string, unit: string) => {\n const n = parseFloat(s.replace(\",\", \".\"));\n if (!isFinite(n)) return NaN;\n return unit === \"pct\" ? n / 100 : n;\n};\n\nexport default function Parameters({ onSaved }: { onSaved?: () => void }) {\n const [rec, setRec] = useState(null);\n const [draft, setDraft] = useState>({});\n const [busy, setBusy] = useState(false);\n const [msg, setMsg] = useState<{ kind: \"ok\" | \"err\"; text: string } | null>(null);\n\n const load = async () => {\n const r = await fetch(\"/api/params\");\n if (!r.ok) { setMsg({ kind: \"err\", text: \"Could not load parameters.\" }); return; }\n const d: Record_ = await r.json();\n setRec(d);\n setDraft(Object.fromEntries(d.defs.map((f) => [f.key, toDisplay(d.values[f.key], f.unit)])));\n };\n\n useEffect(() => { load(); }, []);\n\n if (!rec) return

Loading…

;\n\n const dirty = rec.defs.filter((f) => {\n const v = toStored(draft[f.key] ?? \"\", f.unit);\n return isFinite(v) && Math.abs(v - rec.values[f.key]) > 1e-9;\n });\n\n const invalid = rec.defs.filter((f) => {\n const v = toStored(draft[f.key] ?? \"\", f.unit);\n return !isFinite(v) || v < f.min || v > f.max;\n });\n\n const save = async () => {\n setBusy(true); setMsg(null);\n const values: Record = {};\n for (const f of rec.defs) values[f.key] = toStored(draft[f.key] ?? \"\", f.unit);\n const r = await fetch(\"/api/params\", {\n method: \"PUT\",\n headers: { \"Content-Type\": \"application/json\" },\n body: JSON.stringify({ values }),\n });\n const d = await r.json();\n setBusy(false);\n if (!r.ok || !d.ok) {\n setMsg({ kind: \"err\", text: (d.errors ?? [\"Save rejected.\"]).join(\" · \") });\n return;\n }\n setMsg({\n kind: \"ok\",\n text: d.changed.length === 0\n ? \"Nothing to save.\"\n : `${d.changed.length} parameter${d.changed.length > 1 ? \"s\" : \"\"} saved.`,\n });\n await load();\n onSaved?.();\n };\n\n const reset = () =>\n setDraft(Object.fromEntries(rec.defs.map((f) => [f.key, toDisplay(rec.values[f.key], f.unit)])));\n\n const restoreDefaults = () =>\n setDraft(Object.fromEntries(rec.defs.map((f) => [f.key, toDisplay(f.fallback, f.unit)])));\n\n return (\n

\n \n {rec.updated_at\n ? `Changed on ${new Date(rec.updated_at).toLocaleString(\"fr-FR\")} by ${displayUser(rec.updated_by)}`\n : \"Original values (never changed)\"}\n \n }>\n

\n These values drive breakeven, logistics cost and margins across the dashboard. The\n Pricing Simulator applies them immediately. The Breakeven and\n Logistics columns in Products, however, are computed by the pipeline: they\n will only reflect a change at its next run.\n

\n\n {/* La colonne Definition est verbeuse : sans défilement le tableau imposait\n sa largeur à toute la page sur mobile. */}\n

\n \n \n \n \n \n \n \n \n \n \n {rec.defs.map((f) => {\n const v = toStored(draft[f.key] ?? \"\", f.unit);\n const bad = !isFinite(v) || v < f.min || v > f.max;\n const moved = isFinite(v) && Math.abs(v - rec.values[f.key]) > 1e-9;\n const custom = Math.abs(rec.values[f.key] - f.fallback) > 1e-9;\n return (\n \n \n \n \n \n \n );\n })}\n \n

Parameter	Value	Original	Definition
\n {f.label}\n {f.cell}\n	\n \n setDraft({ ...draft, [f.key]: e.target.value })}\n inputMode=\"decimal\"\n className={`w-20 px-2 py-1 border rounded text-right tabular-nums ${\n bad ? \"border-rose-400 bg-rose-50\" : moved ? \"border-indigo-400 bg-indigo-50\" : \"border-slate-300\"\n }`} />\n {f.unit === \"pct\" ? \"%\" : \"€\"}\n \n	\n {toDisplay(f.fallback, f.unit)}{f.unit === \"pct\" ? \" %\" : \" €\"}\n {custom && ●}\n	\n {f.help}\n {bad && (\n \n Expected between {toDisplay(f.min, f.unit)} and {toDisplay(f.max, f.unit)}\n {f.unit === \"pct\" ? \" %\" : \" €\"}.\n \n )}\n

\n\n

\n \n \n \n

\n\n {msg && (\n

\n {msg.text}\n

\n )}\n \n\n {rec.history.length > 0 && (\n \n

\n \n \n \n \n \n \n \n \n \n {rec.history.map((h, i) => {\n const f = rec.defs.find((d) => d.key === h.key);\n const u = f?.unit ?? \"eur\";\n const suffix = u === \"pct\" ? \" %\" : \" €\";\n return (\n \n \n \n \n \n \n \n );\n })}\n \n

Date	By	Parameter	Before	After
\n {new Date(h.at).toLocaleString(\"fr-FR\")}\n	{displayUser(h.user)}	{f?.label ?? h.key}	{toDisplay(h.from, u)}{suffix}	{toDisplay(h.to, u)}{suffix}

\n \n )}\n\n \n

\n These rates come from the carrier contracts and cannot be edited here. DPD applies to\n small appliances up to 30 kg; above that, or outside the listed categories, shipping\n goes through CCV.\n

DPD — per kg (incl. VAT)

\n {Object.entries(DPD_RATES).map(([kg, r]) => (\n

\n ≤ {kg} kg\n {fmtEur(r)}\n

\n ))}\n

\n DPD categories: {[...DPD_CATEGORIES].join(\", \")}.\n

CCV — per weight bracket (incl. VAT)

\n {CCV_BRACKETS.map(([kg, r]) => (\n

\n ≥ {kg} kg\n {fmtEur(r)}\n

\n ))}\n

\n The CCV cost also includes WEEE take-back: gross cost × take-back rate.\n

\n \n

\n );\n}\n\nfunction Card({ title, children, right }: any) {\n return (\n

\n {title && (\n

{title}

\n {right}\n

\n )}\n

{children}

\n );\n}\n",[],[220,565],"3a751acf-fe18-4ba1-9ea1-4df3a6b0aeab","Icons.tsx","01a033f6-8ab0-7225-822e-d5ac74478f10",["D",1787578190526],"/** @jsxImportSource https://esm.sh/react@18.2.0 */\n// Monochrome stroke icons (inherit currentColor) — no external assets.\n\nconst base = {\n fill: \"none\",\n stroke: \"currentColor\",\n strokeWidth: 1.7,\n strokeLinecap: \"round\" as const,\n strokeLinejoin: \"round\" as const,\n};\n\nconst Svg = ({ children, size = 20 }: any) => (\n \n);\n\nexport const IconOverview = (p: any) => (\n \n);\nexport const IconProducts = (p: any) => (\n \n);\nexport const IconScorecard = (p: any) => (\n \n);\nexport const IconBusiness = (p: any) => (\n \n);\nexport const IconLogistic = (p: any) => (\n \n);\nexport const IconAction = (p: any) => (\n \n);\nexport const IconChanges = (p: any) => (\n \n);\nexport const IconLogout = (p: any) => (\n \n);\nexport const IconSearch = (p: any) => (\n \n);\nexport const IconRefresh = (p: any) => (\n \n);\n\n// KPI tile icons\nexport const IconBox = (p: any) => (\n \n);\nexport const IconCheck = (p: any) => ();\nexport const IconCart = (p: any) => (\n \n);\nexport const IconEuro = (p: any) => (\n \n);\nexport const IconForecast = (p: any) => (\n \n);\nexport const IconTrend = (p: any) => (\n \n);\nexport const IconAlert = (p: any) => (\n \n);\nexport const IconUnits = (p: any) => (\n \n);\nexport const IconTruck = IconLogistic;\nexport const IconScale = (p: any) => (\n \n);\nexport const IconClock = (p: any) => (\n \n);\nexport const IconReceipt = (p: any) => (\n \n);\nexport const IconReturn = (p: any) => (\n \n);\nexport const IconMargin = (p: any) => (\n \n);\nexport const IconPrint = (p: any) => (\n \n);\nexport const IconDownload = (p: any) => (\n \n);\nexport const IconColumns = (p: any) => (\n \n);\n/** Pricing simulator — price tag with a calculator dot grid */\nexport const IconSimulator = (p: any) => (\n \n);\n/** Read-only indicator */\nexport const IconLock = (p: any) => (\n \n);\n/** Ouverture du tiroir de navigation sur petit écran */\nexport const IconMenu = (p: any) => (\n \n);\n/** Fermeture du tiroir / d'une modale */\nexport const IconClose = (p: any) => (\n \n);\nexport const IconSave = (p: any) => (\n \n);\n\n// ── Product category icons (appliances) ──────────────────────────────────────\n/** Cooling — fridge / freezer */\nexport const IconFridge = (p: any) => (\n \n);\n/** Washing — washing machine / dryer */\nexport const IconWasher = (p: any) => (\n \n);\n/** Cooking — oven */\nexport const IconOven = (p: any) => (\n \n);\n/** Floorcare — vacuum cleaner */\nexport const IconVacuum = (p: any) => (\n \n);\n/** TV — television */\nexport const IconTv = (p: any) => (\n \n);\n/** KSDA — coffee machine */\nexport const IconCoffee = (p: any) => (\n \n);\n/** Consumable — detergent bottle */\nexport const IconDetergent = (p: any) => (\n \n);\n/** Fallback — other */\nexport const IconOther = (p: any) => (\n \n);\n",[],[220,572],"afb929e1-91c4-4e2c-b6e0-f8cff737fb87",["D",1787578219985],[],[220,33],"6ccf59a7-484d-46a2-a7ea-b8dcde6aa001","ErrorBoundary.tsx","01a033f6-8ab0-7225-822e-cf88b7de77af",["D",1787578190526],"/** @jsxImportSource https://esm.sh/react@18.2.0 */\n// Filet de sécurité contre l'écran blanc.\n//\n// React démonte TOUT l'arbre dès qu'un composant lève une erreur pendant le rendu.\n// Une seule ligne fautive dans un panneau secondaire suffit donc à faire disparaître\n// le dashboard entier, sans message — c'est exactement ce qui s'est produit en\n// cliquant sur un SKU dans Forecast : le panneau de détail lisait une propriété\n// absente et l'application entière tombait.\n//\n// Une frontière d'erreur contient la casse : seule la zone fautive est remplacée par\n// un message, le reste du dashboard continue de fonctionner. Elle ne corrige aucun\n// bug, elle empêche qu'un bug local devienne une panne totale.\nimport { Component, type ReactNode } from \"https://esm.sh/react@18.2.0\";\n\ninterface Props {\n children: ReactNode;\n /** Nom de la zone protégée, affiché dans le message pour situer le problème. */\n area?: string;\n}\ninterface State { error: Error | null }\n\nexport default class ErrorBoundary extends Component {\n state: State = { error: null };\n\n static getDerivedStateFromError(error: Error): State {\n return { error };\n }\n\n componentDidCatch(error: Error, info: any) {\n // La console reste la source de vérité pour diagnostiquer : on y laisse la trace\n // complète, l'écran n'en montrant qu'un résumé lisible.\n console.error(`[${this.props.area ?? \"dashboard\"}]`, error, info?.componentStack);\n }\n\n render() {\n const { error } = this.state;\n if (!error) return this.props.children;\n return (\n

\n Cette section n'a pas pu s'afficher\n

\n Le reste du dashboard reste utilisable. Changez d'onglet, ou rechargez la page\n pour repartir d'un état propre.\n

Détail technique

\n            {this.props.area ? `[${this.props.area}] ` : \"\"}{String(error?.message ?? error)}\n

\n this.setState({ error: null })}\n className=\"mt-3 px-3 py-1.5 rounded-lg border border-rose-300 text-rose-800 text-xs hover:bg-rose-100\">\n Réessayer\n \n

