Search Index Audit

Goal and scope

Audit one search, full-text, faceted, autocomplete, permissions-filtered, or indexing boundary for relevance, freshness, authorization, consistency, and failure behavior.

Domain invariants

• Indexed documents reflect authoritative data with known freshness and deletion/update semantics.
• Search results enforce tenant, role, ownership, plan, privacy, and visibility rules at the right layer.
• Filters, facets, sorting, pagination, ranking, tokenization, language handling, and empty states match product expectations.
• Reindex/backfill, partial failures, stale documents, and provider/index downtime have defined recovery behavior.
• Tests or replay fixtures prove key recall/freshness/permission cases.

Audit method

Trace selected resource from authoritative write/delete through indexing pipeline, document schema, query builder, filters, result hydration, UI/API state, analytics, and tests.

Define representative exact-ID/title, ordinary task, alias/language and ambiguous queries with expected relevant and misleading documents before judging ranking. Inspect normalization, candidate generation, scoring, explicit filter intersection and hydration separately. Retain the failing query, source/index identity, expected ordering/window and consumer impact for Improve Search Index. Check protected snippets/facets/counts as well as visible result rows where access matters.