Assessment of Test-driven Development in an industrial context

Abstract: Test-driven development (TDD) has been the target of many articles in which the authors try to reveal the impact of TDD compared to the traditional iterative test-last approach (ITL). Most of the existing articles conducts case studies in academic setting or in industrial setting with focus on metrics of little relevance for the industry. The metrics Defect Density per 1000 lines of code (DD/KLOC) and McCabe’s cyclomatic complexity number are frequently used to show differences between ITL and TDD in quality. However, these metrics are outdated and irrelevant in today’s industry according to several articles regarding TDD. The reason is the introduction of high-level programming languages such as Java, C# and Python. To compare these languages with assembler languages is not feasible. In this master thesis, the author suggests measuring defect density per hour instead (DD/h) of per 1000 lines of code to establish the quality and success of a software project. DD/h together with total hours needed to develop the software is a better measurement and is not dependent on programming language and complexity as DD/KLOC is to describe quality and the scope of the development project. By examine DD/h from five case studies comparing ITL and TDD the author shows a possible positive change in quality and resources needed in software development projects.