1. Match the exact product
We normalize retailer identifiers such as SKU, model, UPC, ASIN and TCIN, then check variants and pack sizes. Similar-looking products are not treated as identical.
2. Establish a realistic street price
Manufacturer list price alone can exaggerate a discount. The comparison price should reflect recent, credible selling prices for the same product and configuration.
3. Measure local relevance
Store-specific markdowns are tied to a location and search radius. A reported price in another region is a lead—not proof of a nearby deal.
4. Separate signals from facts
Public chatter can identify an early lead. It remains separate from verified offers until identifiers, time, location and an approved retailer or in-store observation corroborate it.
5. Apply freshness and risk gates
Older observations, uncertain inventory, mismatched variants and checkout failures reduce confidence. Expired or contradicted records are removed from verified results.
6. Rank value before compensation
Deal scores may consider discount quality, distance, evidence confidence, availability and risk. Affiliate commission is excluded from the ranking formula.