The MVP computes a bounded 0-100 score from existing platform activity (published posts, delivered marketplace orders, and seller reviews). High-volume signals are capped to reduce gaming.
- `src/lib/server/reputation.ts`
This MVP proves that reputation is computed from first-party signals, capped to reduce gaming, penalized for safety signals without leaking moderation details, and surfaced on both public and in-app profile pages.
The goal is an honest, explainable reputation baseline that can be expanded later. This is not a full-blown fraud engine or marketplace guarantee.
The MVP computes a bounded 0-100 score from existing platform activity (published posts, delivered marketplace orders, and seller reviews). High-volume signals are capped to reduce gaming.
The model applies penalties for enforcement actions and risk-score presence, but the public API only returns a coarse penalty summary (no raw enforcement rows).
Public profiles show a reputation snapshot in the sidebar. Signed-in users also see it inside the app People surface.