\n );\n }\n}\n",[],[220,583],"b0051b38-b1d8-452c-9887-896d1f42e612","App.tsx","01a033f6-8ab0-7225-822e-cbf3d3b4919d",["D",1787578190526],"/** @jsxImportSource https://esm.sh/react@18.2.0 */\nimport { useEffect, useMemo, useState } from \"https://esm.sh/react@18.2.0\";\nimport {\n Alert, BRAND_COLORS, downloadCSV, fmtEur, fmtEurK, fmtInt, fmtPct, GROUPS, groupBy,\n Product, productAlerts, aggregate, shippedRevenueBySku, shippedDelivery,\n preCutoverRevenue, preCutoverUnits, shopifyWindowStart, DEFAULT_CUTOVER, displayUser,\n fillDays, isWeekend,\n} from \"/frontend/helpers.ts\";\nimport * as Ico from \"/frontend/Icons.tsx\";\nimport Simulator from \"/frontend/Simulator.tsx\";\nimport Parameters from \"/frontend/Parameters.tsx\";\nimport Forecast from \"/frontend/Forecast.tsx\";\nimport { applyParams } from \"/frontend/pricing.ts\";\nimport { useOrders } from \"/frontend/ordersStore.ts\";\nimport ErrorBoundary from \"/frontend/ErrorBoundary.tsx\";\nimport Amazon from \"/frontend/Amazon.tsx\";\n\nconst NAVY = \"#1F3864\";\nconst ACCENT = \"#4F46E5\";\n\n/** Category-group icons, keyed by GROUPS[].icon */\nconst GROUP_ICONS: Record = {\n fridge: Ico.IconFridge, washer: Ico.IconWasher, oven: Ico.IconOven,\n vacuum: Ico.IconVacuum, coffee: Ico.IconCoffee, tv: Ico.IconTv,\n detergent: Ico.IconDetergent, other: Ico.IconOther,\n};\n\n/** Inline \"icon + label\" for a category group. */\nfunction GroupLabel({ g, size = 16 }: { g: any; size?: number }) {\n const I = GROUP_ICONS[g.icon] ?? Ico.IconOther;\n return (\n \n {g.label}\n \n );\n}\n\n/** Just the icon, from a group key. */\nfunction GroupIcon({ k, size = 15 }: { k: string; size?: number }) {\n const g = GROUPS.find((x) => x.key === k);\n const I = GROUP_ICONS[g?.icon ?? \"other\"] ?? Ico.IconOther;\n return ;\n}\nconst BRANDS = [\"CANDY\", \"HOOVER\", \"HAIER\"];\nconst TABS = [\n { id: \"overview\", label: \"Overview\", Icon: Ico.IconOverview },\n { id: \"products\", label: \"Products\", Icon: Ico.IconProducts },\n { id: \"bestsellers\", label: \"Scorecards\", Icon: Ico.IconScorecard },\n { id: \"business\", label: \"Business\", Icon: Ico.IconBusiness },\n { id: \"logistic\", label: \"Logistic\", Icon: Ico.IconLogistic },\n // Amazon Vendor (1P) — périmètre distinct du D2C, jamais agrégé avec lui.\n // Placé après les onglets D2C pour marquer visuellement la séparation.\n { id: \"amazon\", label: \"Amazon Vendor\", Icon: Ico.IconReceipt },\n { id: \"simulator\", label: \"Pricing Simulator\", Icon: Ico.IconSimulator },\n { id: \"forecast\", label: \"Forecast\", Icon: Ico.IconForecast },\n { id: \"actions\", label: \"Action Plan\", Icon: Ico.IconAction },\n { id: \"changes\", label: \"Changes\", Icon: Ico.IconChanges },\n // editorOnly : masqué aux comptes en lecture seule. Ce n'est qu'un confort d'affichage,\n // la vraie protection étant le refus de PUT /api/params côté serveur.\n { id: \"parameters\", label: \"Parameters\", Icon: Ico.IconScale, editorOnly: true },\n] as const;\n\ntype TabId = typeof TABS[number][\"id\"];\n\nexport default function App() {\n const [data, setData] = useState(null);\n const [error, setError] = useState(null);\n const [tab, setTab] = useState(\"overview\");\n const [changes, setChanges] = useState([]);\n // Tiroir de navigation sur petit écran. Sans effet à partir de lg, où la barre\n // latérale redevient permanente.\n const [navOpen, setNavOpen] = useState(false);\n // Sert uniquement à dater les commandes dans l'en-tête. Le téléchargement étant\n // mutualisé, cet appel ne déclenche aucune requête supplémentaire.\n const ordersMeta = useOrders();\n\n const loadChanges = () => {\n fetch(\"/api/changes\").then((r) => r.ok ? r.json() : { entries: [] })\n .then((j) => setChanges(j.entries ?? [])).catch(() => {});\n };\n\n // Paramètres économiques : ils remplacent les constantes de pricing.ts avant tout\n // calcul du simulateur. `paramsTick` force un rendu après application, sinon React\n // ignorerait la mutation de l'objet PARAMS et le simulateur garderait l'ancienne valeur.\n const [paramsTick, setParamsTick] = useState(0);\n const loadParams = () => {\n fetch(\"/api/params\").then((r) => r.ok ? r.json() : null)\n .then((j) => { if (j?.values) { applyParams(j.values); setParamsTick((n) => n + 1); } })\n .catch(() => {});\n };\n\n useEffect(() => {\n fetch(\"/api/data\").then(async (r) => {\n if (!r.ok) {\n const j = await r.json().catch(() => ({}));\n setError(j.message ?? `Error ${r.status}`);\n return;\n }\n setData(await r.json());\n }).catch((e) => setError(String(e)));\n loadChanges();\n loadParams();\n }, []);\n\n if (error) {\n return (\n
\n
\n
📭
\n
No data available
\n
{error}
\n
\n
\n );\n }\n if (!data) {\n return (\n
\n
Loading dashboard…
\n
\n );\n }\n\n const products: Product[] = data.products;\n // Read-only accounts (direction, observers). The server rejects their writes anyway —\n // this only keeps the edit affordances out of their way.\n const canEdit = data.viewer?.canEdit !== false;\n const pendingByCode = new Set(changes.filter((c) => c.status === \"pending\").map((c) => c.code));\n const nPending = pendingByCode.size;\n\n // Modifications parties du dashboard mais pas encore visibles dans l'export du site.\n //\n // C'est la source d'un vrai malentendu : une équipe change un prix, revient sur Products,\n // et voit toujours l'ancienne valeur — parce que la colonne « Site price » vient de\n // l'export Matrixify, qui ne sera régénéré qu'après le passage de la modification dans\n // Shopify. Elle en conclut que rien n'a été enregistré et recommence.\n //\n // On retient « pending » (en file) et « sent » (livré à Matrixify, pas encore confirmé),\n // mais pas « applied » : à ce stade l'export reflète la nouvelle valeur, il n'y a plus\n // rien à signaler. Les entrées arrivant les plus récentes d'abord, la première rencontrée\n // pour un couple SKU + champ est la bonne.\n const queuedByCode: Record> = {};\n for (const c of changes) {\n if (c.status !== \"pending\" && c.status !== \"sent\") continue;\n const byField = (queuedByCode[c.code] ??= {});\n if (!byField[c.field]) byField[c.field] = c;\n }\n\n const visibleTabs = TABS.filter((t) => canEdit || !(t as any).editorOnly);\n const current = visibleTabs.find((t) => t.id === tab) ?? visibleTabs[0];\n\n return (\n
\n {/* Voile d'arrière-plan du tiroir. Uniquement sous le point de rupture :\n en desktop la barre est permanente et ne doit rien masquer. */}\n {navOpen && (\n
setNavOpen(false)} aria-hidden=\"true\" />\n )}\n\n {/* ── Sidebar ─────────────────────────────────────────────────────────\n Permanente à partir de lg, tiroir coulissant en dessous. On la garde\n montée en permanence plutôt que de la démonter : l'animation reste\n fluide et l'état de défilement de la liste d'onglets est conservé. */}\n \n
\n
D2C
\n
\n
D2C France
\n
Haier · Candy · Hoover
\n
\n {/* Fermeture explicite : le voile suffit à la souris, mais au doigt une\n croix reste le geste attendu et évite les fermetures accidentelles. */}\n \n
\n\n
\n {visibleTabs.map((t) => {\n const active = tab === t.id;\n const Icon = t.Icon;\n return (\n \n );\n })}\n
\n\n
\n
\n {data.viewer?.username ?? \"—\"}\n
\n {!canEdit && (\n
\n Read-only access\n
\n )}\n
\n \n
\n view source\n
\n \n\n {/* ── Main ────────────────────────────────────────────────── */}\n
\n
\n
\n {/* Ouvre le tiroir. Masqué dès lg, où la barre latérale est visible. */}\n \n \n
{current.label}
\n
\n {/* Les dates de fraîcheur sont un repère utile mais secondaire : elles\n cèdent la place au titre sur les écrans étroits. */}\n
\n \n \n Catalogue {data.generated_at ? new Date(data.generated_at).toLocaleDateString(\"fr-FR\") : \"?\"}\n \n {data.cms_synced_at && (\n Site {new Date(data.cms_synced_at).toLocaleString(\"fr-FR\", { day: \"2-digit\", month: \"2-digit\", hour: \"2-digit\", minute: \"2-digit\" })}\n )}\n {/* Troisième source, la plus fraîche : c'est elle qui explique qu'un écran\n bâti sur les commandes soit en avance sur un écran bâti sur le BI. */}\n {ordersMeta?.updated_at && (\n \n Commandes {new Date(ordersMeta.updated_at).toLocaleString(\"fr-FR\", { day: \"2-digit\", month: \"2-digit\", hour: \"2-digit\", minute: \"2-digit\" })}\n \n )}\n
\n
\n\n
\n {/* `key={tab}` remonte une frontière neuve à chaque changement d'onglet : un\n onglet tombé en erreur ne reste pas bloqué sur son message une fois qu'on\n en est parti puis revenu. */}\n \n {tab === \"overview\" && }\n {tab === \"products\" && (\n \n )}\n {tab === \"bestsellers\" && }\n {tab === \"business\" && }\n {tab === \"logistic\" && }\n {tab === \"amazon\" && }\n {tab === \"simulator\" && (\n // `key` : le simulateur calcule à partir de PARAMS, muté sur place. Sans\n // remontage, un changement de paramètre ne serait pas répercuté.\n \n )}\n {tab === \"forecast\" && }\n {tab === \"actions\" && }\n {tab === \"changes\" && }\n {tab === \"parameters\" && canEdit && }\n \n
\n
\n
\n );\n}\n\n// ═══════════════════════════════ OVERVIEW ═══════════════════════════════════\nfunction Overview({ products, cutover }: { products: Product[]; cutover?: string }) {\n const orders = useOrders();\n\n // Year-to-date revenue, composed from the two sources so it is both complete and fresh:\n // • before the Shopify cut-over → BI invoiced NSV (Shopify does not cover the migration)\n // • from the cut-over onwards → Shopify shipped revenue, synced nightly\n // Falls back to the BI full-year figure while the orders feed is loading or absent.\n const cut = cutover ?? DEFAULT_CUTOVER;\n const today = new Date().toISOString().slice(0, 10);\n\n const revBySku = useMemo(() => {\n const lines = orders?.lines;\n if (!Array.isArray(lines) || lines.length === 0) return undefined;\n const shipped = shippedRevenueBySku(lines, products, cut, today);\n const out: Record = {};\n for (const p of products) {\n const k = String(p.code);\n out[k] = (p.revPre ?? 0) + (shipped[k] ?? 0);\n }\n return out;\n }, [orders, products, cut]);\n\n // Headline revenue — same scope as the Business tab: goods (catalogue AND service\n // SKUs) plus the delivery charged. The per-brand / per-category tables below stay\n // product-only, since services and delivery cannot be attributed to a brand.\n const totalRevenue = useMemo(() => {\n const lines = orders?.lines;\n if (!Array.isArray(lines) || lines.length === 0) return undefined;\n const bySku = shippedRevenueBySku(lines, products, cut, today);\n const goods = Object.values(bySku).reduce((s, v) => s + v, 0);\n // Avoirs sans détail article : ils ne portent aucune référence, donc n'apparaissent\n // pas dans `bySku`. Sans cette déduction, Overview affichait 40 k€ de plus que\n // l'onglet Business — un écart sans cause visible pour qui compare les deux écrans.\n const orphans: any[] = (orders as any)?.orphan_refunds ?? [];\n const orphanVal = orphans\n .filter((o) => o.amt != null && o.rf >= cut && o.rf <= today)\n .reduce((s, o) => s + o.amt / 1.2, 0);\n return preCutoverRevenue(products, \"2026-01-01\", cut) + goods\n + shippedDelivery(lines, cut, today) - orphanVal;\n }, [orders, products, cut]);\n\n const tot = aggregate(products, revBySku);\n const byBrand = BRANDS.map((b) => ({ brand: b, ...aggregate(products.filter((p) => p.brand === b), revBySku) }));\n const byGroup = GROUPS.map((g) => ({ ...g, ...aggregate(products.filter((p) => p.grp === g.key), revBySku) }))\n .filter((g) => g.count > 0);\n\n return (\n
\n
\n \n \n \n {/* Trois états distincts, jamais confondus : en cours de chargement on ne montre\n AUCUN chiffre. Auparavant un montant de repli s'affichait pendant les\n secondes de téléchargement — il ressemblait à un résultat définitif et ne se\n distinguait que par un « est. » dans un libellé tronqué à l'écran. C'est\n précisément ce qui faisait croire à un écart avec l'onglet Business. */}\n {orders === undefined\n ? (\n
\n
\n
\n
\n
Revenue YTD — chargement…
\n
\n
\n )\n : (\n \n )}\n \n \n \n
\n\n
\n \n ({ key: b.brand, label: , ...b }))}\n total={tot} />\n \n \n ({ ...g, key: g.key, label: }))} />\n \n
\n
\n );\n}\n\nfunction PerfTable({ rows, total }: { rows: any[]; total?: any }) {\n return (\n
\n \n \n \n \n \n \n \n \n {rows.map((r) => (\n \n \n \n \n \n \n \n \n ))}\n {total && (\n \n \n \n \n \n \n \n \n )}\n \n
Products Sales (u) {total?.revIsInvoiced === false ? \"Revenue est.\" : \"Revenue\"} Forecast (u) Sell-thru
{r.label} {fmtInt(r.count)} {fmtInt(r.sales)} {fmtEurK(r.rev)} {fmtInt(r.fc)}
TOTAL {fmtInt(total.count)} {fmtInt(total.sales)} {fmtEurK(total.rev)} {fmtInt(total.fc)} {fmtPct(total.st)}
\n
