Mean time to restore
Also known as MTTR, Mean time to recovery
Definition
Mean time to restore is the average elapsed time needed to restore a service after an incident or failure. The result depends on which events count, how recovery is defined, and whether the distribution is summarized by a mean or another statistic.
Name the failure and the restore event
MTTR is used for several related measurements. One team may start when monitoring pages an engineer and stop when the service is usable. Another may start when a ticket is opened and stop when the postmortem closes. The same acronym can therefore hide different clocks.
An illustrative service recovers in 20 minutes from one incident and in 180 minutes from another. Its mean recovery time is 100 minutes. That average is easy to calculate, but it conceals the long incident that affected users for three hours.
Use a distribution and a clear boundary
Report the number of incidents, median or percentiles, and the events used to start and stop the clock. Separate deployment-related failures from dependency outages when the team wants to understand delivery stability.
Current DORA guidance uses failed deployment recovery time for a narrower delivery measure. Do not silently replace an older MTTR dashboard with that metric. Keep the names distinct and document the relationship.
Recovery is a system capability
The duration can reflect detection, diagnosis, rollback, communication, and validation. A fast rollback may reduce customer impact even when the permanent fix takes longer. Inspect the path through the incident and choose an improvement that addresses the slowest or riskiest step.
How this relates to Weave
Weave connects code changes and quality signals to incidents, helping teams investigate what preceded a recovery event. Keep a broad incident MTTR separate from DORA's failed deployment recovery time when comparing reports, because their scopes can differ.
Explore Engineering intelligence