Software Inspection, Tom Gilb and Dorothy Graham
(Addison-Wesley, Wokingham, England: 1993, 471 pages.)

This treatise aims to satisfy the needs of practitioners interested in implementing or improving inspection processes, while also providing material that management could use to justify inspections as a cost-effective, even productivity-enhancing method to add to existing development methods. Although the subject is software inspection, Gilb and Graham expand the scope to include any process that produces documents as a place where inspections would be beneficial, and they present numerous examples to support their claim.

The section in the preface called "How To Navigate Around This Book" identifies the target audience for each chapter and gives a road map for how the book should be read. Reading the book from cover to cover, we found it somewhat wordy and repetitive. Concentrating only on the sections as prescribed for specific audiences, it was much more readable. However, we would recommend adding Chapter 11, "Overcoming the Difficulties," to the manager's reading list to help avoid pitfalls. If not avoided, some of these pitfalls could ensure that the inspection process would fail.

Chapter 2, "The Benefits and Costs of Inspection," gives several examples of the costs other organizations have experienced by not implementing inspections--delivering poor quality software. Excellent information is presented on how to build a business case to support implementation of inspections. Inspections are contrasted with other forms of peer reviews. Experience-based data documents the costs of implementation with associated benefits seen through productivity gains, increased schedule compliance, and improved quality. Some side benefits--such as transfer of best practices and proving to be an effective training vehicle--are also presented. These examples identify what factors should be considered and what commitments can reasonably be made in developing the business case justification and also defines the metrics collection that should be integrated into the process to demonstrate realization of expected goals.

The sections devoted to helping the practitioners understand and implement the process provide detailed guidance on roles, duties, and responsibilities as well as steps to follow in process implementation. The appendices provide a sample Inspection Handbook, detailing philosophy, objectives, process phases, and metrics to be collected on one page (quite a feat!). Also included are sample procedures by inspection phase defining activities and functions to be performed according to process role, and finally, examples of inspection forms to direct process flow and collect required metrics. All of this information would be very useful to someone trying to implement an inspection process. However, the information is presented in such a dogmatic and autocratic manner as to possibly overwhelm someone new to the process.

Lost in this presentation is the authors' own advice not to follow all the rules initially: start small, collect only metrics you will use, and use what you collect. Some of the rules seem to be based less on research than the authors' own intuition, e.g., the one page per hour scanning rate. No research is cited for this rule, though a study is cited that relates checking rate to the number of defects discovered. Even though maximum benefit is realized by beginning inspections on upstream documents, benefits will be realized wherever in the process inspections are used. Sometimes, beginning where the organization is experiencing the most pain, such as meeting code development dealines with required quality, and showing positive results might be the most effective way to get buy-in for the inspection process.

The case studies show beginners what to expect when trying to implement inspections. They cover a range of environments and situations--from one person trying to apply the techniques, to a large electronics organization of 100 developers with a variety of document types being considered as inspection candidates. The case studies show that the process must be tailored to fit the specific organization's culture. The main advice is to begin with as much rigor as the culture will allow and to build from there as the process and the organization mature.

In summary, this is a good reference book that discusses all aspects of inspections and beyond and also addresses defect prevention. The defect prevention section is extremely useful in helping to explain how the metrics that are collected can be used later, after a good history of defect data has been collected. We also recommend that an organization not try to implement inspections and defect prevention practices at the same time. As mentioned above, start small with the inspection and metrics collection processes, but keep in mind that much of the data recommended for collection in this book will be useful in the future when the organization is ready to implement defect prevention processes. We recommend that anyone seriously interested in implementing or improving inspection processes include this book in their library.