Investigating the effect of a second software inspection cycle. Cost-benefit data from a large-scale experiment on reinspection of a software requirements document

Abstract

The goal of software inspection is to identify defects and to measure the product and inspection process quality. Software development products that are suspected to contain a substantial number of defects after inspection may be subject to a second inspection cycle (re-inspection). So far, very few empirical results can be found on re-inspections. This paper reports on a controlled experiment that compares the effect of two inspection cycles. A software requirements document was re-inspected by 31 teams, after defects found during the initial inspection had been removed. We present data on the effectiveness, efficiency, net gain and return-on-investment from a second inspection cycle, and compare these results with data from the initial inspection. Models in the literature suggest decision criteria with regard to when to re-inspect, based on assumptions on the relationships of variables between the two inspection cycles. With experimental data, we evaluate these assumptions and the correctness of their decision criteria. The results of the experiment strongly support considering re-inspection as an option to improve product quality.