Result changes and failure groups¶
This report shows outcome changes and equal failure texts from visible completed runs. Open it from the report catalogue or the corresponding test planning entry. Windows opens the same Web report, with the same sign-in and project permissions. No AI configuration is required.
You need access to test planning and the included runs and planned cases. Hidden results do not contribute to counts or groups. If an original library case is hidden, its case and variant IDs remain hidden; the permitted planned snapshot supplies the displayed title.
Selection and coverage¶
Optionally select a visible test plan and environment and filter by UTC run completion time. “Unknown” selects runs without a known environment. The run limit selects the latest matching runs: 50 by default and 200 at most. The report includes at most 20,000 latest results. Coverage counts and a warning identify limited selections. “All” includes all visible matching data. Changing the test plan resets the environment selection.
Only the latest result per planned case, variant and run counts. Series separate original case, original variant, environment, origin and external test identity. Unknown identities are not combined into comparison series. Equal display names do not prove equal identity.
Reading outcome series¶
An outcome change means a change between success and failure in adjacent conclusive run observations. Partial or untested outcomes and ambiguous duplicate variants break the sequence. Repeated failures count distinct failed runs. Every evidence entry links to its run and includes the result ID.
The report distinguishes verified automation imports, other or unverified origin, and changed or invalid import evidence. Missing import evidence does not prove manual execution. Different build identifiers or saved descriptions are disclosed. Content comparison covers the saved description only, not the complete historical specification.
These observations do not prove flaky tests. Inspect builds, specifications, test data and environments using the evidence. A recorded import time does not prove the CI execution time.
Reading failure groups¶
Groups require exactly equal text after removing HTML and normalizing whitespace. Case, error codes, numbers and timestamps remain intact. Origin, source and environment remain separate; missing text is not grouped. Equal text does not confirm a shared root cause. The distinct run count distinguishes multiple results within one run from recurrence across runs.
MCP and REST¶
Read-only access uses the same analysis and permission checks as the Web report. Find the operation in the quality reports domain of the MCP catalogue. HTTP/JSON exposes the same operation at /api/v1/operations/{operationId}; /api/v1/operations and /api/v1/openapi.json describe the contracts available to your key. Reading the report does not change results, approvals or planning.