Steve McConnell is an excellent writer of software development books; I thoroughly enjoyed Code Complete, and have Code Complete Second Edition on my self to read. This book, Software Estimation, I enjoyed as well, although when comparing the subject matter to my own history on software projects, I wondered if he paints a much more idealized world than I’ll ever experience.
Liked from this book:
- Estimating in terms of ranges and probabilities. I like the idea of presenting monthly estimates in ranges, and then ranges compress as the project nears completion; much better than single point dates that seem to get pushed back each month.
- Alternatives to “gut instinct” estimation, such as counting and deriving, even if counting something indirectly related. Also the notion of taking different estimation techniques (and estimators) into consideration. The only technique I’ve ever experienced is single person, gut instinct estimates – often set at by project managers without asking the people who will do the work.
- Importance of measuring – and consistent remeasuring – to learn to provide more accurate estimates. The power of tracking consistently over a long period of time, a lesson learned many times, but taking many years to take hold as a habit.
Dislikes from this book:
- Too much detail when discussing different estimation techniques. I care most about the principles, not the exact steps.
- Sounds almost surreal in the amount of measurements made on the example projects discussed. Perhaps they were meant as illustrative, rather than the common project?
- Project management type numbers. Charts demonstrating the non-linear relationship between manpower and/or function points to project duration is not interesting to me, although the fact itself is.
I’m glad I read the book. I learned things I probably wouldn’t learn from the Internet or from the people I typically worth it. I need something lighter for my next book, though.
