An evaluation of the reinsepction decision policies for software code inspections

Thesis Type: Postgraduate

Institution Of The Thesis: Middle East Technical University, Faculty of Engineering, Department of Industrial Engineering, Turkey

Approval Date: 2005

Thesis Language: English

Student: Serkan Nalbant



This study evaluates a number of software reinspection decision policies for software code inspections with the aim of revealing their effects regarding cost, schedule and quality related objectives of a software project. Software inspection is an effective defect removal technique for software projects. After the initial inspection, a reinspection may be performed for decreasing the number of remaining defects further. Although, various reinspection decision methods are proposed in the literature, no study provides information on the results of employing different methods. In order to obtain insight about this unaddressed issue, this study compares the reinspection decision policies by finding out and analyzing their performance with respect to designated measures and preference profiles for cost, schedule, and quality perspectives in the context of a typical Software Capability Maturity Model Level 3 software organization. For this purpose, a Monte Carlo simulation model, which represents the process comprising initial code inspection, reinspection, testing and field use activities, is employed in the study together with the experiment designed in order to consider different circumstances under which the mentioned process operates. The study recommends concluding the reinspection decision by comparing inspection effectiveness measure for major defects with respect to a moderately high threshold value (i.e. 75%). The study also reveals that applying default decisions of ءNever Reinspect̕ and ءAlways Reinspect̕ do not exhibit the most appropriate outcomes regarding cost, schedule, and quality. Additionally, the study presents suggestions for further improving the cost, schedule, and quality of the software based on the analysis of the experiment factors.