Story Points considered Harmful- a psychologist's thoughts on estimation
summary
Argues that story points mask a latent psychological construct and produce bad outcomes unless estimation is instrumented: uses psychometrics and item response theory, compares 15,000 estimates versus actuals and finds a skewed log‑normal distribution (not Gaussian), shows open R code and a Bernardon/Derek Jones model, recommends reference class forecasting, task unpacking, tracking estimates vs actuals, using an exponentially weighted moving average (alpha ≈ 0.28) and Bayesian updating; warns estimates have one‑sprint shelf life.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:05Estimation struggles
- 3:14Psychology of estimation
- 6:44Psychometrics of estimation
- 7:49Estimation fallacies
- 10:09Planning fallacy impact
- 12:18Estimation dimensions
- 14:49Distribution fallacy
- 17:59Velocity fallacy
- 22:57Objectivity fallacy
- 24:06New estimation techniques
- 28:21Reference class forecasting
- 31:33Task breakdown 'oh effect'
- 33:42Bayesian estimation
- 35:30Q&A
related talks
15mEstimation and Trust-Driven Development
40mEstimations: from guessing to weighing in grams
51mYour worth as a developer: not what you do, but why
20mSticking the Landing
33mMeasuring Developer Productivity: the good, the bad and the data
56mDebugger; for developers
35mYour Brain is a Liar: Cognitive Pitfalls at Work & How to Fight Them
11mHow to deal with your demons