\n );\n}\n\n// ═══════════════════════════════ PRODUCTS ═══════════════════════════════════\ninterface ViewDef { name: string; q: string; brand: string | null; grp: string | null; lc: string; ss: string; alertsOnly: boolean }\n\nfunction loadViews(): ViewDef[] {\n try { return JSON.parse(localStorage.getItem(\"d2c_views\") ?? \"[]\"); } catch { return []; }\n}\n\nfunction Products({ products, pendingByCode, queuedByCode = {}, onChanged, canEdit = true }:\n {\n products: Product[]; pendingByCode: Set;\n queuedByCode?: Record>;\n onChanged: () => void; canEdit?: boolean;\n }) {\n const [q, setQ] = useState(\"\");\n const [brand, setBrand] = useState(null);\n const [grp, setGrp] = useState(null);\n const [lc, setLc] = useState(\"\");\n const [ss, setSs] = useState(\"\");\n const [alertsOnly, setAlertsOnly] = useState(false);\n const [sort, setSort] = useState<{ k: keyof Product; dir: 1 | -1 }>({ k: \"sales\", dir: -1 });\n const [sel, setSel] = useState>(new Set());\n const [modal, setModal] = useState(null);\n const openModal = (kind: any, value = \"\") =>\n setModal({ kind, value, rounding: \"none\", when: \"now\", at: \"\" });\n\n const roundPrice = (v: number, rule: string): number => {\n if (rule === \"ceil\") return Math.ceil(v);\n if (rule === \"floor\") return Math.floor(v);\n if (rule === \"psy\") return Math.ceil((v + 0.01) / 10) * 10 - 0.01; // next price ending in 9.99\n return Math.round(v * 100) / 100;\n };\n const [busy, setBusy] = useState(false);\n const [views, setViews] = useState(loadViews());\n const [issueFilter, setIssueFilter] = useState(\"\");\n const [colsOpen, setColsOpen] = useState(false);\n const [hiddenCols, setHiddenCols] = useState>(() => {\n try { return new Set(JSON.parse(localStorage.getItem(\"d2c_hidden_cols\") ?? \"[]\")); } catch { return new Set(); }\n });\n\n const toggleCol = (k: string) => {\n const s = new Set(hiddenCols);\n s.has(k) ? s.delete(k) : s.add(k);\n setHiddenCols(s);\n localStorage.setItem(\"d2c_hidden_cols\", JSON.stringify([...s]));\n };\n const vis = (k: string) => !hiddenCols.has(k);\n\n const COL_DEFS: { k: string; label: string }[] = [\n { k: \"cat\", label: \"Category\" }, { k: \"lc\", label: \"Lifecycle\" }, { k: \"ss\", label: \"Site\" },\n { k: \"s48\", label: \"Stock 48\" }, { k: \"s47p\", label: \"Area 47 stock\" }, { k: \"s47d\", label: \"Area 47 avail. M3\" },\n { k: \"cost\", label: \"Standard cost\" },\n { k: \"pvc\", label: \"RRP\" }, { k: \"be\", label: \"Breakeven\" }, { k: \"mip\", label: \"MIP\" },\n { k: \"price\", label: \"Site price\" }, { k: \"comp\", label: \"Min comp\" }, { k: \"compname\", label: \"Cheapest competitor\" },\n { k: \"gap\", label: \"Gap\" }, { k: \"mcur\", label: \"Margin\" }, { k: \"fc\", label: \"Forecast\" },\n { k: \"sales\", label: \"Sales YTD\" }, { k: \"st\", label: \"Sell-thru\" },\n ];\n\n const lifecycles = useMemo(() => [...new Set(products.map((p) => p.lc))].sort(), [products]);\n const statuses = useMemo(() => [...new Set(products.map((p) => p.ss))].sort(), [products]);\n const allIssues = useMemo(() => {\n const counts: Record = {};\n for (const p of products) for (const a of productAlerts(p)) counts[a.issue] = (counts[a.issue] ?? 0) + 1;\n return Object.entries(counts).sort((a, b) => b[1] - a[1]);\n }, [products]);\n\n const filtered = useMemo(() => {\n const ql = q.trim().toLowerCase();\n let out = products.filter((p) =>\n (!brand || p.brand === brand) &&\n (!grp || p.grp === grp) &&\n (!lc || p.lc === lc) &&\n (!ss || p.ss === ss) &&\n (!alertsOnly || productAlerts(p).length > 0) &&\n (!issueFilter || productAlerts(p).some((a) => a.issue === issueFilter)) &&\n (!ql || String(p.code).includes(ql) || p.name.toLowerCase().includes(ql) ||\n (p.cat ?? \"\").toLowerCase().includes(ql))\n );\n out = [...out].sort((a, b) => {\n const va = a[sort.k] ?? -Infinity, vb = b[sort.k] ?? -Infinity;\n return (va < vb ? -1 : va > vb ? 1 : 0) * sort.dir;\n });\n return out;\n }, [products, q, brand, grp, lc, ss, alertsOnly, issueFilter, sort]);\n\n const toggleSel = (code: number) => {\n const s = new Set(sel);\n s.has(code) ? s.delete(code) : s.add(code);\n setSel(s);\n };\n const selectAllFiltered = () => setSel(new Set(filtered.map((p) => p.code)));\n\n const applyView = (v: ViewDef) => {\n setQ(v.q); setBrand(v.brand); setGrp(v.grp); setLc(v.lc); setSs(v.ss); setAlertsOnly(v.alertsOnly);\n };\n const saveCurrentView = () => {\n const name = prompt(\"Nom de la vue :\");\n if (!name) return;\n const next = [...views.filter((v) => v.name !== name), { name, q, brand, grp, lc, ss, alertsOnly }];\n setViews(next); localStorage.setItem(\"d2c_views\", JSON.stringify(next));\n };\n const deleteView = (name: string) => {\n const next = views.filter((v) => v.name !== name);\n setViews(next); localStorage.setItem(\"d2c_views\", JSON.stringify(next));\n };\n\n const submitChanges = async () => {\n if (!modal) return;\n const selProducts = products.filter((p) => sel.has(p.code));\n const applyAt = modal.when === \"later\" && modal.at\n ? new Date(modal.at).toISOString() : null;\n const changes: any[] = [];\n let skipped = 0;\n for (const p of selProducts) {\n if (modal.kind === \"status\") {\n changes.push({ code: p.code, field: \"status\", newValue: modal.value, oldValue: p.ss, applyAt });\n continue;\n }\n let np: number | null = null;\n if (modal.kind === \"price\") np = Number(modal.value);\n else if (modal.kind === \"pct\") np = p.price ? p.price * (1 + Number(modal.value) / 100) : null;\n else if (modal.kind === \"align_be\") np = p.be ?? null;\n else if (modal.kind === \"align_mip\") np = p.mip ?? null;\n else if (modal.kind === \"align_rrp\") np = p.pvc ?? null;\n else if (modal.kind === \"align_comp\") np = (p.comp && p.comp > 0) ? p.comp : null;\n else if (modal.kind === \"margin\") {\n const m = Number(modal.value) / 100;\n np = (p.cost != null && m < 1) ? ((p.cost + (p.deee ?? 0)) / (1 - m)) * 1.2 : null;\n }\n if (np == null || !isFinite(np) || np <= 0) { skipped++; continue; }\n if ([\"align_comp\", \"margin\", \"pct\"].includes(modal.kind)) np = roundPrice(np, modal.rounding);\n else np = Math.round(np * 100) / 100;\n changes.push({ code: p.code, field: \"price\", newValue: np, oldValue: p.price ?? null, applyAt });\n }\n if (changes.length === 0) { alert(`No applicable product (missing reference price — ${skipped} skipped)`); return; }\n if (skipped > 0 && !confirm(`${skipped} product(s) skipped (missing reference). Continue with ${changes.length}?`)) return;\n setBusy(true);\n const r = await fetch(\"/api/changes\", {\n method: \"POST\", headers: { \"Content-Type\": \"application/json\" },\n body: JSON.stringify({ changes }),\n });\n const j = await r.json();\n setBusy(false); setModal(null); setSel(new Set());\n onChanged();\n alert(`${j.added} change(s) queued.` +\n (j.errors?.length ? `\\n\\nErrors:\\n${j.errors.join(\"\\n\")}` : \"\"));\n };\n\n // sticky top-0 keeps the header visible while the table body scrolls in its own box\n const Th = ({ k, children, right = true }: any) => (\n setSort((s) => ({ k, dir: s.k === k ? (s.dir === 1 ? -1 : 1) : -1 }))}>\n {children}{sort.k === k ? (sort.dir === -1 ? \" ↓\" : \" ↑\") : \"\"}\n \n );\n\n return (\n
\n \n {views.length > 0 && (\n
\n Vues :\n {views.map((v) => (\n \n \n \n \n ))}\n
\n )}\n
\n {/* Pleine largeur sur mobile : à 375 px une recherche de 16 rem laissait\n deux sélecteurs orphelins sur la ligne suivante. */}\n setQ(e.target.value)} placeholder=\"Search SKU / name / category…\"\n className=\"px-3 py-2 border rounded-lg text-sm w-full sm:w-64\" />\n \n \n \n \n setAlertsOnly(e.target.checked)} />\n With alerts only\n \n \n
\n \n {colsOpen && (\n /* Ancré à droite et borné à la largeur d'écran : ouvert depuis un bouton\n proche du bord, le panneau sortait de la page sur mobile. */\n
\n {COL_DEFS.map((c) => (\n \n toggleCol(c.k)} />\n {c.label}\n \n ))}\n
\n )}\n
\n \n
\n
\n setBrand(null)}>All brands\n {BRANDS.map((b) => setBrand(brand === b ? null : b)}>{b})}\n |\n setGrp(null)}>All categories\n {GROUPS.filter((g) => g.key !== \"Other\").map((g) =>\n setGrp(grp === g.key ? null : g.key)}>\n )}\n
\n \n\n {/* Bulk action bar — editors only */}\n {canEdit && sel.size > 0 && (\n
\n {sel.size} selected\n \n \n |\n Align to:\n \n \n \n \n \n |\n \n \n \n
\n )}\n\n \n {/* The table scrolls inside its own box so the header can stay pinned to the top.\n `tscroll` y ajoute le défilement horizontal : avec 18 colonnes il n'y a aucune\n largeur d'écran, même en desktop, où tout tient sans compression illisible. */}\n
\n \n \n {canEdit && (\n \n )}\n \n \n {vis(\"cat\") && }\n {vis(\"lc\") && }\n {vis(\"ss\") && }\n {vis(\"s48\") && }\n {vis(\"s47p\") && }\n {vis(\"s47d\") && }\n {vis(\"cost\") && }\n {vis(\"pvc\") && }\n {vis(\"be\") && }\n {vis(\"mip\") && }\n {vis(\"price\") && }\n {vis(\"comp\") && }\n {vis(\"compname\") && }\n {vis(\"gap\") && }\n {vis(\"mcur\") && }\n {vis(\"fc\") && }\n {vis(\"sales\") && }\n {vis(\"st\") && }\n \n \n {filtered.slice(0, 400).map((p) => {\n const queued = queuedByCode[p.code];\n // `bg-white` explicite (et non l'absence de fond) : la cellule SKU est\n // figée à gauche et hérite du fond de sa ligne. Sans couleur déclarée\n // elle serait transparente et laisserait défiler le texte dessous.\n return (\n \n {canEdit && (\n \n )}\n \n \n {vis(\"cat\") && (\n \n )}\n {vis(\"lc\") && }\n {vis(\"ss\") && }\n {vis(\"s48\") && }\n {vis(\"s47p\") && }\n {vis(\"s47d\") && }\n {vis(\"cost\") && }\n {vis(\"pvc\") && }\n {vis(\"be\") && }\n {vis(\"mip\") && (\n \n )}\n {vis(\"price\") && (\n \n )}\n {vis(\"comp\") && }\n {vis(\"compname\") && }\n {vis(\"gap\") && }\n {vis(\"mcur\") && }\n {vis(\"fc\") && }\n {vis(\"sales\") && }\n {vis(\"st\") && }\n \n );\n })}\n \n
0 && sel.size >= filtered.length}\n onChange={(e) => e.target.checked ? selectAllFiltered() : setSel(new Set())} /> SKU Product Category Lifecycle Site Stock 48 Area 47 stock Area 47 avail. M3 Standard cost RRP Breakeven MIP Site price Min comp Cheapest comp. Gap Margin Forecast Sales YTD Sell-thru
toggleSel(p.code)} /> {p.code}{queued ? \" ⏳\" : \"\"} \n {p.name}\n \n \n {p.cat ?? \"—\"}\n \n {p.lc} {fmtInt(p.s48)} {fmtInt(p.s47p)} {fmtInt(p.s47d)} {fmtEur(p.cost)} {fmtEur(p.pvc)} {fmtEur(p.be)} \n {fmtEur(p.mip)}\n \n \n {fmtEur(p.comp)} {p.compname ?? \"—\"} {fmtPct(p.mcur)} {fmtInt(p.fc)} {fmtInt(p.sales)}
\n {filtered.length > 400 && (\n
Showing first 400 rows — refine filters or export CSV.
\n )}\n
\n \n\n {/* Edit modal */}\n {modal && (\n /* La modale doit rester atteignable au clavier virtuel : largeur bornée à\n l'écran, hauteur bornée à la fenêtre avec défilement interne. Sans quoi les\n boutons Valider/Annuler passaient sous le clavier sur téléphone. */\n
!busy && setModal(null)}>\n
e.stopPropagation()}>\n
\n {modal.kind === \"price\" ? \"💶 Set price (€ incl. VAT)\" :\n modal.kind === \"pct\" ? \"％ Adjust price (%)\" :\n modal.kind === \"align_be\" ? \"⚖️ Align to Breakeven\" :\n modal.kind === \"align_mip\" ? \"⚖️ Align to Min Internal Price\" :\n modal.kind === \"align_rrp\" ? \"⚖️ Align to RRP\" :\n modal.kind === \"align_comp\" ? \"⚔️ Align to cheapest competitor\" :\n modal.kind === \"margin\" ? \"🎯 Set price for target margin\" :\n modal.value === \"Active\" ? \"▶️ Activate listings\" : \"⏸️ Deactivate listings\"}\n
\n
{sel.size} product(s) selected. Changes go to the queue and reach Shopify at the next hourly Matrixify import.
\n {[\"price\", \"pct\", \"margin\"].includes(modal.kind) && (\n setModal({ ...modal, value: e.target.value })}\n placeholder={modal.kind === \"price\" ? \"ex: 499.99\" :\n modal.kind === \"margin\" ? \"marge cible en % — ex: 25\" : \"ex: -5 (baisse de 5%)\"}\n className=\"w-full px-3 py-2 border rounded-lg mb-3\" />\n )}\n {[\"align_comp\", \"margin\", \"pct\"].includes(modal.kind) && (\n
\n Rounding rule\n \n
\n )}\n
\n Apply\n
\n \n setModal({ ...modal, when: \"now\" })} /> Now\n \n \n setModal({ ...modal, when: \"later\" })} /> Schedule (FR time)\n \n {modal.when === \"later\" && (\n setModal({ ...modal, at: e.target.value })}\n className=\"px-2 py-1.5 border rounded-lg text-sm\" />\n )}\n
\n
\n
\n \n \n
\n
\n
\n )}\n
