Find where database time goes, then measure whether a change helps.
Set up a scheduled agent that runs weekly and finds database queries worth optimising. Use the last seven days of query statistics and application traces, and map the queries to the code that calls them.
Rank candidates by total database time and call volume. Look for repeated queries, unnecessary data reads and missing indexes supported by query plans. Skip issues already tracked.
For the strongest candidate, report its share of database time, code references, evidence and one proposed change. Benchmark it locally with representative data and check that results stay the same. Separate measured improvements from estimates. If evidence is missing, say what you need. Don't change production.
Ask me for the repo, read-only query data, local benchmark setup, issue tracker, report destination and schedule, including timezone, before setting anything up.