DevOps Performance From a Different Dataset: What 30M Workflows Reveal
Talkat KubeCon + CloudNativeCon North America 2020 Virtual · on Wednesday, 2020-11-18· at 15:50· Intrado Virtual Event Platform,·
summary
CircleCI analyzes 30M+ builds from 44K orgs and 160K projects and reports concrete metrics: 50% finish under 4 minutes, average pipeline duration ~25 minutes, mean success 54% / median 61%, and mean recovery ~14.85 hours (95th percentile ~3.4 days). Multicontributor projects recover and succeed faster (benefit plateaus near ~200 contributors); single-developer projects show the worst duration and recovery. Default/main branches have higher success and shorter runs than feature branches, weekend throughput drops ~70%, and March–April 2020 shows a pandemic-linked uptick in activity and success.
AI-written summary · about
chapters (12)
- 0:00Start
- 2:55Look at What's Changed Year over Year
- 4:19Deployment Frequency
- 10:57Average Time for a Pipeline
- 11:26Success Rate
- 13:12Recovery Time
- 15:47Success Rate Does Not Correlate with Company Size
- 17:23Longest Recovery Times
- 18:17Software Is Collaborative
- 21:28Throughput
- 22:11Success Rate at the 50th Percentile
- 33:20Final Thoughts
related talks
06mDevops Behavioral Economics: Displayed...
25mContinuous Delivery, high-performing engineering teams, and the Holy Grail
46mObservability for Engineering Teams
47mHow to win in software development with tighter feedback loops
24mUnlocking CI Performance in the Age of Agentic Development
39mSPACE and Beyond - Measuring Developer Productivity
33mMeasuring Developer Efficiency in DevOps: Key Questions, Metrics, Best Practices
46mDeveloper Productivity at Scale