Extended Metrics to Evaluate Cost Effectiveness of Software Inspections

For more than twenty-five years software inspections have been considered an effective and efficient method for software quality improvement [3,4,9]. The goal of inspections is to detect defect before the testing stage begins. According to the data published in the literature, software design inspections save on average 44% of the defect detection costs, and code inspections save on average 39% of the defect detection costs [1]. However, are the software inspections always worthwhile? In order to evaluate the effectiveness of the inspections with respect to software development cost, several metrics have been previously proposed [2,3,6]. Collofello et al. have taken into account of the costs consumed and saved by inspections and proposed a metric, called Cost Effectiveness [2]. Kusumoto et al. proposed a metric Mk for evaluating cost effectiveness of inspections, which is based on the degree to which costs to detect all faults from the software in a project are reduced by inspections [6]. Yet, none of those metrics includes the information about false positives (the issues that require no repair) introduced during inspection, although the rework of false positives is costly and can introduce new defects [8,12]. This letter presents a) an inspection cost model that describes all costs related to inspections, b) extension of the metric Mk to evaluate the cost effectiveness as well as the losses of inspections. Proposed cost model and metrics can be useful for project managers to decide whether inspections are worthwhile. The structure of the letter is as follows. Section 2 describes the inspection process. Section 3 describes the inspection cost model. In Section 4, we review the existing metrics to evaluate software inspections. Section 5 presents the extended metrics for software inspection evaluation. In Section 6, we compare extended metrics to Collofello’s metric and Mk. Section 7 concludes the letter with an overall summary and directions for future research.