Validate recorded policy evaluations

Create a validation plan

Management-group assignments can be selected, but a run queries only the chosen subscription or resource-group scope. Exempt resources do not satisfy the minimum assessed-resource count.

1. Open Azure → Policy Validation and choose New plan.
2. Choose a subscription, an optional resource group, and up to ten policy assignments.
3. Set the maximum evaluation age (1–168 hours), minimum assessed resources, and whether exempt evaluations are allowed.
4. Add expected resource IDs when specific resources must be represented, then save the plan.

Run and inspect checks

Choose Run checks and keep the section open while it runs. The fixed sequence collects existing evaluations, checks your expectations, and saves the results. Cancel aborts the active request. Recent runs opens the saved plan snapshot and evidence; JSON export downloads the same retained evidence sample.

Editing a plan increments its revision. Historical runs retain the original expectations. Deleting a plan preserves its run history.

Interpret the result

• Passed: returned evaluations are fresh, resolved, and meet the saved coverage expectations.
• Failed: complete, fresh evidence includes noncompliance or disallowed exemptions.
• Inconclusive: stale or missing data, unresolved states, inadequate coverage, access errors, or query limits prevent a conclusion.
• Cancelled: the request was cancelled or the client disconnected.
• Interrupted: execution started but no final result was saved within three minutes.

What a passing result means

The run reads Azure’s recorded evaluations. It does not trigger a compliance scan, deploy a test resource, or prove that a policy will deny a future deployment. A pass concerns the returned evaluations and your explicit expectations, not every resource that could exist in the tenant.

Current limits

• Up to 10 assignments, 100 pages, and 10,000 evaluation rows per assignment; a fixed seven-day query window and two-minute request limit.
• Up to 200 evidence rows retained per assignment, prioritizing problematic rows. Counts cover all retrieved, deduplicated evaluations.
• The latest 100 run summaries appear in the history view; older records remain stored. Automatic retention is not implemented.
• Scheduling, active enforcement tests, Robot execution, and workflow-graph integration are not implemented.