What are the odds? Transforming forecasts through probabilities
summary
Claims that probabilistic forecasts built from cycle time (finish − start + 1) and throughput match or beat traditional estimates, needing only ~10–15 completed items of start/finish data. Practical steps: plot cycle times as a scatter to read percentiles (50th, 85th), run a Monte Carlo simulation over historical throughput for fixed‑scope or fixed‑date answers, and pick a confidence level (85% as a pragmatic example). Teams use it at item, release, and portfolio scale and report it works.
AI-written summary · about
chapters (13)
- 0:00Probabilistic forecasting intro
- 2:52Certainty in forecasts impossible
- 7:04From deterministic to probabilistic
- 10:14Data needs for probabilistic
- 12:03Cycle time for single items
- 13:49Interpreting scatter plots
- 17:22Choosing confidence levels
- 21:46Throughput for multiple items
- 23:52Monte Carlo simulations
- 26:42Fixed scope forecasts
- 27:49Fixed date forecasts
- 30:42Practical forecasting takeaways
- 33:28Q&A
related talks
16mPractically predictable – Forecasting delivery when nothing is certai
44mAgile metrics for predicting the future
41mYour Project Behaves Like a Hurricane--Forecast It Like One
11mEstimates as probabilities
30mSoftware Estimation Made Easy
47mCreating efficient, accurate, software estimations
41mThe Development Metrics You Should Use (but Don’t)
51mGet Predictability Regardless of Estimates