COEN Scoring

Coverage — per kind iCoverage — for each kind of governed artifact enrolled in ScoringScope: how many of the population have a current quality score, out of how many exist. Computed against the governed ScoringScope only, never raw counts (crew-pr-02 v2). Over 100% means the instrument scored subjects outside the counted universe — that lands as a Finding, not a denominator fudge.

Outlooks iQualityOutlook — CABE's credit-rating-style verdict per artifact, derived from per-dimension beliefs. healthy/stable: evidence supports quality. developing: young or thin evidence, not a defect. watch: a dimension fell below its activation threshold. negative: aggregate below floor or an outcome-linked dimension below retraction. incomplete: more than half the dimensions are weakly known — the honest "we can't say yet". Outlooks are first-class beliefs with their own evidence and decay, not computed labels.

Sensor iThe sensor family (canon sf-06..08) — how CABE learns whether its own scores are right. Every score emits an append-immutable Prediction with a resolution window; structural dimensions resolve by re-inspection at window close, functional ones by real-world outcomes. The outcome matcher links outcomes back to the predictions that should have anticipated them — until it's built, the functional tier is a governed gap, shown as PENDING rather than hidden.

Belief state (four outputs) icanon-cabe-bs-01 — the engine's target shape: per artifact it maintains structural completeness, functional effectiveness, failure attribution, and uncertainty; the score you see is only a projection of that state. The structural tier is live; the other outputs render here when the belief-state chain lands.

Runs iScoringRun — every scoring operation is a governed run: event-based snapshot cursor, locked policy versions, governance snapshot (canon sf-01). Reproducible by construction — the same run inputs give the same scores.