\n );\n}\n\nfunction exportProducts(list: Product[]) {\n downloadCSV(\"products_export.csv\",\n [\"SKU\", \"EAN\", \"Name\", \"Brand\", \"Category\", \"Group\", \"Lifecycle\", \"Site status\", \"Stock48\", \"Stock47\", \"Standard cost\", \"Eco-part. DEEE\", \"Site price\", \"Breakeven\", \"Min internal\", \"RSP\", \"Min competitor\", \"Cheapest competitor\", \"Gap %\", \"Margin %\", \"Forecast\", \"Sales YTD\", \"Sales/wk\", \"Sell-through %\", \"Revenue est.\", \"Catalogue action\"],\n list.map((p) => [p.code, p.ean ?? \"\", p.name, p.brand, p.cat ?? \"\", p.grp, p.lc, p.ss, p.s48, p.s47p,\n p.cost ?? \"\", p.deee ?? \"\",\n p.price ?? \"\", p.be ?? \"\", p.mip ?? \"\", p.rsp ?? \"\", p.comp ?? \"\", p.compname ?? \"\",\n p.gap != null ? (p.gap * 100).toFixed(1) : \"\", p.mcur != null ? (p.mcur * 100).toFixed(1) : \"\",\n p.fc, p.sales, p.spw ?? \"\", p.st != null ? (p.st * 100).toFixed(1) : \"\",\n p.price ? Math.round(p.price * p.sales) : \"\", p.cact]));\n}\n\n// ═══════════════════════════ SCORECARDS (BEST SELLERS) ══════════════════════\nfunction BestSellers({ products, cutover }: { products: Product[]; cutover?: string }) {\n const [scopeType, setScopeType] = useState<\"all\" | \"brand\" | \"group\">(\"all\");\n const [scopeVal, setScopeVal] = useState(\"\");\n const [q, setQ] = useState(\"\");\n const [rankBy, setRankBy] = useState<\"units\" | \"revenue\">(\"units\");\n // Profondeur du classement. « all » liste toutes les références ayant vendu sur la\n // période, ce qui transforme l'onglet en inventaire des produits vendus.\n const [topN, setTopN] = useState<20 | 50 | 100 | \"all\">(20);\n const orders = useOrders(); // undefined=loading, null=absent\n const [from, setFrom] = useState(() => new Date().getFullYear() + \"-01-01\");\n const [to, setTo] = useState(() => new Date().toISOString().slice(0, 10));\n\n // La recherche agit comme les autres filtres de périmètre : elle restreint tout le\n // scorecard, KPI compris, et pas seulement le classement. Une carte dont le Top 20\n // serait filtré mais dont les totaux porteraient sur le catalogue entier donnerait des\n // parts de marché fausses — et cette carte est faite pour être imprimée et diffusée.\n const scoped = useMemo(() => {\n const ql = q.trim().toLowerCase();\n let out = products;\n if (scopeType === \"brand\" && scopeVal) out = out.filter((p) => p.brand === scopeVal);\n else if (scopeType === \"group\" && scopeVal) out = out.filter((p) => p.grp === scopeVal);\n if (ql) {\n out = out.filter((p) =>\n String(p.code).includes(ql) ||\n p.name.toLowerCase().includes(ql) ||\n (p.cat ?? \"\").toLowerCase().includes(ql));\n }\n return out;\n }, [products, scopeType, scopeVal, q]);\n\n const hasOrders = Array.isArray(orders?.lines) && orders.lines.length > 0;\n\n // ── Period figures from the orders feed ────────────────────────────────────\n // Shipped lines dated on their FULFILMENT date, net of discounts, minus refunds\n // booked on their own refund date — the same accounting basis as the BI.\n // Quand la part BI est ajoutée (période remontant avant la bascule), Shopify doit\n // démarrer à la bascule — sinon janvier–février compterait deux fois.\n const biIncluded = from < (cutover ?? DEFAULT_CUTOVER);\n const shopFrom = shopifyWindowStart(from, cutover ?? DEFAULT_CUTOVER, biIncluded);\n\n const period = useMemo(() => {\n if (!hasOrders) return null;\n const byCode: Record = {};\n for (const p of scoped) byCode[String(p.code)] = p;\n const units: Record = {};\n const rev: Record = {};\n let refundVal = 0, refundUnits = 0, hasRef = false, deeeTotal = 0;\n // Marge RÉELLEMENT réalisée, ligne par ligne, au prix effectivement pratiqué —\n // même méthode que l'onglet Business. La version précédente appliquait le taux de\n // marge du prix du jour pondéré par le CA de la période : elle ignorait remises et\n // tarifs marketplace et surestimait la marge de 5 points, jusqu'à afficher +53 % sur\n // une référence vendue à perte.\n let mSum = 0, mBase = 0;\n // Même calcul par référence : le KPI et la ligne produit doivent dire la même chose.\n const mBySku: Record = {};\n // Revenue excl. VAT AND excl. eco-participation — the BI's NSV definition.\n const goodsOf = (l: any, p: Product) =>\n (l.n != null ? l.n : l.p * l.q) / 1.2 - (p.deee ?? 0) * l.q;\n for (const l of orders.lines) {\n const kind = l.k ?? \"ord\";\n if (kind === \"ord\") continue;\n const d = kind === \"ref\" ? l.rf : (l.f || l.d);\n if (!d || d < shopFrom || d > to) continue;\n const sku = l.sku ?? \"\";\n const p = byCode[sku];\n if (!p) continue;\n const sign = kind === \"ref\" ? -1 : 1;\n const amt = goodsOf(l, p);\n units[sku] = (units[sku] ?? 0) + sign * l.q;\n rev[sku] = (rev[sku] ?? 0) + sign * amt;\n deeeTotal += sign * (p.deee ?? 0) * l.q;\n if (p.cost && l.p > 0) {\n mSum += sign * (amt - p.cost * l.q);\n mBase += sign * amt;\n const e = (mBySku[sku] ??= { gm: 0, base: 0 });\n e.gm += sign * (amt - p.cost * l.q);\n e.base += sign * amt;\n }\n if (kind === \"ref\") { hasRef = true; refundVal += amt; refundUnits += l.q; }\n }\n const totUnits = Object.values(units).reduce((s, v) => s + v, 0);\n const totRev = Object.values(rev).reduce((s, v) => s + v, 0);\n return {\n units, rev, totUnits, totRev, refundVal, refundUnits, hasRef, deeeTotal, mBySku,\n margin: mBase > 0 ? mSum / mBase : null,\n marginBase: mBase,\n };\n }, [hasOrders, orders, scoped, shopFrom, to]);\n\n /** Units and revenue for a product: real period figures, or catalogue YTD as fallback. */\n // Même définition partout : BI avant la bascule, Shopify après — pour le CA comme\n // pour les unités, faute de quoi le prix de vente moyen serait faux.\n const preOf = (p: Product) => (biIncluded ? (p.revPre ?? 0) : 0);\n const preUnitsOf = (p: Product) => (biIncluded ? (p.salesPre ?? 0) : 0);\n const unitsOf = (p: Product) =>\n period ? preUnitsOf(p) + (period.units[String(p.code)] ?? 0) : p.sales;\n const revOf = (p: Product) =>\n period ? preOf(p) + (period.rev[String(p.code)] ?? 0) : (p.rev ?? (p.price ?? 0) * p.sales);\n /** Marge réalisée de la référence sur la période, ou null si rien à mesurer.\n * Même source que le KPI : afficher deux définitions dans un même écran était le\n * défaut le plus visible du Scorecard. */\n const marginOf = (p: Product): number | null => {\n const e = period?.mBySku?.[String(p.code)];\n return e && e.base > 0 ? e.gm / e.base : null;\n };\n /** Prix de vente moyen réellement pratiqué : CA de la période ÷ unités de la période.\n * Mêmes numérateur et dénominateur que les colonnes Revenue et Units de la ligne, donc\n * toujours cohérent avec elles, et construit sur les deux sources (BI avant la bascule,\n * Shopify après) puisque revOf et unitsOf les additionnent l'une et l'autre.\n *\n * Attention à la base : HT et hors éco-participation, net de remises, retours déduits —\n * c'est la définition NSV du BI. Le prix catalogue `p.price` est TTC : les deux ne sont\n * pas comparables directement, d'où le rappel du prix du jour dans l'infobulle.\n *\n * Null dès que les unités nettes ne sont pas strictement positives : une référence dont\n * les retours annulent les ventes donnerait un prix moyen infini ou négatif. */\n const aspOf = (p: Product): number | null => {\n const u = unitsOf(p);\n if (!u || u <= 0) return null;\n const r = revOf(p);\n if (!isFinite(r) || r <= 0) return null;\n return r / u;\n };\n\n const top = useMemo(() => {\n const arr = scoped.filter((p) => unitsOf(p) !== 0 || !period);\n arr.sort((a, b) => rankBy === \"units\" ? unitsOf(b) - unitsOf(a) : revOf(b) - revOf(a));\n return topN === \"all\" ? arr : arr.slice(0, topN);\n }, [scoped, rankBy, period, topN]);\n\n const cat = aggregate(scoped);\n // Marge de la période : celle qui a réellement été encaissée. Sans flux de commandes,\n // repli sur la marge catalogue au prix du jour — c'est alors la seule disponible, et\n // le libellé du KPI le dit.\n const periodMargin = period?.margin ?? cat.avgMargin;\n\n // Les totaux affichés doivent être la somme des lignes du tableau — donc passer par\n // unitsOf / revOf, qui portent déjà la part BI. Utiliser period.totRev directement\n // afficherait la seule part Shopify (le bug des 3,29 M€ au lieu de 4,19).\n const agg = {\n ...cat,\n sales: period ? scoped.reduce((s, p) => s + unitsOf(p), 0) : cat.sales,\n rev: period ? scoped.reduce((s, p) => s + revOf(p), 0) : cat.rev,\n avgMargin: periodMargin,\n };\n const scopeGroup = scopeType === \"group\" ? GROUPS.find((g) => g.key === scopeVal) : null;\n // Le libellé sert d'en-tête à la carte imprimée et de nom de fichier à l'export :\n // il doit dire ce que la carte contient réellement, recherche comprise.\n const baseLabel = scopeType === \"all\" ? \"All catalogue\" :\n scopeType === \"brand\" ? scopeVal : scopeGroup?.label ?? scopeVal;\n const scopeLabel = q.trim() ? `${baseLabel} · « ${q.trim()} »` : baseLabel;\n\n // Bar length is scaled to the best seller (readable), the figure is the true\n // share of the selected scope's total revenue.\n const maxRev = Math.max(...top.map((p) => revOf(p)), 1);\n const scopeRev = scoped.reduce((s, p) => s + revOf(p), 0);\n const topShare = scopeRev > 0\n ? top.reduce((s, p) => s + revOf(p), 0) / scopeRev : 0;\n\n return (\n
\n \n
\n From\n setFrom(e.target.value)}\n className=\"px-2 py-1.5 border rounded-lg text-sm\" disabled={!hasOrders} />\n to\n setTo(e.target.value)}\n className=\"px-2 py-1.5 border rounded-lg text-sm\" disabled={!hasOrders} />\n \n
\n {!period && (\n
\n Orders feed unavailable — showing catalogue Sales YTD. The date filter is inactive.\n
\n )}\n {/* Ligne 2 — ce que la carte contient : recherche libre, puis marque et catégorie */}\n
\n setQ(e.target.value)}\n placeholder=\"Search SKU / name / category…\"\n className=\"px-3 py-2 border border-slate-200 rounded-lg text-sm w-full sm:w-72\" />\n {q && (\n \n )}\n |\n { setScopeType(\"all\"); setScopeVal(\"\"); }}>Whole catalogue\n |\n {BRANDS.map((b) => (\n { setScopeType(\"brand\"); setScopeVal(b); }}>{b}\n ))}\n |\n {GROUPS.filter((g) => g.key !== \"Other\").map((g) => (\n { setScopeType(\"group\"); setScopeVal(g.key); }}>\n \n ))}\n
\n\n {/* Ligne 3 — comment la carte est présentée : profondeur à gauche, tri et exports à droite */}\n
\n {([20, 50, 100, \"all\"] as const).map((n) => (\n setTopN(n)}>\n {n === \"all\" ? \"All sold\" : `Top ${n}`}\n \n ))}\n
\n setRankBy(\"units\")}>Rank by units\n setRankBy(\"revenue\")}>Rank by revenue\n \n \n
\n
\n \n\n {/* Printable scorecard */}\n
\n
\n
\n
\n {scopeGroup && }\n Commercial Scorecard — {scopeLabel}\n
\n
\n D2C France · Haier Candy Hoover ·{\" \"}\n {period\n ? <>Shipped {new Date(from).toLocaleDateString(\"fr-FR\")} → {new Date(to).toLocaleDateString(\"fr-FR\")}\n : <>Catalogue sales YTD}\n {\" \"}· edited {new Date().toLocaleDateString(\"fr-FR\")}\n
\n
\n \n
\n\n
\n \n \n \n \n \n \n
\n\n
\n
\n {topN === \"all\" ? \"Products sold\" : `Top ${topN}`}\n {\" \"}— ranked by {rankBy === \"units\" ? \"units sold\" : \"revenue\"}\n
\n \n {topN === \"all\"\n ? <>{fmtInt(top.length)} products with sales over the period\n : <>These {Math.min(topN as number, top.length)} products = {fmtPct(topShare)} of {scopeLabel} revenue}\n \n
\n
\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n {top.map((p, i) => {\n const rev = revOf(p);\n return (\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n );\n })}\n \n
# SKU Product Category {period ? \"Units shipped\" : \"Sales (u)\"} {period ? \"Revenue\" : \"Revenue est.\"} \n Share of revenue Sell-thru Margin \n Avg selling price Stock 48 Position
{i + 1} {p.code} {p.name} {p.cat ?? \"—\"} {fmtInt(unitsOf(p))} {fmtEurK(rev)} \n
\n
\n
\n
\n \n {scopeRev > 0 ? ((rev / scopeRev) * 100).toFixed(1) + \"%\" : \"—\"}\n \n
\n
\n {(() => {\n const m = marginOf(p);\n if (m == null) {\n // Rien de vendu sur la période, ou coût standard absent : on\n // affiche la marge catalogue en la signalant comme telle plutôt\n // que de laisser croire à une marge constatée.\n return (\n \n {fmtPct(p.mcur)}*\n \n );\n }\n return (\n \n {fmtPct(m)}\n \n );\n })()}\n \n {(() => {\n const asp = aspOf(p);\n if (asp == null) {\n return (\n —\n );\n }\n // Le prix du jour est rappelé en infobulle, converti HT pour être\n // sur la même base que le prix moyen : comparer un TTC à un HT\n // ferait conclure à une remise de 20 % là où il n'y en a aucune.\n const refHt = p.price != null && p.price > 0 ? p.price / (1 + 0.20) : null;\n const gap = refHt ? (asp - refHt) / refHt : null;\n return (\n \n {fmtEur(asp)}\n \n );\n })()}\n {fmtInt(p.s48)}
\n
\n
\n
\n );\n}\n\n/** SVG line chart of daily sales deltas computed from cumulative snapshots. */\nfunction TrendChart({ series }: { series: { date: string; sales: number }[] }) {\n const deltas = series.slice(1).map((s, i) => ({\n date: s.date,\n v: Math.max(0, s.sales - series[i].sales),\n }));\n if (deltas.length < 2) return
Collecting history — chart appears after a few days of snapshots.
;\n const W = 800, H = 120, PAD = 4;\n const max = Math.max(...deltas.map((d) => d.v), 1);\n const pts = deltas.map((d, i) => {\n const x = PAD + (i / (deltas.length - 1)) * (W - 2 * PAD);\n const y = H - PAD - (d.v / max) * (H - 2 * PAD);\n return `${x},${y}`;\n });\n return (\n \n );\n}\n\n/** Export du Scorecard — les colonnes doivent dire exactement ce que dit l'écran.\n * `marginOf` et `aspOf` sont passés par l'appelant plutôt que recalculés ici : ce sont\n * des grandeurs de période, elles dépendent des dates et du périmètre sélectionnés, que\n * cette fonction ne connaît pas. Le CSV écrivait auparavant `p.mcur` (marge catalogue au\n * prix du jour) et `p.price` (prix site TTC) alors que le tableau affichait la marge\n * réalisée et le prix de vente moyen : le fichier contredisait la capture d'écran. */\nfunction exportBestSellers(\n top: Product[], scope: string,\n unitsOf: (p: Product) => number, revOf: (p: Product) => number, dated: boolean,\n marginOf: (p: Product) => number | null, aspOf: (p: Product) => number | null,\n) {\n downloadCSV(`scorecard_${scope.replace(/\\W+/g, \"_\")}.csv`,\n [\"Rank\", \"SKU\", \"Name\", \"Brand\", \"Category\",\n dated ? \"Units shipped (period)\" : \"Sales YTD\",\n dated ? \"Revenue excl. VAT (period)\" : \"Revenue est.\",\n \"Sales/wk\", \"Forecast\", \"Sell-through %\",\n \"Realised margin %\", \"Catalogue margin % (today's price)\",\n \"Avg selling price excl. VAT\", \"Site price incl. VAT\",\n \"Stock48\", \"Price position\"],\n top.map((p, i) => {\n const m = marginOf(p);\n const asp = aspOf(p);\n return [i + 1, p.code, p.name, p.brand, p.cat ?? \"\", unitsOf(p),\n Math.round(revOf(p)), p.spw ?? \"\", p.fc,\n p.st != null ? (p.st * 100).toFixed(1) : \"\",\n m != null ? (m * 100).toFixed(1) : \"\",\n p.mcur != null ? (p.mcur * 100).toFixed(1) : \"\",\n asp != null ? asp.toFixed(2) : \"\",\n p.price ?? \"\", p.s48, p.pos ?? \"\"];\n }));\n}\n\n// ═══════════════════════════ ACTION PLAN ════════════════════════════════════\nfunction ActionPlan({ products }: { products: Product[] }) {\n const [typeFilter, setTypeFilter] = useState<\"all\" | \"commercial\" | \"technical\">(\"all\");\n\n const rows = useMemo(() => {\n const out: { p: Product; a: Alert }[] = [];\n for (const p of products) for (const a of productAlerts(p)) out.push({ p, a });\n return out.sort((x, y) => x.a.severity - y.a.severity || y.p.sales - x.p.sales);\n }, [products]);\n\n const filtered = typeFilter === \"all\" ? rows : rows.filter((r) => r.a.type === typeFilter);\n const byIssue = groupBy(filtered, (r) => r.a.issue);\n const nCom = rows.filter((r) => r.a.type === \"commercial\").length;\n const nTec = rows.filter((r) => r.a.type === \"technical\").length;\n\n return (\n
