The application of subjective estimates of effectiveness to controlling software inspections

One of the recently proposed tools for controlling software inspections is capture-recapture models. These are models that can be used to estimate the number of remaining defects in a software document after an inspection. Ba sed on this information one can decide whether to reinspect a document to e nsure that it is below a prespecified defect density threshold, and that th e inspection process itself has attained a minimal level of effectiveness. This line of work has also recently been extended with other techniques, su ch as the detection profile method (DPM). In this paper, we investigate an alternative approach: the use of subjective estimates of effectiveness by t he inspectors for making the reinspection decision. We performed a study wi th 30 professional software engineers and found that the median relative er ror of the engineers' subjective estimates of defect content to be zero, an d that the reinspection decision based on that estimate is consistently mor e correct than the default decision of never reinspecting. This means that subjective estimates provide a good basis for ensuring product quality and inspection process effectiveness during software inspections. Since a subje ctive estimation procedure can be easily integrated into existing inspectio n processes, it represents a good starting point for practitioners before i ntroducing more objective decision making criteria by means of capture-reca pture models or the defect detection profile method. (C) 2000 Elsevier Scie nce Inc. All rights reserved.