Code Review Quality and Software Reliability: The Mediating Role of Defect Detection and the Moderating Role of Developer Experience

Abstract

This study aims to explore how the quality of code review relates to software reliability, while focusing on the role of defect detection as a mediator and developer experience as a moderator between code review and software reliability in an information technology (IT) organization with the software engineers and software quality assurance (QA) teams. The researcher used a quantitative approach and explanatory research design to gather the data from the software engineers and QA professionals using validated multi-item scale of code review quality, defect detection, developer experience, and software reliability. To assess data's validity and reliability, first data validation was performed, second confirmatory factor analysis was performed to ensure the consistency of data measurements across constructs. Partial Least Squares Structural Equation Modeling (PLS-SEM) was used to analyze the relationships between the variables in this study because of their hierarchical, second-order and interaction-based forms. The impact of defect detection on software fault was estimated using bootstrapped indirect effects, while the moderating effect ofdeveloper experiencein the relation between software and defect parameters was analyzed using a hierarchical interaction modeling approach within the PLS framework. In addition to traditional path significance testing, two measures (effect size (f²) and predictive relevance (Q²)) were introduced to assess the strength of the model. All constructs showed good levels of reliability, convergent and discriminant validity of the measurement model. There was a significant positive effect of code review quality on both defect detection and software reliability, a significant positive predictiveness of software reliability on defect detection, and there was a partial mediation between code review quality and software reliability. It was determined that the relationship between defect detection and software reliability was significantly moderated by whether the developer was highly or less experienced, and was higher in the cases of high experience. The results indicate that such translation of code review investment into software reliability works significantly, but mainly through defect detection, and that this translation depends both on the level of experience of the development team and on the contraction-level of the code, implying software organizations may be able to optimize code review investment and team by substituting experienced developers with less experienced ones, or by implementing non-experimental code reviews on less criticized layers of the contraction.