Search & Ranking Interview Questions
Search and ranking questions cover indexing, retrieval then ranking as separate stages, and the matching systems behind marketplaces and recommendations.
Search systems separate cheaply narrowing a large corpus from expensively ordering a small candidate set. That two-stage shape — retrieve, then rank — recurs across web search, product search, recommendations and marketplace matching, and recognising it is most of what a design answer needs.
What interviewers probe
The structure interviewers look for is exactly that separation, with an inverted index or a vector index doing retrieval and a more expensive model ordering the survivors. Follow-ups concern index freshness — how quickly a new document becomes findable — and how ranking quality is measured, which is where change data capture and evaluation come into the conversation.
Practice search & ranking
6 system design · 0/2/4 easy/medium/hard
Search & Ranking interview questions — FAQ
Why separate retrieval from ranking?
Because ranking is too expensive to run over an entire corpus. Retrieval narrows millions of candidates to hundreds with a cheap index; ranking then spends real computation ordering only those.
How is a search index kept fresh?
Usually by streaming changes from the source database rather than periodically rebuilding. Change data capture reads the replication log, which avoids the dual-write problem where the database and the index silently disagree.