Beyond Estimates: Forecasting with Little's Law
summary
Little's Law gets used to forecast by counting completed stories and applying throughput = WIP / lead time under steady‑state assumptions (arrival rate ≈ departure rate, consistent units). Analysis of 55 projects and simulations show throughput gives comparable accuracy to velocity; typical uncertainty bands are large — p90/p10 ≈ 4 (jellybean estimates ~60 to 360, ratio ≈6) — and many projects show a hardening phase 10–12% near the end. Estimation only helps when item sizes vary massively or estimation accuracy is unusually tight.
AI-written summary · about
chapters (29)
- 0:00Intro
- 0:34Little's Law (John D. C. Little version)
- 1:42Little's Law in Knowledge Work
- 4:10Little's Law - a Relationship of Averages
- 5:50Estimation Exercise
- 7:10Decisions, Decisions, Decisions
- 8:06What do we Estimate?
- 8:27Managing the Coming Storm Inside the Cyclone
- 8:42So can we use Little's Law for Forecasting?
- 9:16About the Data
- 9:55Basic Burnup
- 10:51What does the Story Point Data show?
- 11:11What does the Throughput Data show?
- 12:14Typical Distributions
- 12:50P90 and P10
- 14:27P90/P10 Ratio
- 15:02Estimate the # of Jelly Beans
- 16:19Burnup Projection to P10 and P90
- 18:45Conclusions from the Data
- 20:38How about a Simulation
- 22:20Simulation Results - What If?
- 25:36What do the simulations tell us
- 26:46What does this tell us about Estimation?
- 28:58IEEE Software, May/June 2006
- 29:17Accuracy of Initial Estimate
- 29:52Data From Steve McConnell
- 30:08Actual vs. Original Estimate
- 30:57Jørgensen 2013
- 32:15LITTLE'S LAW EXAMPLE
related talks
30mSoftware Estimation Made Easy
16mPractically predictable – Forecasting delivery when nothing is certai
41mThe Development Metrics You Should Use (but Don’t)
40mWhat are the odds? Transforming forecasts through probabilities
44mAgile metrics for predicting the future
51mGet Predictability Regardless of Estimates
47m[Workshop] NoEstimates: Forecasting with Less Effort and More Accuracy
43mPredictable Delivery