How We Score Deals

Scoring summarizes evidence for review; it is not an automatic endorsement.

Price-change context

The model compares the current observation with earlier observations for the same retailer-specific offer. Absolute savings and percentage change can both matter, but a high percentage receives less confidence when history is sparse or the comparison price is unclear.

Freshness and availability

Recent observations receive more weight than stale ones. Availability, observation time, and a freshness window help reviewers distinguish a current candidate from an old change. A score should fall or the deal should expire when evidence is no longer current.

Quality and penalties

Complete product identity, currency consistency, retailer context, seller information, condition, and shipping details improve confidence. Missing fields, reconciliation concerns, stale data, unusual comparison prices, or uncertain availability can create penalties visible to reviewers.

Human review and ranking

The score helps prioritize work but does not replace editorial judgment. Approved offers are ranked for customer value, with likely buyer cost first. Commission information may only break close ties and cannot compensate for worse cost, stale evidence, or lower quality.