Methodology

Four axes

• Reputation — what current public signals say after sample-size shrinkage and conflict handling.
• Evidence confidence — how much weight a decision-maker can put on that picture.
• Depth — how many independent sources and evidence classes exist.
• Freshness — how old the facts that actually drive the number are.

Evidence classes

PUBLIC, DECLARED, OBSERVED, VERIFIED, INFERRED, and DISPUTED are stored separately. A professional-looking claim does not become verified by style. Inference is named. Disputes are retained; they are not deleted to “clean” a profile.

Insufficient evidence

Identity without ratings is a valid profile. Assessment is withheld. That is not a low score, and it must not lose a comparison as if it were one.

Paying does not buy a score

A future commercial relationship may add OBSERVED or VERIFIED evidence (work actually seen). It must never add a premium constant, hide negative public evidence, or collapse a weak branch into a strong parent because someone subscribes.

Corrections

Factual errors should be reportable against a specific entity and fact, with a public source. The corrections route is the UX foundation. Case management is not yet active.

What this page will not specify

Exact thresholds, blocking keys, and anti-gaming heuristics that would help an attacker industrialise false evidence are withheld. The estimator identifier, axes, explanation contract, and paid-boost ban are public.