Introduction of a framework and tool for evaluating and improving software measurement programmes

This thesis deals with improving software measurement programmes. The original motivation for the research for this thesis was to compare the effectiveness of software inspections to testing in an organization that maintains mission software for military aircraft. The initial goal was to obtain quantitative evidence, by comparing the density of defects found during inspection to defects found during testing, to support the assertion that inspection effort reduces testing effort. Because of problems experienced during the initial study of defect data the research for this thesis is refined from its original goal of comparing the effectiveness of inspections to testing to a study into improving software measurement programmes. An approach introduced by Meondoca et al provided a means to improve an existing measurement framework. Their approach, however, is restrictive in that it relies on the Goal/Question/Metric (GQM) paradigm from which to rebuild the measurement framework. The GQM paradigm is not ideally suited for rebuilding a measurement programme for an organization such as the one studied for this research. Other approaches provide a framework for evaluating a software measurement programme, but do not provide guidance for making the improvements. In the absence of an existing comprehensive approach, a framework and tool are developed for this research that provide a generalized, tailorable, and scalable approach for evaluating and improving a software measurement programme without a quantitative evaluation. A software measurement improvement project conducted at a software engineering organization, resulting in numerous improvements to the existing programme, provides support to validate this approach.