Engineering benchmarks from 802,979 merged pull requests in the GitHub Archive
802,979
Merged PRs
262,212
Repositories
191,099
Contributors
Oct 2025
Study Window
Based on data from DORA, Stack Overflow, GitHub, and proprietary research
© 2026 CodePulse • codepulsehq.com
The 2025 engineering landscape is defined by three major trends: the widespread adoption of AI tools, a persistent burnout crisis, and a widening gap between high-performing and struggling teams.
75%+
AI Adoption
of developers now use AI assistants
83%
Burnout Crisis
report experiencing burnout
4x
Performance Gap
difference between elite and low performers
Cycle time—the duration from first commit to production—is the single best predictor of engineering team performance. Here's how teams compare:
| Performance Level | Cycle Time | % of Teams |
|---|---|---|
| Elite | < 1 day | 18% |
| High | 1-3 days | 23% |
| Medium | 3-7 days | 35% |
| Low | > 7 days | 24% |
Mid-sized companies (250-499 engineers) tend to have the fastest cycle times. Smaller teams often lack process maturity, while larger organizations face coordination overhead.
Code review practices have a significant impact on both code quality and team velocity. Here's what the data shows:
Optimal PR Size
<400 lines
PRs over 400 lines have 50% lower review quality
Optimal Reviewers
2 reviewers
Diminishing returns beyond 2 reviewers
Target Wait Time
<4 hours
First review within 4 hours for elite teams
Review Coverage
80%+
Minimum for critical systems
| Wait Time | % of PRs | Impact |
|---|---|---|
| < 4 hours | 35% | Optimal - minimal context loss |
| 4-24 hours | 40% | Acceptable - some overhead |
| 1-3 days | 18% | Problematic - significant delays |
| > 3 days | 7% | Critical - blocking delivery |
of software developers report experiencing burnout
Up from 61% in 2021
There's a dangerous disconnect between how leadership perceives team health and reality:
Leaders who believe
46%
they're effectively preventing burnout
Employees who agree
34%
that burnout is being prevented
Teams of 10 or fewer engineers report 23% higher burnout rates than larger teams, often due to lack of coverage and constant on-call pressure.
AI coding assistants have become ubiquitous, with over 75% of developers now using tools like GitHub Copilot, Claude, or ChatGPT. But the impact is nuanced:
Documentation
+7.5%
AI excels at generating docs
Code Readability
+3.4%
More consistent style
Delivery Stability
-7.2%
More bugs reaching prod
Embrace AI tools for productivity, but strengthen review processes and automated testing to catch AI-introduced issues before they reach production.
"Bus factor"—how many people would need to leave before critical knowledge is lost—remains a significant risk for many organizations:
Average Bus Factor
2.3
people per critical system
Teams at Risk
40%
have bus factor of 1
Analysis across public GitHub repositories reveals concerning concentration patterns:
Benchmarks are useful for context, but remember: your goal is to improve your own team, not to match an industry average.
Answer these questions to identify your priorities:
Stop guessing where you stand. CodePulse connects to your GitHub in minutes and shows you exactly how your team compares to these benchmarks—with actionable insights to improve.
Read-only GitHub access • SOC 2-aligned controls • No credit card required