Version: 1.0 Status: Foundational scoring specification

A score is a calculated representation of defined evidence. It is never a substitute for evidence and must never be generated merely because a user expects a number.

Every measurement must retain: audit reference, metric name, definition, source/evidence reference, method, input values, unit, timestamp where relevant, calculation, result, limitations, and measurement status.

• Measured: required inputs were actually obtained and the defined method was executed.
• Partially measured: some required inputs were obtained, but the result has material limitations.
• Estimated: result depends on explicit assumptions and is labelled as such.
• Unavailable: required evidence/input could not be obtained.
• Not applicable: metric does not apply to the scoped subject.
• Inconclusive: evidence exists but cannot support a reliable measurement.

A score is eligible only if:

1. a published rubric exists;
2. all mandatory inputs for that rubric are available;
3. the measurement method was actually executed;
4. missing values were not silently substituted;
5. the calculation is reproducible; and
6. the limitations do not invalidate the interpretation.

If any mandatory condition fails, the result is Not Scorable.

Missing, inaccessible, stale, contradictory, or unverified inputs must remain explicitly labelled. They may not automatically become zero, average, worst-case, best-case, or a neutral value unless the declared rubric explicitly defines that treatment before the audit.

Where a rubric permits scoring, each component must identify its weight, allowed range, input source, normalization rule, and calculation. The final score must be reproducible from the stored component records. Rounding must occur only after the underlying calculation.

A benchmark may be shown only when its source, population, period, definition, and comparability are known. A generic industry number is not evidence about an individual business.

A measured correlation, observation, or diagnostic finding must not be presented as proof that a particular business outcome was caused by a particular factor unless the evidence and methodology support causal attribution.

Scores and measurements describe the assessed state. Recommendations describe proposed action. A recommendation must never be used as evidence for a score.

Before publishing a scored result, verify: rubric version, mandatory inputs, evidence references, calculation trace, missing-data treatment, benchmark provenance where applicable, timestamp, limitations, and reviewer/system identity.

Rubric versions must be retained. A later audit must identify whether a change in score comes from changed evidence, changed inputs, changed methodology, or changed weighting.

Use precise states such as Measured, Estimated, Unavailable, Inconclusive, and Not Scorable. Never use a numerical score to create an appearance of certainty that the evidence does not support.

No evidence + no valid measurement method = no score.