The Development Metrics You Should Use (but Don’t)
summary
Shows how to use historical delivery data to answer concrete questions: time in process (cycle time) plotted as a scatter lets you read probabilities (example: 90% = 53 days), not misleading standard deviations; a 45‑day mandatory babysit item creates a long tail that skews predictions; throughput measured per month gives probabilities (about 85% → ≥6 items, 50% → ≥11 items for one team); warns against using Little’s Law for estimates and recommends attacking waiting time because flow efficiency 10–15% touch time is typical.
AI-written summary · about
chapters (8)
- 0:00Intro
- 1:57Know what you are measuring
- 4:20Software development performance index
- 6:42Quality
- 7:50Responsiveness
- 26:55Productivity
- 32:17Predictability
- 36:16Things to notice
related talks
36mThe Development Metrics You Should Use (but Don’t)
49mThe Development Metrics You Should Use (but Don’t)
23mThe Shape of Uncertainty
51mGet Predictability Regardless of Estimates
43mPredictable Delivery
44mAgile metrics for predicting the future
14mI don't think these metrics mean what you think they mean!
16mPractically predictable – Forecasting delivery when nothing is certai