Does Socio-Technical Congruence Have an Effect on Continuous Integration Build Failures? An Empirical Study on 10 GitHub Projects

Abstract

Coordination is important in software development. Socio-Technical Congruence (STC) is proposed to measure the match between coordination requirements and actual coordination activities, and has been proved to have impact on software failures in commercial projects. Continuous defect prediction is aimed to predict defects just in time, which is more meaningful than traditional defect prediction in practice. In this paper, we compute the build-level STC and investigate its usefulness in continuous defect prediction based on 10 GitHub projects. We find that adding STC metrics into logistic regression models can significantly improve both the explanatory power and the predictive power when predicting build failures. Furthermore, we compare the performance of STC and MDL from the aspects of regression and prediction. MDL is short for Missing Developer Links, a deviation of the STC metric. We find that MDL usually performs better than STC. Our work is promising to help detect coordination issues during real time process of software development.