\n
\n \n \n \n
\n\n
\n \n
\n\n {Object.entries(byIssue).map(([issue, items]) => (\n 1 ? \"s\" : \"\"}`}>\n
\n \n \n \n \n \n \n \n \n \n \n \n {items.slice(0, 100).map(({ p, a }, i) => (\n \n \n \n \n \n \n \n \n \n ))}\n \n
SKU Product Lifecycle Stock 48 Site price Sales YTD Detail
{p.code} {p.name} {p.lc} {fmtInt(p.s48)} {fmtEur(p.price)} {fmtInt(p.sales)} {a.detail ?? \"\"}
\n
\n \n ))}\n {filtered.length === 0 && (\n
No open actions for this filter.
\n )}\n
\n );\n}\n\nfunction exportActions(rows: { p: Product; a: Alert }[]) {\n downloadCSV(\"action_plan.csv\",\n [\"Type\", \"Issue\", \"SKU\", \"Name\", \"Brand\", \"Category\", \"Lifecycle\", \"Stock48\", \"Site price\", \"Sales YTD\", \"Detail\"],\n rows.map(({ p, a }) => [a.type, a.issue, p.code, p.name, p.brand, p.cat ?? \"\", p.lc, p.s48, p.price ?? \"\", p.sales, a.detail ?? \"\"]));\n}\n\n// ═══════════════════════════ CHANGES (audit + queue) ═════════════════════════\nfunction ChangesTab({ entries, refresh, canEdit = true }:\n { entries: any[]; refresh: () => void; canEdit?: boolean }) {\n const pending = entries.filter((e) => e.status === \"pending\");\n const others = entries.filter((e) => e.status !== \"pending\").slice(0, 100);\n\n const cancel = async (id: string) => {\n await fetch(\"/api/changes/cancel\", {\n method: \"POST\", headers: { \"Content-Type\": \"application/json\" },\n body: JSON.stringify({ id }),\n });\n refresh();\n };\n\n const fmtVal = (e: any, v: any) => e.field === \"price\" ? fmtEur(Number(v)) : String(v ?? \"—\");\n\n const Row = ({ e, showCancel }: any) => (\n \n {new Date(e.at).toLocaleString(\"fr-FR\")}\n {displayUser(e.user)}\n {e.code}\n {e.name ?? \"—\"}\n {e.brand}\n {e.field === \"price\" ? \"💶 Price\" : \"📄 Status\"}\n {fmtVal(e, e.oldValue)}\n {fmtVal(e, e.newValue)}\n \n {e.status === \"pending\" && (\n e.applyAt && new Date(e.applyAt) > new Date()\n ? 🕐 scheduled {new Date(e.applyAt).toLocaleString(\"fr-FR\")}\n : ⏳ pending\n )}\n {e.status === \"sent\" && 📤 sent {e.sentAt ? new Date(e.sentAt).toLocaleString(\"fr-FR\") : \"\"}}\n {e.status === \"applied\" && ✅ live {e.appliedAt ? new Date(e.appliedAt).toLocaleString(\"fr-FR\") : \"\"}}\n {e.status === \"cancelled\" && ✖ cancelled}\n \n {showCancel && canEdit && (\n \n )}\n \n );\n\n const Tbl = ({ rows, showCancel }: any) => (\n
\n \n \n \n \n \n \n \n \n {rows.map((e: any) => )}\n
Date User SKU Product Brand Field Before After Status
\n
\n );\n\n return (\n
\n \n {pending.length === 0\n ?
No pending changes. Select products in the Products tab to queue price or status updates.
\n : }\n
\n ⏳ pending → 📤 sent (picked up by Matrixify) → ✅ live (confirmed on the site by the CMS sync).\n Sent changes stay in the feed for 12 h, so a Matrixify retry re-applies them safely.\n A change stuck on \"sent\" means Matrixify fetched it but the import did not apply it — check the Matrixify job history.\n
\n \n \n {others.length === 0\n ?
No history yet.
\n : }\n \n
\n );\n}\n\n// ═══════════════════════════ LOGISTIC ═══════════════════════════════════════\nfunction LogisticTab({ products }: { products: Product[] }) {\n const [brand, setBrand] = useState(null);\n const [grp, setGrp] = useState(null);\n const [from, setFrom] = useState(() => new Date().getFullYear() + \"-01-01\");\n const [to, setTo] = useState(() => new Date().toISOString().slice(0, 10));\n const orders = useOrders(); // undefined=loading, null=absent\n\n const skuInfo = useMemo(() => {\n const m: Record = {};\n for (const p of products) m[String(p.code)] = p;\n return m;\n }, [products]);\n\n const hasOrders = orders !== undefined && orders !== null && Array.isArray(orders?.lines) && orders.lines.length > 0;\n\n // Units per SKU on the selected period:\n // - with orders feed: real units shipped in [from, to]\n // - fallback: catalog Sales YTD (date filter inactive)\n const unitsBySku = useMemo(() => {\n const m: Record = {};\n if (hasOrders) {\n for (const l of orders.lines) {\n if ((l.k ?? \"ord\") !== \"ful\") continue; // logistics cost applies to shipped units\n if (!l.sku || l.d < from || l.d > to) continue;\n const p = skuInfo[l.sku];\n if (!p || p.logi == null) continue;\n if (brand && p.brand !== brand) continue;\n if (grp && p.grp !== grp) continue;\n m[l.sku] = (m[l.sku] ?? 0) + l.q;\n }\n } else {\n for (const p of products) {\n if (p.logi == null || !p.sales) continue;\n if (brand && p.brand !== brand) continue;\n if (grp && p.grp !== grp) continue;\n m[String(p.code)] = p.sales;\n }\n }\n return m;\n }, [hasOrders, orders, skuInfo, products, brand, grp, from, to]);\n\n const rows = useMemo(() => {\n const out = Object.entries(unitsBySku)\n .map(([sku, units]) => ({ p: skuInfo[sku], units }))\n .filter((r) => r.p);\n out.sort((a, b) => ((b.p.logi ?? 0) * b.units) - ((a.p.logi ?? 0) * a.units));\n return out;\n }, [unitsBySku, skuInfo]);\n\n const totUnits = rows.reduce((s, r) => s + r.units, 0);\n const totSpend = rows.reduce((s, r) => s + (r.p.logi ?? 0) * r.units, 0);\n const avgCost = totUnits > 0 ? totSpend / totUnits : null; // true weighted avg per shipped unit\n const nDpd = rows.filter((r) => r.p.carrier === \"DPD\").length;\n const nCcv = rows.filter((r) => r.p.carrier === \"CCV\").length;\n\n // Shipping stats from orders feed (if configured)\n const shipStats = useMemo(() => {\n const lines = orders?.lines;\n if (!Array.isArray(lines) || lines.length === 0) return null;\n const skuBrand: Record = {};\n for (const p of products) skuBrand[String(p.code)] = p.brand;\n const orderMap: Record = {};\n for (const l of lines) {\n if (l.d < from || l.d > to) continue;\n if (brand && l.b !== brand) continue;\n if (grp) {\n const p = products.find((x) => String(x.code) === l.sku);\n if (!p || p.grp !== grp) continue;\n }\n const e = orderMap[l.o] ?? { d: l.d, f: null, ful: false };\n if ((l.k ?? \"ord\") === \"ful\") { e.ful = true; if (l.f && !e.f) e.f = l.f; }\n orderMap[l.o] = e;\n }\n const os = Object.values(orderMap);\n if (os.length === 0) return null;\n const shipped = os.filter((o) => o.ful);\n const inProgress = os.length - shipped.length;\n const days = shipped.filter((o) => o.f && o.d)\n .map((o) => (new Date(o.f!).getTime() - new Date(o.d).getTime()) / 864e5)\n .filter((d) => d >= 0 && d < 60);\n const avgDays = days.length ? days.reduce((s, d) => s + d, 0) / days.length : null;\n return { total: os.length, shipped: shipped.length, inProgress, avgDays };\n }, [orders, products, brand, grp, from, to]);\n\n return (\n
\n \n
\n From\n setFrom(e.target.value)}\n className=\"px-2 py-1.5 border rounded-lg text-sm\" disabled={!hasOrders} />\n to\n setTo(e.target.value)}\n className=\"px-2 py-1.5 border rounded-lg text-sm\" disabled={!hasOrders} />\n \n {!hasOrders && (\n Date filter needs the Orders exports — using Sales YTD for now.\n )}\n
\n
\n setBrand(null)}>All brands\n {BRANDS.map((b) => setBrand(brand === b ? null : b)}>{b})}\n |\n setGrp(null)}>All categories\n {GROUPS.filter((g) => g.key !== \"Other\").map((g) =>\n setGrp(grp === g.key ? null : g.key)}>\n )}\n
\n \n\n
\n \n \n \n \n \n
\n\n {shipStats ? (\n
\n \n \n \n
\n ) : orders === null ? (\n \n
\n Configure the Matrixify Orders exports (see the Business tab) to see orders in progress,\n shipped orders and average days to ship (order date → fulfilment date).\n
\n \n ) : null}\n\n \n
\n \n \n \n \n \n \n \n \n \n \n \n \n {rows.slice(0, 300).map(({ p, units }) => (\n \n \n \n \n \n \n \n \n \n \n ))}\n \n
SKU Product Category Weight (kg) Carrier Unit cost Units {hasOrders ? \"(period)\" : \"(YTD)\"} Spend
{p.code} {p.name} {p.cat ?? \"—\"} {p.wt != null ? p.wt.toFixed(1) : \"—\"} \n \n {p.carrier}\n \n {fmtEur(p.logi)} {fmtInt(units)} {fmtEurK((p.logi ?? 0) * units)}
\n
\n \n
\n );\n}\n\n// ═══════════════════════════ BUSINESS ═══════════════════════════════════════\nfunction BusinessTab({ products, cutover }: { products: Product[]; cutover?: string }) {\n // Year to date by default, so this tab lines up with the Overview headline at a glance\n const [from, setFrom] = useState(() => new Date().getFullYear() + \"-01-01\");\n const [to, setTo] = useState(() => new Date().toISOString().slice(0, 10));\n const [chans, setChans] = useState>(new Set()); // empty = all channels\n const [brand, setBrand] = useState(null);\n const [grp, setGrp] = useState(null);\n const [basis, setBasis] = useState<\"ful\" | \"ord\">(\"ful\"); // shipped vs ordered\n const [exclRefund, setExclRefund] = useState(true); // drop refunded orders\n\n const toggleChan = (c: string) => {\n const s = new Set(chans);\n s.has(c) ? s.delete(c) : s.add(c);\n setChans(s);\n };\n\n const data = useOrders(); // undefined=chargement, null=flux absent\n\n const skuInfo = useMemo(() => {\n const m: Record = {};\n for (const p of products) m[String(p.code)] = p;\n return m;\n }, [products]);\n\n if (data === undefined) return
Loading business data…
;\n\n if (data === null || !Array.isArray(data.lines)) {\n return (\n \n
\n
To activate this tab, create 3 scheduled Matrixify exports (Orders) — one per store — with:
\n
\n
Format: Matrixify: CSV, not zipped, repeat every 1 day
\n
Custom file names: Export_Hoover_Orders, Export_Haier_Orders, Export_Candy_Orders (no date placeholder)
\n
Columns: ID, Name, Tags, Created At, Source, Line: Type, Line: SKU,\n Line: Quantity, Line: Price, Line: Discount, Line: Total,\n Refund: Created At, Transaction: Status, Fulfillment: Status,\n Fulfillment: Created At, Fulfillment: Location
\n
Line: Type is what exposes the Refund Line rows — without it,\n refunds carry no amount and cannot be booked in their own month
\n
Add Transaction: Kind and Transaction: Amount: refunds entered in\n Shopify without item quantities have no line detail at all, and their amount lives\n only on the refund transaction. Without these two columns such refunds can be\n counted but not valued.
\n
Date filter: last 12 months (rolling) to keep files light
\n
\n
The dashboard checks the same stable URLs pattern as the product exports — data appears after the first daily sync (or ask Claude to trigger it).
\n
\n \n );\n }\n\n const lines: any[] = data.lines;\n\n // Channel from marketplace tags. Tags hold: marketplace order no. + integrator (lengow)\n // + marketplace identifier. Klarna is a payment method → DtoC. No marketplace tag → DtoC.\n const chanOf = (l: any) => {\n const t = (l.t ?? \"\").toLowerCase();\n if (t.includes(\"but_mirakl\")) return \"BUT\";\n if (t.includes(\"cdiscount_octopia\")) return \"Cdiscount\";\n if (t.includes(\"boulanger\")) return \"Boulanger\";\n if (t.includes(\"showroomprive\")) return \"Showroomprive\";\n return \"DtoC\";\n };\n const CHANNELS = [\"DtoC\", \"BUT\", \"Cdiscount\", \"Boulanger\", \"Showroomprive\"];\n\n // ── Dating rules ──────────────────────────────────────────────────────────\n // basis \"ord\" → ordered lines, dated on the ORDER creation date (l.d)\n // basis \"ful\" → shipped lines, dated on the FULFILMENT date (l.f), falling back to\n // the order date when the export carries no fulfilment date\n // refunds → ALWAYS dated on the REFUND date (l.rf), never on the order date:\n // a June order refunded in July is deducted from July.\n const dateOf = (l: any) =>\n l.k === \"ref\" ? l.rf : (basis === \"ful\" ? (l.f || l.d) : l.d);\n // How many shipped lines lack a fulfilment date → shown as a caveat in the UI\n const noFulDate = basis === \"ful\"\n ? lines.filter((l: any) => (l.k ?? \"ord\") === \"ful\" && !l.f).length\n : 0;\n // True once the export carries \"Line: Type\" = Refund Line rows. Without them we only\n // know an order *was* refunded, not how much nor on which line.\n const hasRefundLines = lines.some((l: any) => l.k === \"ref\");\n\n // La part BI n'est ajoutée que si la période remonte avant la bascule, et jamais\n // avec un filtre canal actif (le BI n'a pas de ventilation par canal).\n const biIncluded = chans.size === 0 && from < (cutover ?? DEFAULT_CUTOVER);\n // Quand le BI couvre les mois d'avant, Shopify doit démarrer à la bascule — sinon\n // janvier–février serait compté deux fois.