Beyond matching
Finding products that match a query is the easy half. The hard half is ranking the matches so the most relevant and purchasable items appear first.
Signals that drive rank
- Text relevance: how well the title and attributes match the query.
- Popularity: clicks, conversions, and recent sales.
- Availability: in stock items should outrank out of stock ones.
- Business rules: margin, sponsored placement, and freshness.
Putting it together
- Retrieve candidates fast with an inverted index, then re rank the top set with richer signals.
- Blend signals into a score and sort, or learn weights from past behavior.
- Watch for feedback loops where popular items get more exposure and stay popular.
Key idea
Catalog search first retrieves matching candidates quickly, then re ranks the top set by blending relevance, popularity, availability, and business rules into a single ordering.