Developing and using defect classification schemes

Defects play a crucial role in software development. This is because on one hand, defects, when detected, should be corrected so that the final version of the developed software artifact is of higher quality. On the other hand, defects carry a lot of information that can be analyzed in order to characterize the quality of processes and products, to track the progress of a project and control it, and to improve the process. Therefore, defect measurement plays a crucial role in many software measurement programs. Consequently, in many measurement programs defect data are collected. Generally there are several pieces of information that can be collected about defects. The most often used pieces of information relate to the quantity of defects (i. e., their number) and their type. For the latter one, defect classification schemes are used to quickly characterize the nature of defects. Two important questions arise, when using defect classification: "How can a defect classification be designed?" and "How can defect classification data be analyzed?" In order to answer the first question, this report presents the aspects of a defect that have been measured in the literature and it presents the possible structures of a defect classification scheme. Finally, examples of frequently used defect classification schemes are presented. In order to answer the second question, this report presents general methods to analyze defect classification as reported in the literature as well as concrete analyses for a variety of purposes. This report is a result of the project "Integrierte Qualitätssicherung und Anforderungsanalyse zur Softwareentwicklung im Umfeld Fahrzeug (QUASAR)", which is funded by the German BMBF under the grant VFG0004A