Change failure rate
Change failure rate (CFR) is the percentage of deployments that cause a degradation in service requiring remediation (a hotfix, rollback, or patch). It is the fourth DORA metric and the noisiest of the four — published year-over-year DORA data shows the largest annual variance on CFR, particularly for non-elite teams. DORA 2024 thresholds: elite ≤5%; high 5–10%; medium 10–15%; low ≥15%.
CFR is hard to measure cleanly because the definition of 'change failure' is operationalised differently across organisations: some count any incident, some count only customer-impacting outages, some count only rollbacks. The DORA survey question wording asks respondents to self-classify — making cross-org comparisons informative directionally but unreliable absolutely. The most interesting trend in recent DORA data: CFR for elite-quartile teams has held steady at ~5% across years, while the spread between elite and low quartiles has widened. The popular interpretation that 'AI raises change failure rate' is not supported by DORA's published data at the elite quartile but does appear (with explicit non-causal framing) in the lower three quartiles in 2024.
Related terms
- Four key metrics
The four key metrics are the DORA framework for measuring software-delivery performance: deployment frequency, lead time for changes, mean time to restore, and change failure rate.
- Deployment frequency
Deployment frequency is the rate at which a team deploys to production.
- MTTR
Mean Time To Recovery is the average elapsed time between an incident's detection and its resolution.