A General Model of Unit Testing Efficacy

Much of software engineering is targeted towards identifying and removing existing defects while preventing the injection of new ones. Defect management is therefore one important software development process whose principal aim is to ensure that the software produced reaches the required quality standard before it is shipped into the market place. In this paper, we report on the results of research conducted to develop a predictive model of the efficacy of one important defect management technique, that of unit testing. We have taken an empirical approach. We commence with a number of assumptions that led to a theoretical model which describes the relationship between effort expended and the number of defects remaining in a software code module tested (the latter measure being termed correctness). This model is general enough to capture the possibility that debugging of a software defect is not perfect and could lead to new defects being injected. The Model is examined empirically against actual data and validated as a good predictive model under specific conditions. The work has been done in such a way that models are derived not only for the case of overall correctness but also for specific types of correctness such as correctness arising from the removal of defects contributing to shortcoming in reliability (R-type), functionality (F-type), usability (U-type) and maintainability (M-type) aspects of the program subject to defect management.