Skip to main content
Search results are ranked on three axes, combined into a single score.

​
The axes

  1. Textual relevance. BM25 against describe.title, describe.description, and describe.tags. Titles weigh heaviest.
  2. Price. Cheaper listings rank higher within a relevance bucket, subject to a floor so paying $0 does not automatically win. Buyers can override this with sort=price to sort strictly by price.
  3. Recent activity. Endpoints that have settled successfully in the last 24 hours rank higher. New listings get a small “fresh” boost for 48 hours to avoid a cold-start problem.

​
What is not a signal

  • Editorial ranking. There is no allowlist and no manual boosting.
  • Payment to the operator. The Bazaar does not sell placement.
  • Historical volume beyond the 7-day window. Old listings do not benefit from ancient traffic.

​
Query examples

​
Pagination

All queries use cursor-based pagination:
Pass ?cursor=<value> to the next request. Do not construct cursors yourself — they are opaque.

​
Debugging rank

Append ?explain=1 to any search request. The response gains a debug block per result showing the contribution of each axis. Useful for sellers trying to understand why their endpoint sits below a competitor.