An Overview of Software Defect Density: A Scoping Study

Context: Defects are an ineludible component of software, Defect Density (DD) - defined as number of defects divided by size - is often used as a related measure of quality. Project mangers and researchers alike would benefit a lot from overview DD figures from software projects, the former for decision making the latter for state-of-the-practice assessment. Objective: In this paper, we collect and aggregate DD figures published in literature, in addition we characterize DD as a function of different project factors in terms of central tendency and dispersion. The factors considered include development mode –open vs. closed source–, programming language, size, and age. Results: We were able to identify 19 papers reporting defect density figures concerning 109 software projects. The mean DD for the studied sample of projects is 7.47 post release defects per thousand lines of code (KLoC), the median is 4.3 with standard deviation of 7.99. Development mode, is characterized by statistically meaningful different DD, the same for Java vs. C. Besides, in the studied sample large projects exhibited lower DD than medium and small projects. Conclusion: The study is a first step in collecting and analyzing DD figures for the purpose of characterizing one important aspect of software quality. These figures can be used both by researchers and project managers interested to evaluate their projects. Further work is needed to extend the data set and to identify predictors of defect density.