NoEstimates
summary
Claims that estimating is harmful are backed with concrete data and experiments: a global program of ~100 teams runs 18 months then cancels after being ~6 months late; industry surveys show ~60–80% schedule overruns; a Shinkansen line achieves average delay ~54 seconds by redesigning the process. Practical alternatives include shortening feedback cycles, measure lead time not guessed effort, running two-week experiments that produced revenue in 3 days, and using CFD forecasting (treating all backlog items as size 1) to drop forecast error from ~25% to ~8%.
AI-written summary · about
chapters (19)
- 0:00The problem with estimates
- 2:25Unicorns and estimations
- 5:03Introduction to no estimates
- 7:12Lessons from other domains
- 10:53Reliability and predictability
- 14:40Rethinking process design
- 17:08Why projects really ship
- 19:26Shortening feedback loops
- 21:17Accidental complication
- 25:12The reality of queue time
- 26:54Industry failure data
- 28:58Estimates and company risk
- 31:45Definition of a good estimate
- 33:34Value over functionality
- 36:23Testing for value
- 39:26Measuring progress
- 42:13Dysfunctions of estimation
- 43:31Forecasting alternatives
- 45:51Final principles and summary
related talks
42mBeyond Estimates (Estimates and "NoEstimates") - Let's Explore the Possibilities
45mBeyond Estimates
47mCreating efficient, accurate, software estimations
03h 00mBeyond Estimates (Estimates and NoEstimates ) - Let's Explore the Possibilities
45mHow to not shoot yourself in the foot with estimation
58mThe State of NoEstimates
01h 45mEstimates or NoEstimates
48mRespect estimates