\n const shopFrom = shopifyWindowStart(from, cutover ?? DEFAULT_CUTOVER, biIncluded);\n\n const inScope = (l: any) => {\n const d = dateOf(l);\n if (!d || d < shopFrom || d > to) return false;\n if (chans.size > 0 && !chans.has(chanOf(l))) return false;\n if (brand && l.b !== brand) return false;\n if (grp) {\n const p = skuInfo[l.sku ?? \"\"];\n if (!p || p.grp !== grp) return false;\n }\n return true;\n };\n\n // Positive side: ordered or shipped lines, per the selected basis\n const fAll = lines.filter((l) => (l.k ?? \"ord\") === basis && inScope(l));\n // Negative side: refund lines falling in the period by their own refund date\n const refLines = hasRefundLines ? lines.filter((l) => l.k === \"ref\" && inScope(l)) : [];\n // Legacy fallback (no Refund Line rows yet): drop whole refunded orders when asked\n const fLines = hasRefundLines ? fAll : (exclRefund ? fAll.filter((l) => !l.rf) : fAll);\n\n // Net amount actually charged when the export provides it (Line: Total / Line: Discount),\n // otherwise the pre-discount line value. `hasNet` says whether the data is available.\n const hasNet = lines.some((l: any) => l.n != null);\n const amountOf = (l: any) => (l.n != null ? l.n : l.p * l.q);\n const grossOf = (l: any) => l.p * l.q;\n\n // KPIs — Shopify prices are VAT-inclusive; Business figures are shown excl. VAT\n const VAT_RATE = 1.2;\n // The eco-participation is included in the site price but excluded from the BI's NSV.\n // We strip it too, so that the dashboard and the BI measure the same thing.\n const deeeOf = (l: any) => skuInfo[l.sku ?? \"\"]?.deee ?? 0;\n /** Goods revenue excl. VAT and excl. eco-participation — the BI's definition. */\n const goodsOf = (l: any) => amountOf(l) / VAT_RATE - deeeOf(l) * l.q;\n\n // Delivery charged to the customer — the equivalent of the BI's AESFR001 line.\n // `ship` is an order-level total repeated on every line, so dedupe by order.\n const shippingOf = (lines: any[]) => {\n const seen = new Set();\n let s = 0;\n for (const l of lines) {\n if (!l.ship || seen.has(l.o)) continue;\n seen.add(l.o);\n s += l.ship / VAT_RATE;\n }\n return s;\n };\n\n const refundVal = refLines.reduce((s, l) => s + goodsOf(l), 0);\n const refundUnits = refLines.reduce((s, l) => s + l.q, 0);\n const goodsBeforeRefunds = fLines.reduce((s, l) => s + goodsOf(l), 0);\n const shipping = shippingOf(fLines);\n // Complément BI pour les mois d'avant la bascule (voir biIncluded / shopFrom).\n const scopedForPre = products.filter((p) =>\n (!brand || p.brand === brand) && (!grp || p.grp === grp));\n const preBi = biIncluded ? preCutoverRevenue(scopedForPre, from, cutover) : 0;\n\n // ── Avoirs sans détail article ──────────────────────────────────────────────\n // Remboursements saisis dans Shopify sans quantité sur les lignes : ils n'ont aucun\n // `refund_line_item`, donc ni SKU ni montant dans les colonnes de ligne. Le pipeline\n // les range à part (`orphan_refunds`) plutôt que parmi les lignes produit — les y\n // mêler fausserait marge, quantités et classements, puisqu'on ignore ce qui a été rendu.\n //\n // Deux conséquences assumées :\n // • le montant vient de la transaction, donc TTC et éco-participation comprise ;\n // on ne retire que la TVA. L'écart résiduel est l'éco-part, non ventilable.\n // • un filtre catégorie ne peut pas s'y appliquer (pas de SKU) : dans ce cas on ne\n // déduit rien et l'encart le dit, plutôt que d'imputer un avoir au mauvais rayon.\n const orphanAll: any[] = Array.isArray(data.orphan_refunds) ? data.orphan_refunds : [];\n const orphanScoped = grp ? [] : orphanAll.filter((o) =>\n o.rf >= shopFrom && o.rf <= to &&\n (!brand || o.b === brand) &&\n (chans.size === 0 || chans.has(chanOf(o))));\n const orphanKnown = orphanScoped.filter((o) => o.amt != null);\n const orphanVal = orphanKnown.reduce((s, o) => s + o.amt / VAT_RATE, 0);\n const orphanUnknown = orphanScoped.length - orphanKnown.length;\n\n const ca = goodsBeforeRefunds - refundVal - orphanVal + shipping + preBi;\n const caGross = fLines.reduce((s, l) => s + grossOf(l), 0) / VAT_RATE;\n const deeeTotal = fLines.reduce((s, l) => s + deeeOf(l) * l.q, 0);\n const discountTotal = caGross - fLines.reduce((s, l) => s + amountOf(l), 0) / VAT_RATE;\n // Unités : même composition que le CA — Shopify sur la période, plus la part BI\n // quand elle est comptée, sinon le panier moyen serait faussé.\n const preUnits = biIncluded ? preCutoverUnits(scopedForPre, from, cutover) : 0;\n\n // ── Appareils vs services ───────────────────────────────────────────────────\n // Les références FR_AES… ne sont pas des appareils : ce sont des prestations\n // rattachées à la livraison (installation, reprise, mise en service). Elles\n // génèrent du chiffre d'affaires et de la marge, et restent donc pleinement\n // comptées dans le CA — mais les additionner aux appareils dans un compteur\n // « Units sold » gonflait le volume de plus d'un tiers et rendait le panier\n // moyen unitaire ininterprétable. On les suit donc sur leur propre vignette.\n const isService = (l: any) => /AES/i.test(String(l.sku ?? \"\"));\n const svcUnitsFul = fLines.filter(isService).reduce((s, l) => s + l.q, 0);\n const svcUnitsRef = refLines.filter(isService).reduce((s, l) => s + l.q, 0);\n const svcUnits = svcUnitsFul - svcUnitsRef;\n const svcRevenue = fLines.filter(isService).reduce((s, l) => s + goodsOf(l), 0)\n - refLines.filter(isService).reduce((s, l) => s + goodsOf(l), 0);\n\n // « Units sold » ne compte plus que les appareils. La part BI d'avant la bascule\n // ne contient que des appareils, elle s'y ajoute donc sans retraitement.\n const deviceUnitsFul = fLines.filter((l) => !isService(l)).reduce((s, l) => s + l.q, 0);\n const deviceUnitsRef = refLines.filter((l) => !isService(l)).reduce((s, l) => s + l.q, 0);\n const units = deviceUnitsFul - deviceUnitsRef + preUnits;\n // Product lines with no catalogue match get no eco-participation deduction. Service\n // SKUs (FR_AES…) legitimately carry none, so they are not flagged.\n const deeeUnknown = fLines.filter(\n (l) => !skuInfo[l.sku ?? \"\"] && !/AES/i.test(String(l.sku ?? \"\")),\n ).length;\n const orderIds = new Set(fLines.map((l) => l.o));\n const nOrders = orderIds.size;\n const aov = nOrders > 0 ? ca / nOrders : null;\n // Margin estimate from catalog costs (lines with known SKU); refunds reverse the sale\n let mSum = 0, mBase = 0;\n const addMargin = (l: any, sign: number) => {\n const p = skuInfo[l.sku ?? \"\"];\n if (p?.cost && l.p > 0) {\n const ht = goodsOf(l); // net of discount and of eco-participation\n mSum += sign * (ht - p.cost * l.q);\n mBase += sign * ht;\n }\n };\n for (const l of fLines) addMargin(l, 1);\n for (const l of refLines) addMargin(l, -1);\n const margin = mBase > 0 ? mSum / mBase : null;\n\n // Refunds: number of distinct orders refunded IN this period\n const refunded = hasRefundLines\n ? new Set(refLines.map((l) => l.o)).size\n : new Set(fAll.filter((l) => l.rf).map((l) => l.o)).size;\n const refundedVal = hasRefundLines\n ? refundVal\n : fAll.filter((l) => l.rf).reduce((s, l) => s + goodsOf(l), 0);\n\n // By channel — refunds deducted from the channel that made the sale,\n // delivery added to it (counted once per order)\n const byChannel: Record; units: number; shipSeen: Set }> = {};\n const bump = (l: any, sign: number) => {\n const src = chanOf(l);\n const c = byChannel[src] ?? { ca: 0, orders: new Set(), units: 0, shipSeen: new Set() };\n c.ca += sign * goodsOf(l);\n c.units += sign * l.q;\n if (sign > 0) {\n c.orders.add(l.o);\n if (l.ship && !c.shipSeen.has(l.o)) { c.shipSeen.add(l.o); c.ca += l.ship / VAT_RATE; }\n }\n byChannel[src] = c;\n };\n for (const l of fLines) bump(l, 1);\n for (const l of refLines) bump(l, -1);\n\n // Daily revenue\n const byDate: Record = {};\n const shipSeenDay = new Set();\n for (const l of fLines) {\n const d = dateOf(l);\n byDate[d] = (byDate[d] ?? 0) + goodsOf(l);\n if (l.ship && !shipSeenDay.has(l.o)) {\n shipSeenDay.add(l.o);\n byDate[d] = byDate[d] + l.ship / VAT_RATE;\n }\n }\n for (const l of refLines) { // deducted on the refund day\n const d = dateOf(l);\n byDate[d] = (byDate[d] ?? 0) - goodsOf(l);\n }\n // fillDays réintroduit les journées sans expédition (week-ends, jours fériés).\n // Sans elles, un vendredi et un lundi apparaissaient côte à côte : l'axe des dates\n // n'était plus régulier et une journée faible se confondait avec une journée absente.\n // On prolonge jusqu'à la borne de fin choisie, sans dépasser aujourd'hui : la journée\n // en cours doit apparaître même vide. Elle reste partielle par nature — l'export\n // Matrixify est filtré sur « Created At — Last: 2160 hours », qui s'arrête à l'heure\n // ronde précédente — d'où le marquage « en cours » sur la dernière barre.\n const today = isoDay(new Date());\n const dailySeries = fillDays(\n Object.entries(byDate).sort((a, b) => a[0] < b[0] ? -1 : 1),\n to < today ? to : today,\n );\n\n return (\n
\n \n
\n From\n setFrom(e.target.value)} className=\"px-2 py-1.5 border rounded-lg text-sm\" />\n to\n setTo(e.target.value)} className=\"px-2 py-1.5 border rounded-lg text-sm\" />\n \n \n Orders data updated {data.updated_at ? new Date(data.updated_at).toLocaleString(\"fr-FR\") : \"?\"}\n \n
\n
\n Basis:\n setBasis(\"ful\")}>📤 Shipped (fulfilled)\n setBasis(\"ord\")}>🛒 Ordered\n \n {basis === \"ful\"\n ? \"dated on the fulfilment date — comparable to invoiced revenue\"\n : \"dated on the order creation date — gross intake, incl. not-yet-shipped\"}\n \n {!hasRefundLines && (\n <>\n |\n \n setExclRefund(e.target.checked)} />\n Exclude refunded orders\n \n \n )}\n
\n
\n Channels (multi):\n setChans(new Set())}>All\n {CHANNELS.map((c) => (\n toggleChan(c)}>\n {c === \"DtoC\" ? \"🌐 DtoC\" : \"🏪 \" + c}\n \n ))}\n
\n
\n setBrand(null)}>All brands\n {BRANDS.map((b) => setBrand(brand === b ? null : b)}>{b})}\n |\n setGrp(null)}>All categories\n {GROUPS.filter((g) => g.key !== \"Other\").map((g) =>\n setGrp(grp === g.key ? null : g.key)}>\n )}\n
\n \n\n
\n \n \n \n \n \n \n \n {/* Montant remboursé sur la période, imputé à la DATE DE REMBOURSEMENT et non à\n celle de la commande : une commande de juillet remboursée en août est déduite\n d'août. Le total additionne les avoirs détaillés (rattachés à des références)\n et les avoirs sans détail, sans quoi la vignette contredirait le CA juste\n au-dessus, qui déduit bien les deux. */}\n 0\n ? `Refunded this period — ${fmtInt(orphanUnknown)} not valued`\n : \"Refunded this period (excl. VAT)\"}\n value={fmtEurK(-(refundedVal + orphanVal))}\n hint={[\n `Imputé sur la date de remboursement, pas sur celle de la commande.`,\n `Avoirs détaillés : ${fmtEurK(refundedVal)} sur ${fmtInt(refunded)} commande(s).`,\n orphanScoped.length > 0\n ? `Avoirs sans détail article : ${fmtEurK(orphanVal)} sur ${fmtInt(orphanScoped.length)} commande(s).`\n : null,\n orphanUnknown > 0\n ? `${fmtInt(orphanUnknown)} avoir(s) sans montant connu ne sont PAS compris ici — activez Transaction: Kind et Transaction: Amount dans les exports Matrixify.`\n : null,\n ].filter(Boolean).join(\"\\n\")}\n Icon={Ico.IconReturn} tone=\"rose\" />\n
\n\n {orphanScoped.length > 0 && (\n
\n
\n \n
\n
\n {fmtInt(orphanScoped.length)} avoir{orphanScoped.length > 1 ? \"s\" : \"\"} sans détail article sur la période\n {orphanVal > 0 && <> — {fmtEurK(orphanVal)} déduits du CA}\n
\n
\n Ces remboursements ont été saisis dans Shopify sans quantité sur les lignes\n d'articles. Ils n'ont donc aucune référence associée : ils sont retirés du chiffre\n d'affaires mais restent invisibles dans la marge par produit et les classements.\n
\n {orphanUnknown > 0 && (\n
\n {fmtInt(orphanUnknown)} d'entre eux n'ont aucun montant connu et ne sont\n donc pas déduits. Pour les chiffrer, ajoutez les colonnes\n {\" \"}Transaction: Kind et Transaction: Amount aux trois exports\n Matrixify — le montant apparaîtra à la synchro suivante, sans autre intervention.\n
\n )}\n
\n Pour que les prochains avoirs soient ventilés par référence, le remboursement doit\n être saisi en indiquant les quantités sur les lignes d'articles plutôt qu'un montant\n global. Les avoirs marketplace, créés par API, le sont déjà et sont comptés normalement.\n
\n
\n
\n
\n )}\n\n {grp && orphanAll.some((o) => o.rf >= shopFrom && o.rf <= to) && (\n
\n Filtre catégorie actif : les avoirs sans détail article ne portent aucune référence et ne\n peuvent pas être rattachés à une catégorie. Ils ne sont donc pas déduits de ce chiffre —\n retirez le filtre pour un CA net complet.\n
\n )}\n\n {dailySeries.length > 1 && (\n \n \n \n )}\n\n \n
\n \n \n \n \n \n \n \n {Object.entries(byChannel).sort((a, b) => b[1].ca - a[1].ca).map(([src, c]) => (\n \n \n \n \n \n \n \n \n ))}\n \n
Channel Revenue Share Orders Units AOV
{src === \"DtoC\" ? \"🌐 DtoC (sites)\" : \"🏪 \" + src} {fmtEurK(c.ca)} {fmtPct(ca > 0 ? c.ca / ca : null)} {fmtInt(c.orders.size)} {fmtInt(c.units)} {fmtEur(c.orders.size > 0 ? c.ca / c.orders.size : null)}
\n
\n \n
