⭐ Trustpilot Tracker

Every day at 09:00 UTC this val:

Pulls the Metabase public question and gathers all distinct URLs.
Rewrites Trustpilot /evaluate/ links → /review/.
Visits each Trustpilot page and scrapes the trust score (e.g. 4.1) and review count (e.g. 45 reviews), plus the date of the most recent review.
Saves a row per tenant per day into SQLite.

The dashboard pivots everything into the wide shape you asked for:

Tenant | Link | Score DayX | Reviews DayX | LastReview DayX | Score DayY | Reviews DayY | … | Δ Score | Δ Reviews | Δ Reviews %

Files

Rendering mermaid diagram...

File	Role
`daily.ts`	Cron trigger (09:00 UTC). Change to `0 7 * * *` for 9 AM Paris in winter.
`app.ts`	HTTP dashboard — set this as the val's HTTP preview. Routes: `/` table, `/run` manual scrape, `/csv` export.
`runScrape.ts`	Orchestrates a full daily run.
`scrape.ts`	Gathers URLs from Metabase + parses Trustpilot pages.
`pivot.ts`	Reshapes long rows into the wide per-day table with growth columns.
`db.ts`	SQLite schema + upsert/read helpers.
`dashboard.html`	Static UI shell.

Open the dashboard and click ▶ Run scrape now to do the first pull immediately (don't wait for 9 AM).
⬇ Download CSV exports the full wide table.

Timezone: crons run in UTC. 0 9 * * * = 9 AM UTC. For 9 AM Europe/Paris use 0 7 * * * (summer) or 0 8 * * * (winter).
Non-Trustpilot platforms (Google, Trustedshops, Verified-Reviews…) are recorded with platform != trustpilot and a needs guidance note — share their structure and I'll add parsers.
Metabase access: the URL is …internal.bigblue.co. If it requires VPN/auth that Val Town can't reach, the scraper will get nothing — in that case we'll need an auth token or to mirror the question's results somewhere fetchable.