Software delivery teams often report recovery time from the first page ping to the “all clear” message. That interval is useful, but it hides whether the fix was a rollback, a forward patch, or a feature flag flip — each with different implications for pipeline design.
For CI/CD workflow analytics, tag recoveries by mechanism and by who initiated them. Automated rollback that completes in four minutes is a different story from a manual hotfix that took the same wall-clock time after forty minutes of diagnosis.
Agree on the start marker before the next incident: customer report, synthetic check, or deploy job failure. Mixing those starts across quarters makes trend lines decorative. Keep the definition dull and stable; let the post-mortem carry the narrative colour.
If your team ships from Busan evening windows into global traffic, also record timezone of detection. Overnight detections without on-call coverage inflate recovery without teaching you anything about the pipeline itself.