Engineering Notes
Search Is Different When Devices Are Imperfect
Primary retail search assumes clean catalogs. Secondary markets require intent understanding, condition awareness, and trust-aware ranking.
Search engines built for new-device retail optimise for brand, model, and promotional inventory. Secondary markets add condition language, negotiation intent, regional slang, and incomplete attributes. Ranking that ignores those realities returns technically relevant results that still fail the buyer.
Design goals for PhoneMark AI Search
- Interpret device intent across languages and informal naming.
- Surface condition and verification context alongside relevance.
- Avoid promoting high-engagement listings that are structurally untrustworthy.
- Degrade gracefully when catalog data is incomplete.
AI ranking here is not about maximising clicks at all costs. It is about decision quality: helping participants find devices they can evaluate fairly. That objective changes how we measure success.
Evaluation
We evaluate search with offline relevance judgements, trust-sensitive metrics, and production feedback from Launch Market #001. Public write-ups will summarise methodology without exposing ranking features that invite gaming.