How would you test a search feature for a marketplace site? List your test scenarios.

Search is multi-dimensional and most candidates only test the first dimension. A structured walkthrough:

Relevance / correctness: exact match, typo tolerance ("iphne" still finds iPhone), stemming, synonyms, case insensitivity, whitespace handling.

Input edge cases: empty query, single space, single character, very long query (1000+ chars); special characters, SQL injection patterns, XSS, unicode (Cyrillic, CJK, emoji); boolean operators if supported (AND, OR, quoted phrases).

Filters / facets: single filter, combined filters, filters with no matching results; filter persistence across pagination; price range, category, location radius, condition; filter UI (chips removable, count badges accurate).

Sorting: each sort option works (relevance, price low-high, price high-low, newest), persists across pagination, stable under ties.

Pagination: page navigation forward/back, page size, deep pagination (page 50 of 50), URL state preserved (you can bookmark page 3).

No-results UX: helpful empty state with suggestions, "did you mean", related searches.

Indexing: newly created listing appears in search within SLA (10s? 1m?), updated listing reflects updates, deleted listing disappears.

Performance: latency under typical load (p95), auto-suggest latency (<100ms perceived), behaviour at high concurrency.

Personalisation (if applicable): logged-out vs logged-in result differences, location-based ranking, recently-viewed influence.

In an interview I'd flag the questions I'd ask the PM: what's the relevance algorithm (BM25? a custom ranker?), what's the indexing latency target, is search personalised, what's the no-results SLA.