Know who can query. Understand what it costs.
Governance, performance, and cost meet at the workload. Establish the access boundary, observe behavior, then test the smallest useful change.
Make access boundaries explicit.
Use role design and least privilege as the foundation. Evaluate masking policies and row access policies against the sensitivity of the data and the needs of each consumer.
- Document role ownership, grant paths, service identities, and review cadence.
- Test policy behavior with representative roles, including negative access tests.
- Confirm edition requirements; a policy mechanism does not establish compliance.
Diagnose before adding compute.
Inspect query profiles, concurrency, queueing, pruning, and workload isolation. Establish a reproducible baseline before testing warehouse sizing or an optimization service.
- Separate query shape problems from contention and capacity.
- Compare runtime and consumption under equivalent test conditions.
- Evaluate clustering and search optimization selectively; include maintenance cost.
Build a complete consumption picture.
Attribute warehouse, serverless, storage, and transfer consumption separately. Resource monitors apply to warehouses; they do not track serverless features or AI services. Use budgets for monitoring those categories.
- Review auto-suspend, warehouse responsibilities, and idle patterns.
- Assign owners for budget notifications and operational responses.
- Check each control’s coverage rather than treating every alert as a hard account-wide spending cap.
Start with the workload. Build a clearer plan.
Tell us what needs to change: a legacy warehouse, unreliable pipelines, access boundaries, or an unexplained consumption pattern.
