Estimation: The Art of the SWAG
summary
Practical methods and hard numbers for real estimates: use comparison-based estimation (compare to past projects), decompose into user stories or go bottoms‑up, or count artifacts like content types and modules; T‑shirt sizing speeds early guesses. Early estimates sit in a cone of uncertainty (McConnell: 2.5x–4x spread — 1,000 hours → 2,500–4,000 for 90% coverage). State assumptions, prefer ranges, bias errors up for fixed‑price work, and reestimate as discovery narrows uncertainty.
AI-written summary · about
chapters (22)
- 0:00Introduction and speaker background
- 1:57What is an estimate?
- 5:04What an estimate is not
- 7:07Key questions for accurate estimates
- 11:58The cone of uncertainty explained
- 14:27Comparison-based estimation
- 17:13Decomposition in estimation
- 19:54Bottom-up estimation
- 22:19Counting method for estimates
- 24:38T-shirt sizing by feature
- 25:20Staff planning and feasibility
- 27:28Principles of good estimation
- 35:17Project control and leadership
- 37:30Refining estimates over time
- 39:36Underestimating risks and pricing
- 41:53Summary and final tips
- 43:55Q&A: Handling client assumptions
- 46:28Q&A: Bottom-up vs decomposition
- 48:03Q&A: Value-based pricing vs hourly
- 50:55Q&A: Developer pressure and culture
- 54:32Q&A: Updating clients on estimates
- 59:13Closing remarks and Q&A wrap-up
related talks
29mSoftware estimation – embracing nuance and controlled chaos
40mEstimations: from guessing to weighing in grams
30mSoftware Estimation Made Easy
42mBeyond Estimates (Estimates and "NoEstimates") - Let's Explore the Possibilities
03h 00mBeyond Estimates (Estimates and NoEstimates ) - Let's Explore the Possibilities
47mCreating efficient, accurate, software estimations
58mThe State of NoEstimates
01h 45mEstimates or NoEstimates