COLLABORATION STRUCTURE AND PERFORMANCE IN NEW SOFTWARE DEVELOPMENT: FINDINGS FROM THE STUDY OF OPEN SOURCE PROJECTS

Abstract

The state of software development performance is far from being exemplar, with a success rate well below that of other industries, and understanding how to improve these projects is not only substantive but urgent. Software development is at most times a collaborative effort, yet we do not understand clearly how different collaboration structures associate with development performance metrics. This paper empirically examines the association between collaboration patterns, project productivity and product quality through a field study of working software open source new product development teams, using archival data from electronic sources related to Open Source Software projects. Collaboration structures are measured using Social Network Analysis and in terms of their network density, network centralization and the level of boundary spanning activity of team members. Productivity and quality are measured using "hard", code-based metrics. Results of regression analyses testing relevant hypotheses suggest that project managers need to strongly encourage internal collaboration, but be wary of allowing team members to participate in multiple projects. The breadth of skills within the team is a tactical asset that may increase the efficient frontier in the quality–productivity trade-off. Results also show that centralized structures associate with higher product quality and development productivity.