\n );\n}\n\n// ═══════════════════════════ SHARED WIDGETS ═════════════════════════════════\n\n/** Histogramme journalier avec les montants inscrits au-dessus des barres.\n *\n * Trois partis pris, tirés de ce que le graphique précédent rendait illisible :\n * • les jours à zéro sont conservés et marqués « — », pour qu'une journée sans\n * expédition ne puisse pas être confondue avec une donnée manquante ;\n * • les week-ends sont grisés, ce qui explique visuellement ces zéros ;\n * • les barres non nulles gardent une hauteur minimale afin qu'une petite valeur\n * reste visible — c'est ce qui faisait passer 2 579 € pour une barre vide.\n *\n * Au-delà de 45 jours les étiquettes se chevauchent : elles disparaissent et\n * l'infobulle au survol prend le relais. */\nfunction DailyBars({ series }: { series: [string, number][] }) {\n if (series.length < 2) return null;\n const today = isoDay(new Date());\n const max = Math.max(...series.map(([, v]) => Math.abs(v)), 1);\n const dense = series.length > 45;\n const short = (v: number) =>\n Math.abs(v) >= 1000 ? Math.round(v / 1000) + \"k\" : String(Math.round(v));\n // Les étiquettes occupent le haut du cadre : on plafonne les barres pour éviter\n // que la plus haute ne les pousse hors du graphique.\n const ceiling = dense ? 100 : 86;\n\n return (\n <>\n
\n {series.map(([d, v]) => {\n const zero = Math.round(v) === 0;\n const we = isWeekend(d);\n const partial = d >= today; // journée en cours : jamais complète (cf. filtre Matrixify)\n return (\n
\n {!dense && (\n
\n {partial ? \"ongoing\" : zero ? \"—\" : short(v)}\n
\n )}\n
\n
\n );\n })}\n
\n
\n {series.map(([d]) => (\n
\n {dense ? \"\" : d.slice(8)}\n
\n ))}\n
\n
\n {series[0][0]} → {series[series.length - 1][0]} · pale bars = weekends ·\n “—” = no data ·{\" \"}\n hatched = day in progress, still partial\n {dense && \" · hover a bar for details\"}\n
\n \n );\n}\n\nfunction Card({ title, children, className = \"\", right }: any) {\n return (\n
\n {title && (\n // flex-wrap : sur mobile un titre long et son action à droite ne tiennent\n // pas sur une ligne — sans quoi l'un des deux débordait de la carte.\n
\n
{title}
\n {right}\n
\n )}\n
{children}
\n
\n );\n}\n\n/** KPI tile: monochrome icon in a soft tinted circle, value, label. */\n/** `hint` : infobulle portée par toute la vignette, pour expliquer une composition de\n * chiffre que le libellé seul ne peut pas dire. Sans lui, seul le libellé tronqué\n * était survolable. */\nfunction Kpi({ label, value, icon, Icon, accent = false, tone = \"indigo\", hint }: any) {\n const tones: Record = {\n indigo: [\"#EEF2FF\", \"#4F46E5\"], green: [\"#ECFDF5\", \"#059669\"],\n amber: [\"#FFFBEB\", \"#D97706\"], sky: [\"#F0F9FF\", \"#0284C7\"],\n rose: [\"#FFF1F2\", \"#E11D48\"], slate: [\"#F1F5F9\", \"#475569\"],\n };\n const [bg, fg] = tones[accent ? \"amber\" : tone] ?? tones.indigo;\n return (\n
\n
\n {Icon ? : {icon}}\n
\n
\n
{value}
\n
{label}
\n
\n
\n );\n}\n\nfunction Pill({ active, onClick, children }: any) {\n return (\n \n );\n}\n\n/** Local calendar date as YYYY-MM-DD.\n * Not toISOString(): that converts to UTC first, so 1 July 00:00 in Paris comes back\n * as 30 June — the preset would silently start a day early. */\nfunction isoDay(d: Date): string {\n const p = (n: number) => String(n).padStart(2, \"0\");\n return `${d.getFullYear()}-${p(d.getMonth() + 1)}-${p(d.getDate())}`;\n}\n\n/** Quick period presets, shared by every dated tab so they all offer the same ranges. */\nfunction DateRangePills(\n { from, to, setFrom, setTo, disabled = false }: {\n from: string; to: string;\n setFrom: (v: string) => void; setTo: (v: string) => void;\n disabled?: boolean;\n },\n) {\n const now = new Date();\n const today = isoDay(now);\n const presets: [string, string][] = [\n [\"This month\", isoDay(new Date(now.getFullYear(), now.getMonth(), 1))],\n [\"Last 30 days\", isoDay(new Date(Date.now() - 30 * 864e5))],\n [\"Last 90 days\", isoDay(new Date(Date.now() - 90 * 864e5))],\n [\"YTD\", `${now.getFullYear()}-01-01`],\n ];\n return (\n <>\n {presets.map(([label, start]) => (\n { if (!disabled) { setFrom(start); setTo(today); } }}>\n {label}\n \n ))}\n \n );\n}\n\nfunction BrandBadge({ brand }: { brand: string }) {\n return (\n \n {brand}\n \n );\n}\n\nfunction BrandDot({ brand }: { brand: string }) {\n const c = brand === \"CANDY\" ? \"bg-sky-500\" : brand === \"HOOVER\" ? \"bg-red-500\" : \"bg-blue-900\";\n return ;\n}\n\n/** Product thumbnail (from Shopify \"Image Src\" column) with brand-dot fallback. */\nfunction Thumb({ p }: { p: Product }) {\n if (p.img) {\n return (\n $\"\"$ { e.target.style.display = \"none\"; }} />\n );\n }\n return ;\n}\n\nfunction SiteBadge({ ss, queued }: { ss: string; queued?: any }) {\n const cls = ss === \"Active\" ? \"bg-green-100 text-green-700\" :\n ss === \"Inactive\" ? \"bg-orange-100 text-orange-700\" : \"bg-red-100 text-red-700\";\n return (\n \n {ss}\n {/* Shopify parle de « Draft », la colonne Site affiche « Inactive » : on garde le\n vocabulaire de la colonne pour ne pas donner l'impression de deux notions. */}\n {queued && String(v) === \"Draft\" ? \"Inactive\" : String(v)} />}\n \n );\n}\n\nfunction StBadge({ st }: { st: number | null | undefined }) {\n if (st == null || !isFinite(st)) return —;\n const cls = st >= 1.1 ? \"text-purple-700 font-bold\" : st >= 0.7 ? \"text-green-700\" : \"text-orange-600\";\n return {fmtPct(st)};\n}\n\nfunction PosBadge({ pos }: { pos?: string | null }) {\n if (!pos) return —;\n const map: Record = {\n competitive: [\"🟢 Competitive\", \"text-green-700\"],\n average: [\"🟡 Average\", \"text-amber-700\"],\n above: [\"🔴 Above market\", \"text-red-700\"],\n };\n const [label, cls] = map[pos] ?? [pos, \"\"];\n return {label};\n}\n\n/** Explique en clair où en est une modification partie du dashboard. */\nfunction queuedTooltip(c: any): string {\n const when = c.at ? new Date(c.at).toLocaleString(\"fr-FR\") : \"?\";\n const who = displayUser(c.user);\n return c.status === \"sent\"\n ? `Changed on ${when} by ${who}.\\nDelivered to Matrixify on ${\n c.sentAt ? new Date(c.sentAt).toLocaleString(\"fr-FR\") : \"?\"\n }.\\nThe value shown comes from the last site export and will update at the next one.`\n : `Changed on ${when} by ${who}.\\nWaiting for the next Matrixify import.\\nThe value shown is still the live site price.`;\n}\n\n/** Pastille « en attente » : ancienne valeur barrée, nouvelle valeur mise en avant.\n * Sans cela, l'équipe revenait sur l'onglet, voyait l'ancien prix et refaisait la\n * modification en croyant que rien n'avait été enregistré. */\nfunction QueuedTag({ c, format }: { c: any; format: (v: any) => string }) {\n const sent = c.status === \"sent\";\n return (\n \n → {format(c.newValue)}\n \n );\n}\n\n/** Cellule prix, éditable au clic.\n *\n * La saisie NE MODIFIE PAS le site directement : elle empile une modification dans la\n * même file que l'édition en masse, avec la même piste d'audit et le même passage par\n * le flux Matrixify. Un prix saisi ici et un prix saisi via la sélection multiple\n * suivent exactement le même chemin — c'est la raison pour laquelle on réutilise\n * POST /api/changes plutôt que d'ouvrir une voie parallèle.\n *\n * Trois garde-fous à la saisie :\n * • un prix déjà en attente ne se ré-édite pas — il faut d'abord annuler la\n * modification en file, sinon on empilerait deux valeurs contradictoires ;\n * • un prix sous le seuil de rentabilité demande une confirmation explicite ;\n * • une valeur identique à l'existante n'envoie rien, plutôt que de polluer\n * l'historique d'une modification sans effet.\n */\nfunction PriceCell(\n { p, queued, canEdit = false, onChanged }:\n { p: Product; queued?: any; canEdit?: boolean; onChanged?: () => void },\n) {\n const price = p.price;\n const [editing, setEditing] = useState(false);\n const [draft, setDraft] = useState(\"\");\n const [busy, setBusy] = useState(false);\n const tag = queued ? fmtEur(Number(v))} /> : null;\n\n const save = async () => {\n const np = Math.round(Number(draft.replace(\",\", \".\")) * 100) / 100;\n if (!isFinite(np) || np <= 0) { setEditing(false); return; }\n if (price != null && Math.abs(np - price) < 0.005) { setEditing(false); return; }\n if (p.be != null && np < p.be &&\n !confirm(`${fmtEur(np)} is below breakeven (${fmtEur(p.be)}). Queue it anyway?`)) return;\n setBusy(true);\n try {\n const r = await fetch(\"/api/changes\", {\n method: \"POST\",\n headers: { \"Content-Type\": \"application/json\" },\n body: JSON.stringify({\n changes: [{\n code: p.code, field: \"price\", newValue: np,\n oldValue: price ?? null, applyAt: null,\n }],\n }),\n });\n if (!r.ok) {\n const j = await r.json().catch(() => ({}));\n alert(j.message ?? \"Change refused.\");\n } else {\n onChanged?.();\n }\n } finally {\n setBusy(false);\n setEditing(false);\n }\n };\n\n if (editing) {\n // `onBlur={save}` : sortir du champ vaut validation. Un prix saisi puis abandonné\n // par un clic ailleurs serait sinon perdu sans que rien ne le signale.\n return (\n \n setDraft(e.target.value)}\n onKeyDown={(e) => {\n if (e.key === \"Enter\") save();\n if (e.key === \"Escape\") setEditing(false);\n }}\n onBlur={save}\n className=\"w-20 px-1 py-0.5 border border-indigo-400 rounded text-right tabular-nums text-xs\"\n title=\"Enter to queue · Escape to cancel\" />\n €\n \n );\n }\n\n const startEdit = () => {\n setDraft(price != null ? String(price) : \"\");\n setEditing(true);\n };\n\n // Une modification déjà en file bloque l'édition : le prix affiché n'est plus celui\n // qui fait foi, et empiler une seconde valeur rendrait la file ambiguë.\n const locked = !!queued;\n const editable = canEdit && !locked;\n\n let cls = \"\";\n let title = \"\";\n if (price != null && price !== 0) {\n if (p.be != null && price < p.be) {\n cls = \"bg-red-100 text-red-800 px-1 rounded\"; title = `Below breakeven (${fmtEur(p.be)})`;\n } else if (p.mip != null && price < p.mip) {\n cls = \"bg-orange-100 text-orange-800 px-1 rounded\"; title = `Below min internal price (${fmtEur(p.mip)})`;\n }\n }\n if (editable) {\n title = title ? `${title} — click to change` : \"Click to change the price\";\n } else if (locked) {\n title = \"A change is already queued — cancel it in the Changes tab to edit again\";\n }\n\n const label = price == null || price === 0\n ? —\n : (\n \n {fmtEur(price)}\n \n );\n\n return (\n \n {editable\n ? (\n \n )\n : label}\n {tag}\n \n );\n}\n\nfunction GapCell({ p }: { p: Product }) {\n if (p.gap == null) return —;\n const cls = p.gap <= -0.01 ? \"text-green-700\" : p.gap <= 0.05 ? \"text-amber-700\" : \"text-red-700 font-bold\";\n return {fmtPct(p.gap)};\n}\n",[],[220,590],"7028cfb8-621d-435e-a23c-f93a1ebcc67b","Amazon.tsx","01a033f6-8ab0-7225-822e-c718ea0931c3",["D",1787578190526],"/** @jsxImportSource https://esm.sh/react@18.2.0 */\n// Amazon Vendor tab — DELIBERATELY ISOLATED from D2C.\n//\n// Amazon Vendor is 1P: Amazon buys from us, then resells to the end customer. Two\n// notions of revenue coexist, and confusing them invalidates everything:\n//\n// • SELL-IN — what Amazon pays US, gross of contract commission.\n// Source: `Dispatched COGS` in the Retail Analytics reports.\n// • NSV — what we actually book: sell-in less the contract commission we pay\n// back to Amazon, less any SOA funding. This is the figure comparable\n// to the NSV used elsewhere in the business.\n// • SELL-OUT — what the end customer pays Amazon. NOT our revenue; it measures demand.\n//\n// No figure on this tab may be added to the D2C tabs.\n//\n// Two margins also coexist, from different sources:\n// • our vendor net margin, computed from our PO and standard costs, using the same\n// formulas as the Amazon_Vendor_Rentabilite workbook;\n// • Amazon's Net PPM, TAKEN AS PUBLISHED. We never recompute it: the data review\n// showed it cannot be reliably rebuilt from revenue and COGS (the contract\n// retrocession explains the mechanism on only half the references).\nimport { useEffect, useMemo, useState } from \"https://esm.sh/react@18.2.0\";\nimport { fmtEur, fmtEurK, fmtInt, fmtPct } from \"/frontend/helpers.ts\";\nimport * as Ico from \"/frontend/Icons.tsx\";\n\nconst NAVY = \"#1F3864\";\nconst ACCENT = \"#4F46E5\";\n\ninterface Sellout {\n orderedRev: number | null; orderedUnits: number | null;\n shippedRev: number | null; shippedCogs: number | null; shippedUnits: number | null;\n returns: number | null; unfilled: number | null; confirmedUnits: number | null;\n netPpm: number | null; confirmPct: number | null; poReal: number | null;\n brand?: string | null; cat?: string | null; title?: string | null;\n}\ninterface AmzProduct {\n asin: string; ean: string | null; sku: string; name: string;\n cat: string | null; grp: string | null;\n po: number | null; scti: number | null; deee: number | null; rrp: number | null;\n mnEur: number | null; mnPct: number | null; mbPct: number | null;\n soaMax: number | null; soa: number | null;\n fc: number; sales: number;\n stockPhy: number | null; stockM3: number | null;\n offCatalogue?: boolean;\n sellout?: Record;\n}\n\nexport default function Amazon() {\n const [data, setData] = useState(undefined); // undefined = loading\n const [period, setPeriod] = useState(\"\");\n const [grp, setGrp] = useState(null);\n const [q, setQ] = useState(\"\");\n const [view, setView] = useState<\"all\" | \"opportunities\">(\"opportunities\");\n // SOA simulator: free per-ASIN input, never persisted. A point-in-time projection\n // tool, not a commitment — the Excel workbook remains the record.