Increasing the Understanding of Effectiveness in Software Inspections Using Published Data Sets

Since its inception into software engineering, software inspection has been viewed as a costeffective way of increasing software quality. Despite this, many questions remain unanswered regarding, for example, ideal team size or cost effectiveness. This paper addresses some of these questions by performing an analysis using 30 published data sets from empirical experiments of software inspections. The main question is concerned with determining a suitable team size for software inspections. The effectiveness of different team sizes is also studied. Furthermore, the differences in mean effectiveness between different team sizes are investigated based on the inspection environmental context, document types and reading technique. It is concluded that it is possible to choose a suitable team size based on the effectiveness of inspections. This can be used as a tool to assist in the planning of inspections. A particularly interesting result is that variation in the effectiveness between different teams is considerably higher for certain types of documents than for others. Our findings contain important information for anyone planning, controlling or managing software inspections. ACM Classification: D.2 (Software Engineering), D 2.5 Testing and Debugging (Inspections and Walkthroughs), D 2.8 Metrics, D 4.8 Performances