Proof score (discovery)
RANKING score computed from the already-loaded fields of a listing, so a result list can be ordered without one store read per entity. It shares the tier scale of the other scores but not their inputs: for the same listing it may differ from the transparency score. The list badge is therefore labelled « proof », never « transparency ».
Not to be confused with
- Trust Index
Score from 0 to 100 qualifying an ACTOR, aggregated from its verifiable proofs: anchored attestations 30 %, certifications 27 %, HACCP audits 18 %, carbon transparency 15 %, B2B endorsements 10 %. Recomputed on demand — freshness is what makes it credible. Tiers: platinum ≥ 85, gold ≥ 70, silver ≥ 50, bronze ≥ 25, otherwise unrated.
- Transparency score
Score from 0 to 100 measuring what a producer DISCLOSES, not how good they are: profile completeness 20 %, lots signed over twelve months 30 %, valid certifications 30 %, recency of activity 20 %. An impeccable but silent producer scores low — by design, since the measure is about disclosure.
- Trust capital
Per-owner aggregate of the signed attestations they have accumulated: a volume, not a grade. Where the Trust Index weights and normalises to a hundred, trust capital counts what has been produced and surfaces it inside its owner’s workspace.
What the blog says
- No One Will Read the Label
Herbert Simon, bounded rationality: four seconds facing thirty options. The label is a rational shortcut — but it still has to be decompressible.
Where it is implemented
lib/discovery/rank.ts