Mean time to recovery (MTTR) is one of the most quoted operations metrics, and one of the easiest to get wrong. This note looks at how to measure it in a way that actually helps.
Define the clock
Agree on exactly when the clock starts and stops:
- Start: when the impact began — not when someone noticed it.
- Stop: when users are no longer affected — not when the ticket was closed.
Measuring from detection instead of impact makes your numbers look better and hides slow detection.
Don’t report only the mean
A handful of long incidents can dominate the average, or a flood of tiny ones can hide them. Report alongside the mean:
- The median, to show a typical incident.
- The longest incident in the period, with a short summary.
- The count of incidents, so trends in frequency are visible.
Use it to ask better questions
MTTR is most useful when it prompts questions: Why did detection take 40 minutes? Why did we need three teams to restart one service? The number is the start of the conversation, not the conclusion.
