DORA (DevOps Research and Assessment) identified four metrics that predict software delivery performance and organizational outcomes. They're the closest thing we have to objective engineering metrics. > "Elite performers deploy 208 times more frequently, have 106 times faster lead time from commit to deploy... and have 7 times lower change failure rates." —Nicole Forsgren, Accelerate ## The Four Metrics | Metric | What It Measures | Elite | Poor | | --------------------- | --------------------------- | ------------ | ---------- | | Deployment Frequency | How often you ship | Multiple/day | < Monthly | | Lead Time for Changes | Commit to production | < 1 hour | > 6 months | | Change Failure Rate | % of deploys causing issues | < 5% | > 45% | | Time to Restore | Recovery from incidents | < 1 hour | > 1 week | ## Why These Metrics Matter > "Improvement of daily work is even more important than doing daily work." —Gene Kim, The Phoenix Project Teams that score well on DORA metrics: - Ship 46x more frequently - Are 2,555x faster from commit to deploy - Have 7x lower change failure rate - Recover 2,604x faster from incidents This isn't correlation—it's predictive. Improving these metrics improves outcomes. ## Measuring DORA Metrics I want to measure DORA metrics for my team. Current setup: - Source control: [GitHub/GitLab] - CI/CD: [GitHub Actions/etc.] - Deployment: [Vercel/etc.] - Incident tracking: [Linear/etc.] Help me design a measurement approach: 1. DEPLOYMENT FREQUENCY: - What counts as a deployment? - How to track (automated vs. manual)? - What's our current estimate? 2. LEAD TIME: - From first commit or PR open? - To staging or production? - How to measure automatically? 3. CHANGE FAILURE RATE: - What counts as a failure? - How to attribute to specific deploy? - How to track rollbacks? 4. TIME TO RESTORE: - What counts as "restored"? - How to track incident duration? - Include off-hours? For each metric, suggest: - How to track it automatically - Target for our team size - Leading indicators to watch ## Improving Each Metric Improving DORA metrics often starts with improving [Developer Experience](/chapter/developer-experience)—fast feedback loops, frictionless environments, and reduced cognitive load directly impact Lead Time and Deployment Frequency. **Deployment Frequency** | Blocker | Solution | | ----------------------- | ------------------------------------------- | | Long test suites | Parallelize, prioritize, delete flaky tests | | Manual QA | Automate critical paths | | Fear of breaking things | Better tests, feature flags, fast rollback | | Batch releases | Deploy smaller changes more often | **Lead Time** | Blocker | Solution | | --------------------- | ------------------------------- | | Slow code review | Smaller PRs, async review, SLAs | | Slow CI | Cache, parallelize, trim fat | | Manual deployment | Fully automate | | Long feature branches | Trunk-based development | **Change Failure Rate** | Blocker | Solution | | ---------------------- | -------------------------------------- | | Bugs in production | Better testing, especially integration | | Deploy the wrong thing | Preview environments, better QA | | Config errors | Config as code, review config changes | | Dependency issues | Lock versions, test upgrades | **Time to Restore** | Blocker | Solution | | ------------------- | ------------------------------------- | | Can't detect issues | Monitoring, alerting, health checks | | Slow rollback | One-click rollback, immutable deploys | | Can't diagnose | Better logging, observability | | No runbooks | Document incident response | ## Common Mistakes **Mistake 1: Gaming the metrics** Deploying empty commits to improve frequency helps nobody. The metrics indicate, not define, good performance. **Mistake 2: Measuring too precisely** Rough tracking is fine. Don't build a complex system to track metrics about shipping fast. **Mistake 3: Ignoring context** An agency with many clients has different deployment frequency than a single-product company. Compare to yourself, not others. **Mistake 4: All metrics at once** Pick one to improve at a time. They're related—improving one often improves others. ## Starting Point for Startups | Metric | Reasonable Starting Target | | -------------------- | -------------------------- | | Deployment Frequency | Daily | | Lead Time | < 1 day | | Change Failure Rate | < 15% | | Time to Restore | < 4 hours | These aren't elite numbers, but they're achievable and a huge improvement over many teams. ## Your Action Items Measure current deployment frequency (rough count is fine) Track lead time for last 10 PRs Count incidents in last month, calculate failure rate Review last 3 incidents for time to restore Identify biggest blocker for each metric Pick one metric to improve this quarter Set up basic tracking (even a spreadsheet works) Review metrics monthly with team DORA metrics aren't about being elite. They're about understanding where your bottlenecks are and systematically removing them.