Archetypes of delay: An analysis of online developer conversations on delayed work items in IBM Jazz

Abstract

Context.: A widely adopted methodology, agile software development provides enhanced flexibility to actively adjust a project scope. In agile teams, particularly in distributed environment, developers interact, manage requirements knowledge, and coordinate primarily in online collaboration tools. Developer conversations become invaluable sources to track and understand developers’ interactions around implementation of requirements, as well as the progress of implementation relative to the project scope and the planned iterations in agile projects. Although extensive research around iteration planning exists, there is a lack of research that leverages developer conversation data to understand delays in implementing planned requirements in agile projects. Objective.: By using developer conversations in a large agile project at IBM, this work aims to analyze conversation in work items (WIs) that are delayed and derive patterns that suggest reasons for delay in the project. Method.: We conducted a case study of the IBM Jazz project, and used thematic analysis to code the developer conversations as time-series, and cluster analysis to identify patterns that differentiated the evolution of discussions in WIs that were late vs. not late in the project. Results.: We identified six main patterns of WI delay. Through semantic analysis of developer conversations within particular clusters we were able to explain the reasons for delays in each pattern. In comparison to non-late WIs, we find that the major reason for delay is a lack of frequent communication associated with a poor project management of WIs. Similarly, non-late tasks more often delegate to children tasks to accelerate the implementation of requirements, in addition to processing requests quickly to resolve bottlenecks in implementation. Conclusion.: Our study complements existing research in bringing evidence that developer conversations are a useful resource that can highlight delays in requirement implementation, as well as recommend patterns in the dynamics of developers interactions relevant to such delays.