Software Engineering Institute

The software community has been slow to use data to measure software quality. This paper discusses the reasons for this problem and describes a way to use process measurements to assess product quality. The basic process measures are time, size, and defects. When these data are gathered for every engineer and for every step of the development process, a host of quality measures can be derived to evaluate software quality. Extensive data from the Personal Software Process (PSP) are used to derive the profiles of software processes that generally produce high quality software products. By examining these profiles, one can judge the likelihood that a program will have defects found in its subsequent testing or use. Examples are given of defect profiles, together with guidelines for their use.