Software Estimation: Demystifying the Black Art
Steve McConnell
Microsoft Press, 2006
ISBN: 0735605351

This book delivered on my typical expectations for a Steve McConnell book—detailed, easy to follow, comprehensive, practical, and organized in a way that it can be used over and over as a reference. (If you're in software development and haven't ever read his book Rapid Development, I highly recommend it for the same reasons.)

Software Estimation is aimed at people who really want to take estimation of software projects further than the cursory work we all too often do—and have even accepted as all we reasonably can do since software development is "a creative art that can't be estimated." The book includes material on (just to name a few):

• Typical issues with estimating and sources of error estimation
• How to use different estimating techniques
• How the estimating process should unfold early in a project
• How to express the level of uncertainty of your estimates
• How to handle large vs. small projects

Software Estimation features several unique aspects. For example, each techniques chapter begins with an "Applicability" table indicating what the technique estimates (e.g. features, lines of code); what size of project it applies to; what development stage it can be used at; whether it applies to iterative or sequential development, or both; and how much accuracy you get or gain with the technique. It also contains a great deal of information and backup from research studies and other sources for its discussion of estimating accuracy using different techniques, industry benchmarks, and rules-of-thumb.

Part III of the text covers Specific Estimation Challenges in a way that helps readers determine the best techniques for their environment, given the typical shortfalls of different techniques in different situations. The individual sections provide guidance for alternative approaches with good explanations of why different techniques are most appropriate.

This book is appropriate for:

• people who are interested in a more scientific, data-centric look at estimating techniques and what studies say about what really works;
• people wondering how far to go with estimating techniques for small projects (that often turn out to be not so small after all);
• people managing complex projects—often with high degrees of risk and uncertainty—who want to take their estimating accuracy to an entirely new level, with some trustworthy backup on why certain techniques are useful, accurate, and worth the team's time;
• anyone who wants to understand the practical implementation of such techniques in their environment.