In 30 seconds
- Failed changes divided by total changes.
- One of the four original DORA metrics.
- Measures stability, balancing speed metrics.
How to calculate it
Change failure rate = changes causing failure / total changes deployed x 100
A failure is anything that degrades service and needs remediation: rollback, hotfix or patch.
Why it matters
Speed metrics like deployment frequency could be gamed by shipping recklessly. Change failure rate keeps speed honest. It's one of the DORA metrics.
How to improve it
- Smaller changes, deployed more often
- Stronger automated tests
- Gradual rollouts with feature flags
- Blameless reviews of each failure
Myth vs reality
Myth · Zero failures is the goal.
Reality · Very low rates can mean the team is too cautious.
Myth · Deploying more often raises failure rates.
Reality · Smaller changes usually lower them.
Myth · Only outages count.
Reality · Any change needing remediation counts.
FAQ
What's a good rate?
DORA research finds top performers at low single to low double-digit percentages; trends matter more.
Do failed deploys caught before production count?
No; it measures failures in production.
How often should we review it?
Monthly or quarterly trends are enough.