📖 Business
Lean Management Practices
The Accelerate research identified four lean management practices that predict software delivery performance. These aren't new ideas — they come from Toyota's production system and decades of lean manufacturing — but the research validated their application to software with statistical rigor. The most provocative finding: external change approval processes (heavyweight change advisory boards) predict LOWER performance. Teams using peer review for changes have 3x higher deployment frequency and lower change failure rates.
2
Minutes
2
Concepts
+45
XP
1
How It Works
Four practices that form a lean management system:
- Limit Work in Progress (WIP) — Constrain the number of things being worked on simultaneously. WIP limits reduce context switching, expose bottlenecks, and increase throughput (counterintuitively, doing less at once means finishing more). Make WIP visible. When you hit the limit, finish something before starting something new. The physics are clear: high WIP = high lead times, low WIP = fast flow.
- Visual management — Dashboards and boards showing current work status, flow metrics, and production health. Visible to the whole team — not locked in a manager's spreadsheet. Includes deployment pipeline status, current incidents, and work-in-progress counts. The principle: if the team can't see it, they can't manage it.
- Feedback from production — Monitoring and observability data actively used to improve products and processes. Not just alerting on failures — actual learning loops where production data informs product decisions, architectural improvements, and process changes. Teams that use production data to learn (not just firefight) build better systems over time.
- Lightweight change approvals — Peer review (pull requests, pair programming) instead of heavyweight change advisory boards. The research is unambiguous: external approval processes that require people outside the team to approve changes are correlated with worse outcomes. Peer review by people with context about the code is faster, catches more real issues, and doesn't create a bottleneck.