The incident process
Five phases, one target for Q4 — cut time-to-mitigate in half.
Assumptions
Five phases carry every incident from alert to lesson
Every Sev1 and Sev2 runs the same five phases in the same order. Ownership changes at each boundary: the on-call engineer owns Detect and Triage, the incident commander owns Mitigate, the service team owns Resolve and Review. The customer is affected from the first alert until Mitigate ends.
Incident lifecycle
Mitigate holds most of the customer-facing minutes
Customers feel an incident from Detect through Mitigate. Of the median 97 minutes in that window, 68 sit in Mitigate. Detect and Triage together take 29.
Resolve and Review are longer, but the customer no longer sees them. The largest cut to customer-facing time comes from the phase with the most minutes.
Median minutes per phase, Sev1 and Sev2, Apr–Sep 2026
The Q4 target is a 34-minute Mitigate
The two known causes are a runbook that exists for 41% of paged services and a rollback that needs a manual approval outside business hours. Q4 closes both: a runbook per paged service, and self-serve rollback for the commander.
Down from 68 min. Measured on Sev1 and Sev2 incidents from October to December 2026.
What to remember
- 1Five phases, unchanged
Detect, Triage, Mitigate, Resolve, Review. The order and the owners stay as they are.
- 2Q4 improves Mitigate only
It holds 68 of the 97 customer-facing minutes.
- 3The target is 34 minutes
A runbook for every paged service and self-serve rollback for the commander.