\n const [soaSim, setSoaSim] = useState>({});\n\n useEffect(() => {\n fetch(\"/api/amazon\").then((r) => r.ok ? r.json() : null)\n .then((d) => {\n setData(d);\n const ps = Object.keys(d?.sellout_periods ?? {}).sort();\n if (ps.length) setPeriod(ps[ps.length - 1]);\n })\n .catch(() => setData(null));\n }, []);\n\n // ⚠️ RULES OF HOOKS — every useMemo below must run on EVERY render, therefore\n // BEFORE any conditional return. An early return placed above would change the\n // number of hooks between the \"loading\" render and the \"data ready\" render, which\n // React rejects (error #310 — the section is replaced by an error message).\n // Values are computed defensively, on an empty array when needed, and the\n // conditional returns only appear once every hook has run.\n const products: AmzProduct[] = Array.isArray(data?.products) ? data.products : [];\n const periods: string[] = Object.keys(data?.sellout_periods ?? {}).sort();\n const so = (p: AmzProduct): Sellout | undefined => p.sellout?.[period];\n\n /** Contract commission rate, or null when the category is unknown. References sold\n * by Amazon but absent from the profitability file have no category, hence no\n * contract: their NSV is left uncomputed rather than guessed. */\n const commRate = (p: AmzProduct): number | null =>\n (p.grp && data?.contracts?.[p.grp]?.comm != null) ? data.contracts[p.grp].comm : null;\n\n /** SOA per unit entered in the simulator, 0 when empty. */\n const soaPerUnit = (p: AmzProduct): number => {\n const raw = soaSim[p.asin];\n const v = raw ? parseFloat(raw.replace(\",\", \".\")) : NaN;\n return isFinite(v) ? v : 0;\n };\n\n /** NSV = sell-in − contract commission − SOA.\n * Returns null when the commission rate is unknown: presenting sell-in as NSV\n * would overstate revenue by roughly 20 %. */\n const nsvOf = (p: AmzProduct, s: Sellout | undefined) => {\n const k = commRate(p);\n const cogs = s?.shippedCogs ?? 0;\n if (k == null || !cogs) return null;\n const commission = cogs * k;\n const afterComm = cogs - commission;\n const soa = soaPerUnit(p) * (s?.shippedUnits ?? 0);\n return { commission, afterComm, soa, afterSoa: afterComm - soa };\n };\n\n const groups = useMemo(\n () => [...new Set(products.map((p) => p.grp).filter(Boolean))] as string[],\n [products],\n );\n\n const scoped = useMemo(() => {\n const ql = q.trim().toLowerCase();\n return products.filter((p) =>\n (!grp || p.grp === grp) &&\n (!ql || p.asin.toLowerCase().includes(ql) || String(p.sku).toLowerCase().includes(ql) ||\n (p.name ?? \"\").toLowerCase().includes(ql)));\n }, [products, grp, q]);\n\n // Only references that sold in the selected period feed the totals: including the\n // dormant catalogue would flatten every average rate.\n const sold = useMemo(() => scoped.filter((p) => so(p)), [scoped, period]);\n\n const kpi = useMemo(() => {\n let sellIn = 0, sellOutRev = 0, sellOutU = 0, shippedU = 0;\n let unfilled = 0, returns = 0, ordered = 0, confirmed = 0;\n let npmW = 0, npmBase = 0, mnW = 0, mnBase = 0;\n let nsv = 0, nsvSoa = 0, commission = 0, soaTotal = 0, sellInNoContract = 0;\n for (const p of sold) {\n const s = so(p)!;\n const cogs = s.shippedCogs ?? 0;\n sellIn += cogs;\n sellOutRev += s.orderedRev ?? 0;\n sellOutU += s.orderedUnits ?? 0;\n shippedU += s.shippedUnits ?? 0;\n unfilled += s.unfilled ?? 0;\n returns += s.returns ?? 0;\n ordered += s.orderedUnits ?? 0;\n confirmed += s.confirmedUnits ?? 0;\n const n = nsvOf(p, s);\n if (n) {\n nsv += n.afterComm; nsvSoa += n.afterSoa;\n commission += n.commission; soaTotal += n.soa;\n } else {\n // Tracked separately and shown on screen: 900 k€ silently missing from NSV\n // would be far more damaging than an NSV flagged as partial.\n sellInNoContract += cogs;\n }\n // Weighted by sell-in: a margin on 200 k€ does not carry the same weight as a\n // margin on 200 €.\n if (s.netPpm != null && cogs > 0) { npmW += s.netPpm * cogs; npmBase += cogs; }\n if (p.mnPct != null && cogs > 0) { mnW += p.mnPct * cogs; mnBase += cogs; }\n }\n // The NSV rate is expressed on the sell-in it actually covers, not on total\n // sell-in — otherwise the uncovered references would drag it down artificially.\n const covered = sellIn - sellInNoContract;\n return {\n sellIn, sellOutRev, sellOutU, shippedU, unfilled, returns,\n nsv, nsvSoa, commission, soaTotal, sellInNoContract, covered,\n refs: sold.length,\n nsvRate: covered > 0 ? nsvSoa / covered : null,\n npm: npmBase > 0 ? npmW / npmBase / 100 : null, // Net PPM is a %, not a fraction\n mn: mnBase > 0 ? mnW / mnBase : null,\n retRate: shippedU > 0 ? returns / shippedU : null,\n confRate: ordered > 0 ? confirmed / ordered : null,\n };\n }, [sold, period, soaSim, data]);\n\n // ── Opportunities ───────────────────────────────────────────────────────────\n // Four signals, each paired with the action it calls for. A metric that does not\n // tell you what to do is just another number.\n const opps = useMemo(() => {\n const lowConfirm: AmzProduct[] = [];\n const unfilledDemand: AmzProduct[] = [];\n const lowMargin: AmzProduct[] = [];\n const rrpTrap: AmzProduct[] = [];\n for (const p of sold) {\n const s = so(p)!;\n if ((s.confirmPct ?? 100) < 60 && (s.orderedUnits ?? 0) >= 50) lowConfirm.push(p);\n if ((s.unfilled ?? 0) > 0) unfilledDemand.push(p);\n if (p.mnPct != null && p.mnPct < 0.10 && (s.shippedCogs ?? 0) > 10000) lowMargin.push(p);\n // Below an RRP/PO ratio of 1.2, VAT alone wipes out Amazon's margin: it sells at\n // a loss at list price, which is why it declines to promote the item.\n if (p.rrp && p.po && p.rrp / p.po < 1.2) rrpTrap.push(p);\n }\n const byCogs = (a: AmzProduct, b: AmzProduct) =>\n (so(b)!.shippedCogs ?? 0) - (so(a)!.shippedCogs ?? 0);\n const byUnfilled = (a: AmzProduct, b: AmzProduct) =>\n (so(b)!.unfilled ?? 0) - (so(a)!.unfilled ?? 0);\n return {\n lowConfirm: lowConfirm.sort(byCogs),\n unfilledDemand: unfilledDemand.sort(byUnfilled),\n lowMargin: lowMargin.sort(byCogs),\n rrpTrap: rrpTrap.sort(byCogs),\n };\n }, [sold, period]);\n\n const rows = useMemo(() => [...sold].sort(\n (a, b) => (so(b)!.shippedCogs ?? 0) - (so(a)!.shippedCogs ?? 0)), [sold, period]);\n\n // ── Conditional returns — AFTER every hook, never before ─────────────────────\n if (data === undefined) {\n return
Loading Amazon data…
;\n }\n if (products.length === 0) {\n return (\n \n
\n
\n Double-click “Envoyer vers le dashboard.command” in the\n Amazon France folder.\n
\n
\n It reads the catalogue, PIPO, standard costs, SCP and the Retail Analytics\n exports, computes profitability, then sends the result here.\n
\n
\n \n );\n }\n\n return (\n
\n {/* Standing reminder: this tab is not comparable to D2C. */}\n
\n Amazon Vendor (1P) scope, isolated from D2C. Sell-in is what Amazon\n pays us, gross of contract commission. NSV is what we actually book:\n sell-in less commission and SOA. Sell-out is what the end customer pays\n Amazon — it measures demand, not our revenue. No figure here may be added to the\n D2C tabs.\n
\n\n \n
\n setQ(e.target.value)}\n placeholder=\"ASIN, SKU or name…\"\n className=\"px-3 py-1.5 border rounded-lg text-sm w-full sm:w-64\" />\n \n {periods.map((y) => (\n setPeriod(y)}>{y}\n ))}\n \n setGrp(null)}>All categories\n {groups.map((g) => (\n setGrp(g)}>{g}\n ))}\n
\n {data.sellout_periods?.[period] && (\n
\n Sell-out: {data.sellout_periods[period].period} · Manufacturing view ·\n {\" \"}{data.sellout_periods[period].file}\n
\n )}\n \n\n
\n \n \n \n \n \n \n \n \n
\n\n {/* An NSV covering only part of the scope must say so, plainly. */}\n {kpi.sellInNoContract > 0 && (\n
\n {fmtEurK(kpi.sellInNoContract)} of sell-in carries no contract\n parameters — references Amazon sells that are absent from the profitability\n file, so their category and commission rate are unknown. They are\n excluded from NSV rather than valued at a guessed rate. Adding them to\n the Amazon catalogue would close the gap.\n
\n )}\n\n
\n setView(\"opportunities\")}>\n Opportunities\n \n setView(\"all\")}>\n All references ({rows.length})\n \n
\n\n {view === \"opportunities\" ? (\n
\n fmtPct((s.confirmPct ?? 0) / 100), bad: true },\n { h: \"Ordered\", get: (s) => fmtInt(s.orderedUnits) },\n { h: \"Sell-in\", get: (s) => fmtEurK(s.shippedCogs) },\n ]} />\n\n fmtInt(s.unfilled), bad: true },\n { h: \"Ordered\", get: (s) => fmtInt(s.orderedUnits) },\n { h: \"Sell-in\", get: (s) => fmtEurK(s.shippedCogs) },\n ]} />\n\n fmtEurK(s.shippedCogs) },\n { h: \"Amz Net PPM\", get: (s) => fmtPct((s.netPpm ?? 0) / 100) },\n ]}\n extra={(p) => {fmtPct(p.mnPct)}}\n extraHead=\"Our margin\" />\n\n fmtEurK(s.shippedCogs) }]}\n extra={(p) => (\n \n {p.rrp && p.po ? (p.rrp / p.po).toFixed(2) : \"—\"}\n \n )}\n extraHead=\"RRP / PO\" />\n
\n ) : (\n \n
\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n {rows.slice(0, 400).map((p) => {\n const s = so(p)!;\n const n = nsvOf(p, s);\n const k = commRate(p);\n return (\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n );\n })}\n \n
ASIN SKU Product Cat. PO Std cost RRP Our margin Amz Net PPM Sell-in Commission NSV Units Confirmed Unfilled SOA / unit NSV after SOA
\n {p.asin}\n {p.offCatalogue && (\n \n off-cat\n \n )}\n {p.sku} {p.name} {p.grp ?? \"—\"} {fmtEur(p.po)} {fmtEur(p.scti)} {fmtEur(p.rrp)} \n {fmtPct(p.mnPct)}\n \n {s.netPpm != null ? fmtPct(s.netPpm / 100) : \"—\"}\n {fmtEurK(s.shippedCogs)} \n {n ? \"−\" + fmtEurK(n.commission) : \"—\"}\n \n {n ? fmtEurK(n.afterComm) : n/a}\n {fmtInt(s.orderedUnits)} \n {s.confirmPct != null ? fmtPct(s.confirmPct / 100) : \"—\"}\n 0 ? \"text-amber-700\" : \"text-slate-400\"}`}>\n {s.unfilled ? fmtInt(s.unfilled) : \"—\"}\n \n setSoaSim({ ...soaSim, [p.asin]: e.target.value })}\n placeholder=\"—\" inputMode=\"decimal\"\n className=\"w-16 px-1 py-0.5 border rounded text-right tabular-nums text-[11px]\"\n title=\"Simulation: SOA amount per unit. Not saved.\" />\n \n {n\n ? \n {fmtEurK(n.afterSoa)}\n \n : n/a}\n
\n
\n
\n NSV = sell-in − contract commission − SOA. The SOA / unit column\n is a projection tool: values are not saved and disappear on reload.\n Commitments stay in the Amazon_Vendor_Rentabilite workbook.\n References marked off-cat have no contract, hence no NSV.\n
\n \n )}\n
\n );\n}\n\n/** Opportunity card: the signal, why it matters, and the action it calls for. */\nfunction OppCard(\n { title, why, action, items, period, cols, extra, extraHead }: {\n title: string; why: string; action: string;\n items: AmzProduct[]; period: string;\n cols: { h: string; get: (s: Sellout) => any; bad?: boolean }[];\n extra?: (p: AmzProduct) => any; extraHead?: string;\n },\n) {\n const [open, setOpen] = useState(false);\n if (items.length === 0) {\n return (\n \n
No reference affected in this period.
\n \n );\n }\n const shown = open ? items : items.slice(0, 6);\n return (\n \n
{why}
\n
→ {action}
\n
\n \n \n \n \n {extraHead && }\n {cols.map((c) => )}\n \n \n {shown.map((p) => {\n const s = p.sellout![period];\n return (\n \n \n \n {extra && }\n {cols.map((c) => (\n \n ))}\n \n );\n })}\n \n
ASIN Product {extraHead} {c.h}
{p.asin} {p.name} {extra(p)} \n {c.get(s)}\n
\n
\n {items.length > 6 && (\n \n )}\n \n );\n}\n\nfunction Card({ title, children }: any) {\n return (\n
\n {title && (\n
\n
{title}
\n
\n )}\n
{children}
\n
\n );\n}\n\nfunction Kpi({ label, value, Icon, tone = \"indigo\", accent = false, hint }: any) {\n const tones: Record = {\n indigo: [\"#EEF2FF\", \"#4F46E5\"], green: [\"#ECFDF5\", \"#059669\"],\n amber: [\"#FFFBEB\", \"#D97706\"], sky: [\"#F0F9FF\", \"#0284C7\"],\n rose: [\"#FFF1F2\", \"#E11D48\"], slate: [\"#F1F5F9\", \"#475569\"],\n };\n const [bg, fg] = tones[accent ? \"amber\" : tone] ?? tones.indigo;\n return (\n
\n
\n \n
\n
\n
{value}
\n
{label}
\n
\n
\n );\n}\n\nfunction Pill({ active, onClick, children }: any) {\n return (\n \n );\n}\n",[],[220,597],["D",1787578190526],[237,238,239,240,241,242,243,244,245,246,247,248,249,250],[220],["D",1785139688485],["D",1789474853273],{"_103":102,"_111":107},{"_103":102,"_104":105,"_106":107,"_18":56,"_108":109,"_110":25},{"_13":90,"_32":91,"_44":43,"_92":45,"_93":94,"_95":96,"_97":51,"_98":45,"_99":66,"_100":46,"_101":102},"ctx",{"_83":-5,"_132":644,"_645":646,"_647":648,"_649":25,"_650":25,"_651":25,"_652":60,"_653":-5,"_654":-7,"_655":-7,"_656":64,"_657":25,"_658":25,"_659":-5,"_660":25,"_661":25,"_662":25,"_663":-5},"ENV",{"_615":616,"_617":618,"_619":138,"_620":621,"_622":623,"_624":625,"_626":627,"_628":629,"_630":631,"_632":633,"_634":635,"_636":637,"_638":639,"_640":641,"_642":643},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","e87eb04e707cf64e7e532c8e8ec395f209e17f15","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]

	Products	Sales (u)	{total?.revIsInvoiced === false ? \"Revenue est.\" : \"Revenue\"}	Forecast (u)	Sell-thru
{r.label}	{fmtInt(r.count)}	{fmtInt(r.sales)}	{fmtEurK(r.rev)}	{fmtInt(r.fc)}
TOTAL	{fmtInt(total.count)}	{fmtInt(total.sales)}	{fmtEurK(total.rev)}	{fmtInt(total.fc)}	{fmtPct(total.st)}

Diff · v2