Software Estimation: Demystifying the Black Art [Paperback]

McConnell immediately differentiates between software estimation as a science and software estimation as an art within the first pages of the introduction to this work - he explains that while software estimation research is currently focused on improving estimation techniques so that project results are achieved within +/- 5% of esimated results, the techniques of which are best included in commerical software estimation tools that are used to pursue the science of estimation for very large projects, the typical software organization struggles to avoid estimates that are incorrect by 100% or more, and Software Estimation discusses the estimation techniques comprising, although not limited to, the art of estimation that will reduce estimation error to about 25% or less, which is what the vast majority of software projects need. This book is absolutely the best software estimation text I have read to date. As Joel Spolsky mentions on the back cover of the book, the vast majority of software project managers still think that estimates are based on multiples of a gut feel. This philosophy completely ignores the vast body of research and hands-on experience of the last few decades. McConnell explains with sufficient depth all of the ways to create "ball park" estimates, and if there is one chapter of this book that you read, Chapter 4 "Where Does Estimation Error Come From?" is the one that I recommend - the explanation of the Cone of Uncertainty is so well explained that anyone can understand it. Although I have never read Code Complete by the same author, I did purchase Rapid Development several years ago - and I must say that McConnell (or at least his editors) has vastly improved his writing skills. This book is less than half the size of Rapid Development, and while I fell asleep while reading that book I was able to conduct a cursory reading of over half of Software Estimation in one evening and immediately apply some of its concepts on a work project the following day. A great modern follow-up to The Mythical Man-Month by Brooks and an excellent primer to reading SEI's Estimating Software-Intensive Systems by Stutzke on the science of estimation.

I must admit, I was very surprised to see an excellent source on software development time and cost estimation from Microsoft Press. However, when I saw that the author was Steve McConnell, the author of Code Complete (which I thought was a great book) I knew that this book would have something to offer.

His definition of the purpose of software estimation is important: "The primary purpose of software estimation is not to predict a project's outcome; it is to determine whether a project's targets are realistic enough to allow the project to be controlled to meet them."

From this premise, the author continues to prove that software estimation is about determining if you can control a project to bring it in reasonably close to the organization's targets. In addition to this, Mr. McConnell's arguments for overestimation being better than underestimation are rather good and have convinced me to continue using my risk time and budget items that I've been using and recommending to others.

In chapter 4, the author identifies four main sources of project estimation errors:

-Inaccurate information about the project being estimated
-Inaccurate information about the capabilities of the organization that will perform the project
-Too much chaos in the project to support accurate estimation (that is, trying to estimate a moving target)
-Inaccuracies arising from the estimation process itself

That last one presents the need for me to give a little more information. The author suggests that bias and subjectivity of the estimator cause major estimation errors in the estimation process. This is proven through exercise case studies of time estimation in training classes. The author found that the more "control knobs" (impacting factors considered when estimating the project) used in the estimation process the greater the probability of variance in the estimates. He attributes this to bias and estimator subjectivity and it makes a lot of sense.

In the end, this book will make you a better estimator and project manager and I highly recommend it.

Tom Carpenter, Author: Wireless# Certification Official Study Guide, CWSP Certification Official Study Guide, Project Management for the IT Pro, and Foundations of Effectiveness

This is Steve McConnell's latest published effort on software engineering. The book is divided into three parts: estimation concepts, techniques, and challenges respectively. The first part of the book, on concepts, is just a broad overview designed to convince the software engineer of the value of estimation and how, done incorrectly, it can mess up what would have been a successful project. The second part, on techniques, is the meatiest part of the book. Here the author combines tried and true computational techniques with your own judgement and reminds you to get the expert opinion of other people too when estimating a project. Finally, the third and last part of the book, on challenges, talks about how you must think "out of the box" when you confront a software project whose size, effort, and schedule may be hard to quantify completely. Finally, McConnell points out the importance of politics and group dynamics in general in estimation. This is important, as many software engineers and project managers tend to be lone wolves that want to go off in a corner and solve problems by themselves. I highly recommend this book to all software engineers and particularly to software project managers. Amazon does not show the table of contents, so I do that here:

CRITICAL ESTIMATION CONCEPTS

1 What is an "estimate"? 3

2 How good an estimator are you? 15

3 Value of accurate estimates 21

4 Where does estimation error come from? 33

5 Estimate influences 55

FUNDAMENTAL ESTIMATION TECHNIQUES

6 Introduction to estimation techniques 77

7 Count, compute, judge 83

8 Calibration and historical data 91

9 Individual expert judgment 105

10 Decomposition and recomposition 113

11 Estimation by analogy 127

12 Proxy-based estimates 135

13 Expert judgment in groups 149

14 Software estimation tools 157

15 Use of multiple approaches 165

16 Flow of software estimates on a well-estimated project 171

SPECIFIC ESTIMATION CHALLENGES

17 Standardized estimation procedures 181

18 Special issues in estimating size 197

19 Special issues in estimating effort 207

20 Special issues in estimating schedule 221

21 Estimating planning parameters 233

22 Estimate presentation styles 249

23 Politics, negotiation, and problem solving 259

A Estimation Sanity Check 271

B Answers to Chapter 2 Quiz, "How Good an Estimator Are You?" 273

C Software Estimation Tips 275