Investigate a failed run
A failed run tells you a check did not complete as expected. Start with what the browser observed, then decide what needs to change. This guide applies to saved background runs; the agent's Live View case result is reviewed inside its card.
- LocateOpen the failed case within the Run.
- InspectRead the reason and watch available evidence.
- ClassifyIdentify product, case, data, or environment causes.
- VerifyChange the cause, then run the check again.
1. Open the failing case
Open Runs, choose the failed run, and inspect its individual case results. For a plan run, identify which cases failed and which actually executed. Check the trigger and timing so you are investigating the intended execution.
Read the failed reason, including any reported step number. Open a linked Issue if one is available. A failed background case with an issue report can create or link an Issue automatically; check before adding another report.
2. Compare the evidence with the expected result
Open the case recording when available. Compare the observed UI with the case's actions, precondition, and expected result. Ask whether the browser reached the intended page, signed in to the right workspace, and used valid data.
If video is still Processing, return later or refresh. The test result and recording processing finish separately. A missing recording does not change the saved test status. See Run evidence.
3. Choose the follow-up
| What the evidence shows | Likely cause | Next action |
|---|---|---|
| The intended workflow is reached, but behavior contradicts the expected result | Product problem | Review or save the Issue with reproduction details and investigate the product. |
| The UI or intended behavior changed and the case describes the old flow | Stale instructions | Update the case's actions and expected results to the current requirement. |
| Login fails, access is missing, or a required record is unavailable | Account or test data | Correct the account, role, tenant, or prepared data. |
| The application or a dependency is unavailable | Environment problem | Restore the environment or dependency before another run. |
These are investigation paths, not automatic diagnoses. A login failure, for example, may itself be a product regression if the credentials and environment are correct.
4. Verify the change
After correcting the cause, run the affected case or plan again and review the new result. Reruns consume credits and use current case settings; plan reruns also resolve the plan's current scope. They do not replay an immutable historical configuration.
If you resolve an Issue, write a reason that explains what changed. Resolution does not start another run or prove a fix; use the new execution evidence to verify the behavior.
Check your progress
You can explain the failure using evidence, identify what changed, and point to a new result that verifies the affected workflow.Keep the next result useful
Prefer a clear expected result such as “The invited email appears as Pending” over “The invitation works.” Give each case an independent starting point and maintain its account and data. See test case maintenance and Issue triage.