Abstract

The success of a software project is dependent on the expertise and knowledge of its developers.
In this dissertation, we use empirical studies to develop an understanding of the impact of knowledge loss on software projects. First, we studied the damage done to projects from turnover, the susceptibility of the project to future turnover, and the suggestion of potential successors to assume abandoned files. Based on the project vulnerability to turnover, project leaders can induce key developers to stay with the project and to mitigate files abandonment. Second, we did an empirical research on the impact of turnover on the quality of a software project. Third, we performed an examination of the impact of inactive files (dormant files).
Our findings on the first research topic showed that the greater the spread of knowledge the less likely a project is to be affected by turnover. Moreover, we found that knowledgeable developers, rather than newcomers, take over abandoned code. In our second study, we observed an unexpected result that in the Chrome web-browser project, the number of developers who leave and join both decreased the number of post-release defects. We discuss this unexpected result. The third study on dormant files, i.e. inactive files, contrasted a legacy system with a popular system. We found that for a legacy system, the developers that take on dormant files were experienced developers.