Tuesday, June 04, 2002

 
Back to Me? Linda is absolutely correct - the foundation of any process improvement or quality initiative is measurement. There are two excellent books on the subject that are specifically for software professionals:
  1. Applied Statistics for Software Managers. If you're working in SQA or managing software development projects this book is an excellent introductory text to statistical analysis.
    What I like about this book is that it's a tutorial on the statistical skills and knowledge that you'll need, and it combines this learning goal with the basics of software metrics and how they can be employed to measure productivity, estimate projects, and manage costs and organizational quality. The core approach is data analysis, and the main tools that the book employs are multi-variate techniques, regression analysis and correlation and sensitivity tests. The author has a talent for clearly explaining a dry subject, and while it will take a good deal of effort to master the material because of its nature, the excellent writing and illustrations will make it easy to quickly grasp statistical fundamentals and put them to use.
    The lessons are taught within the framework of four case studies that are realistic and apply to the real world. The case study topics are: productivity analysis, analysis of time to market factors, development cost analysis, and maintenance cost drivers. These cover the full range of both internal development and product-line software engineering. I especially like the inclusion of maintenance costs as a topic of study because this area contributes significantly to total costs of ownership, but is often overlooked.
  2. Measuring the Software Process. This book contains the keys to meeting core CMM level 5 requirements, which defines key processes for optimizing and continuous improvement, and for achieving 6-sigma processes. However, you need not be striving for either (or both) of these goals to use the techniques and approach in this book to full advantage.
    Implementing and employing statistical process controls are the basis of this book. The authors lead you through the steps and techniques necessary to implement and use SPC, starting with background information on processes and a process measurement framework, and moving through topics such as planning your measurement strategy, data collection and analysis, and developing and interpreting process behavior charts using common SPC chart types. The most common controls are x-bar (mean) and r (range) charts. Be aware that any SPC approach requires two conditions to be met:
    • defined processes
    • the processes are in statistical control (meaning that the data points being measured have settled into a normal distribution that are randomly clustered around a mean and have defined upper and lower control limits)
    New processes, or processes that are not managed well enough to have these characteristics are not candidates for SPC.
    This book requires knowledge and skills in basic statistical analysis. If you require a refresher I recommend reading Visual Statistics before tackling this book.
Deciding which of the two books is better is a matter of assessing your needs. The key strengths of Measuring the Software Process are the tutorial nature and the wide range of case studies that are used to reinforce the learning. The key strengths of Applied Statistics for Software Managers are that it goes much deeper into analysis and also includes statistical process controls and other techniques that are present in highly mature development organizations. Regardless of which book you choose (or if you choose both), the information and knowledge to be gained is the foundation of SQA and best practices in project management.

Sunday, June 02, 2002

 
On Software Process Improvement. Before we become mired down in Oracle topics I am going to take an abrupt turn back towards quality and process improvement. One excellent book on the subject that covers both process assessment and improvement, is Software Process Improvement. With exceptions that I've noted below this is an in-depth examination of standards, initiatives and methods for software process improvement (SPI) and software process assessment (SPA).

The book is divided into twelve chapters, each of which contains two or more papers written by top experts in the field, including Mark Paulk (of CMM fame), Watts S. Humphrey (creator of PSP and TSP, and prolific author of software engineering process papers), Robert B. Grady (author of three standard references on metrics), and others who key players, but are not as widely known outside of the SPI and SPA community.

Chapter 1 covers software process assessment with an article by Paulk that surveys the more common models for SPI and SPA, and a reprint of Sarah Sheard's excellent article from CrossTalk Magazine titled "The Frameworks Quagmire". Chapter 2 contains three articles on the SW-CMM, which seems to be the centerpiece of this book. Chapter 3, "Other Approaches to Software Process Assessment" contains four articles that add balance by covering non-CMM approaches that are in common use, especially in Europe (Bootstrap). I especially liked the article by David N. Card titled "Sorting out Six Sigma and the CMM", which combines two hot topics. One of the exceptions that I cited at the beginning of this review is the article on Trillium, which in my opinion has been superseded by TL 9000 in the telecommunications industry.

The three articles in Chapter 4 (Software Process Improvement: How To Do It) address common concerns and barriers to any SPI initiative, and each add well thought out ideas, especially Sandra McGill's "Overcoming Resistance to Standard Processes, or, Herding Cats", and William Florac's "Statistically Managing the Software Process".

Watts Humphrey's Personal and Team Software Processes, and CMMI are the key topics in Chapter 5, which covers developments inspired by the SW-CMM. All of Chapter 6's Software Product Evaluation articles were my favorites from among the collection in this book, and I particularly liked Jørgen Bøegh's "Quality Evaluation of Software Products" and Geoff Dromey's "A Model for Software Product Quality" because they go to the heart of key issues in both product line engineering challenges and user acceptance testing.

Chapter 7, ISO 9000 Series and TickIT, is the second exception that I previously noted. Much has changed in ISO 9000 with the 2000 standard, which renders this entire chapter moot in my opinion. I also thought the five articles in Chapter 8, The SPICE Project, would have been a better fit in Chapter 3. The same goes for Chapter 9, Experiences of Software Process Assessment, which is nearly an extension of Chapter 8, and is closely related to Chapter 3.

Two other favorite chapters are 10 (Software Process Improvement for Small Organizations) and 11 (Benefits of Software Process Improvement). Chapter 10's three articles dispell any notion that SPI is only feasible for large organizations, and the three articles in Chapter 11 focus on the benefits of SPI, especially Herb Krasner's article titled "Accumulating the Body of Evidence for the Payoff of Software Process Improvement". I also liked the final chapter, which covers software processes in general, including an excellent article on modeling. I felt that this chapter should have been at the beginning of the book instead of the end.

Overall, this is a book for those of us who are nearly religious about SPI; but is not a good introductory text. It's main value will be to IT consultants who specialize in either SPI or SPA (or both), and who need to be familiar with the mainstream standards and approaches.

This page is powered by Blogger. Isn't yours?

Subscribe to Posts [Atom]