Modeling and Predicting Software Failure Costs

For software, the costs of failures are not clearly understood. Often, these costs disappear in the costs of testing, the general developments costs, or the operating expenses. In a general manufacturing context, the British standard BS-6143-2:1990 classifies quality-related costs into prevention costs, appraisal costs, and failure costs. It furthermore recommends to identify the activities carried out within each of these categories, and to measure the costs connected with the activities. The standard thus presents a framework for recording and structuring costs once they have occurred. In this paper, we propose an approach for structuring the information on internal and external software failure costs such that their development over time can be represented by stochastic models. Based on these models, future failure costs can be predicted. In two case studies we show how the approach was applied in an industrial software development project.