Query Optimization Skill

Phase 1 — Discovery

Ask only what context doesn't reveal:

Database engine and version? Planner behavior, available hints, and index types differ significantly. Postgres 14+ has improved parallel query and CTE materialization defaults vs. older versions.
Do you have EXPLAIN ANALYZE output? Without it, optimization is guesswork. Ask them to run it and share — not just EXPLAIN (estimates only), but EXPLAIN (ANALYZE, BUFFERS) which shows actual row counts and I/O.
Table row counts for involved tables? The planner's choices make sense or don't in context of data volume.
Is this a one-time slow query or a recurring endpoint? One-time queries tolerate different tradeoffs (parallel workers, temp indexes) vs. a hot path that runs thousands of times per minute.
Are statistics up to date? Ask if ANALYZE has been run recently on large tables. Stale statistics cause the planner to make systematically wrong choices.