MTTR (Mean Time to Repair)

MTTR is a reliability measure, but the letters do not always expand the same way. Teams use them for mean time to repair, recover, resolve, or restore; each version can start and stop the clock differently. A useful report therefore defines the event population, start timestamp, endpoint, and exclusions before presenting an average.

For mean time to repair, divide the total repair time across qualifying failures by the number of those failures. If three outages require 20, 40, and 60 minutes of repair work, MTTR is 40 minutes. That does not mean every outage took 40 minutes or that customers were affected for the same duration. Delays in detection, waiting for approvals, and partial workarounds can change a customer's experience without changing a narrowly defined repair clock.

A local business might monitor how quickly a booking system or call-routing integration is repaired. It should also track the longest outage, number of incidents, and impact on missed appointments. Google SRE research cautions that simple averages can be poor guides to production incident decisions. MTTR is one descriptive measure, not a complete service-level objective or proof of improving reliability.