Abstract

Communication between developers and testers can be a rich source of insights into software development processes and practices, which may not be easily discoverable from other means like retrospectives or project roadmaps. With the objective of deriving and capitalizing on potential development-related insights, we analyzed developer-tester communication in an industrial setting. We conducted a case study at a software-intensive Agile company, within the context of the development of one of their flagship products from 2016 to 2018. We applied Latent Dirichlet Allocation (LDA) to analyze communication between developers and testers, and then invited two case-company practitioners to study the results for insights into their developments processes: The findings reveal the case company’s efforts to improve their product stability, the growing emphasis on addressing end-user concerns and other quality-related issues. The practitioners interpreted these findings as indicators of evolution in their development process. Based on these findings and the state of the art, we propose an insight classification to highlight insights discoverable from developer-tester communication: Recognizing LDA’s potential for deriving insights, the practitioners are keen on incorporating it into their software development practices. The findings from this study serve as evidence for use and benefits of text-mining techniques like LDA in industrial setting, which other practitioners could adapt to elicit their own context-influenced insights. Furthermore, the insight classification can serve as a foundation for further investigation into the extent and type of insights discoverable from developer-tester communication.