Software inspections: Collaboration and feedback

Abstract

This dissertation studies impact of collaboration and feedback on software inspection productivity. Used as a software-engineering validation technique, software inspections can be a cost-effective method for identifying latent issues (defects) within design documents and program code. For over two years, Baan Company has used a generalized Electronic Meeting System (EMS also referred to as GroupWare) to support software inspections and reported EMS to be more productive than face-to-face paper-based inspections (Genuchten, 1998). Validation of this phenomenon and initial development of a potentially more effective specialized EMS (SEMS) tool is the basis of this dissertation. Explanations of the collaborative phenomenon are presented within a theoretical framework along with testable hypotheses. The framework is derived from Media Synchronicity Theory (Dennis and Valacich) and Focus Theory of Productivity (Briggs and Nunamaker). Two main research questions are explored. (1) Do collaboration tools improve software inspection productivity? (2) Can feedback dimensions that significantly improve productivity be identified and incorporated within software inspections? The first research question is supported. In a detailed reevaluation of the Baan study, EMS inspections are shown to be 32% more efficient than paper-based inspections. During the subsequent period, the results were more pronounced with EMS inspections being 66% more efficient even controlling for inspector proficiency. Significantly more conveying communication than convergent communication occurs during inspection meetings. EMS inspections enable more deliberation, less attention for communication, and more attention for information access compared to face-to-face paper-based inspections. The second research question is explored. Surveys and analysis probe some previously unexplored feedback dimensions (review rate, inspector proficiency and inspection process maturity). Experienced inspectors are surveyed regarding process maturity, inspector proficiency, and collaborative aspects of inspections. Preparation and review rates are necessary but not sufficient to explain productivity. Inspector proficiency is perceived to be important and multi-dimensional. Participation by highly proficient inspectors resulted in 49-76% more effective inspections. Significant inspection process variations exist within mature development organizations. Based on theory and experiences, the SEMS inspection tool is developed and a quasi-experiment proposed. Initial results using the SEMS inspection tool are reported and suggestions made for future enhancements.