Work out mean time to restore instantly with clear inputs, formula shown and shareable results.
Mean time to restore is total downtime divided by incident count, and it is one of the four DORA metrics. Separating detection from repair matters because they have different fixes: detection is monitoring and alerting, repair is runbooks, rollback automation and deployment speed. The DORA elite band is restoring service in under an hour.
MTTR and availability
MTTR = total downtime / incidents; repair time = MTTR - detection time; availability = 1 - downtime / period
It is skewed by one catastrophic outage. Report the median and a high percentile alongside it, since those describe the typical and worst experiences better.
Automated rollback and the ability to deploy quickly. Teams that deploy in minutes can revert in minutes, which is why deployment frequency and MTTR correlate so strongly.