Investigating the effect of expert ranking of use cases for design inspection

Inspection is an important approach to reduce defects in software engineering artifacts. Reading techniques such as usage-based reading (UBR) can focus inspector attention on specific types of defects (missing and wrong information) and specific defect severity classes (critical, important and not important defects). For empirical investigation the replication of experiments can help increase the confidence in results. This work presents a large-scale external experiment replication in the UBR family of experiments in an academic environment. Additionally to experiment replication we introduce a new reading technique variant UBR-ir to investigate the impact of expert know-how with ranking of use cases on the number and severity of defects found. Main results of the study are: (a) UBR expert know-how had significant effects on inspection effort distribution, effectiveness, and efficiency; (b) the share of false positives was higher for checklist-based reading (CBR) than for all UBR variants; and (c) both UBR variants perform significantly better than CBR, possibly due to the active guidance of use cases.