MTTR is the wrong metric for Engineering Teams today
10 hours ago
- MTTR (Mean Time to Recovery) is becoming obsolete for AI-era engineering teams due to increased bug rates and pull request volumes.
- AI tools cause 41% more bugs and 98% more pull requests, making MTTR inadequate for measuring failure prevention.
- MTTF (Mean Time to Failure) shifts focus from incident response to prevention, emphasizing customer experience as the key metric.
- Goodhart's Law highlights the danger of metrics like MTTR becoming targets, leading to harmful practices even with honest pursuit.
- Code review is a bottleneck as AI accelerates code output, with a 10x increase in code requiring a similar increase in review capacity.
- The blast radius of bad code is growing with AI agents autonomously generating more code, potentially causing inscrutable failures.
- MTTF requires rethinking workflows, combining static analysis, runtime analysis, human review, and tests to prevent customer-impacting failures.
- Prevention compounds over time, while recovery resets teams to day zero, making MTTF a more sustainable metric for quality.