Scoring with receipts.

Scores are computed within each category cohort from measured 24-hour signals. No provider payment, user traffic, or editorial placement changes the result.

score = 0.40 · success_24h
      + 0.25 · inverse_normalized(latency_p90)
      + 0.20 · inverse_normalized(price)
      + 0.15 · schema_completeness
      − 0.30 · price_drift_penalty
      − 0.20 · staleness_penalty