Change failure rate

Change failure rate is the percentage of deployments to production that result in a degraded service requiring remediation, such as a rollback, hotfix, or incident, calculated as failed changes divided by total changes over a period. It is one of the four DORA metrics and the primary stability counterweight to deployment frequency and lead time for changes.

Change failure rate differs from simply counting incidents in that it is normalized against deployment volume, so a team deploying far more often is judged on the proportion of failures, not the raw count, avoiding an unfair penalty for shipping frequently. What counts as a "failure" varies by organization, some count only incidents severe enough to page someone, others count any rollback, so comparing this metric across organizations requires checking the underlying definition.

Change failure rate matters because it is the clearest signal that a team's release process is not silently trading speed for reliability, and it is typically tracked alongside mean time to resolve to judge how quickly failures, once they occur, get fixed. A common pitfall is defining "failure" too narrowly, which understates real instability, or so broadly that unrelated incidents get misattributed to a recent change.