What is MTTR?
MTTR most often means mean time to recovery (or restore): the average time from the start of an incident until the service works again, calculated as total recovery time divided by the number of incidents. The same letters are also used for repair, respond and resolve, so always say which one you report.
Why MTTR matters
MTTR measures how much each failure costs users. Because failures are inevitable, cutting recovery time is often the cheapest reliability improvement: faster detection, clear runbooks and quick rollbacks all shorten it.
Recovery time is also one of the four DORA software delivery metrics (time to restore service, renamed failed deployment recovery time in recent reports), so it is a common way to compare how teams operate.
How to measure MTTR
MTTR = Σ (restored − started) ÷ number of incidents
Worked example
Last month four incidents took 12, 45, 8 and 35 minutes from start to recovery.
- Total recovery time = 12 + 45 + 8 + 35 = 100 minutes.
- MTTR = 100 ÷ 4 = 25 minutes.
- Sorted, the times are 8, 12, 35 and 45, so the median is (12 + 35) ÷ 2 = 23.5 minutes.
Result: With only four incidents, report the count and the median next to MTTR, and look at the 45-minute outlier on its own: one slow diagnosis moves the whole month.
Common mistakes
- Not defining the R. Recovery, repair, respond and resolve give very different numbers for the same incidents.
- Averages over tiny samples. One long outage dominates a monthly MTTR. Show the count and the median.
- Starting the clock at the alert. Users were affected before the alert; leaving out detection time flatters MTTR.
- Gaming it. Closing incidents early or splitting one outage into several lowers MTTR without helping anyone.
How Vigil handles MTTR
- Vigil's incident timeline records when each incident started, was acknowledged and was resolved, and shows MTTR from those times; a dashboard tile shows MTTR over the last 30 days.
- Faster recovery comes from evidence: failed checks keep the status, headers, timings and first 4 KB of the body, and Time Detective shows deploys, errors, logs and traces around the start time.
- Escalation and on-call paging cut the time before someone starts working on it.
MTTR (mean time to recovery): frequently asked questions
What does the R in MTTR stand for?
Recovery or restore in most SRE writing, but also repair, respond or resolve. Recovery is when users are fine again; resolve can include follow-up work, so it is longer.
What is a good MTTR?
It depends on the system and the incidents. Compare it with your own history and SLO: if your monthly error budget is 43 minutes, an MTTR near an hour means one incident breaks the SLO.
How is MTTR different from MTTD?
MTTD covers only the time until the problem is noticed; MTTR measured from the start covers detection, diagnosis, fix and verification.