An investigation of the utility and value of process patterns in the management of software development projects.

Pattern theory has engendered much controversy in the field of architecture; yet it has brought new insights to the field of software engineering. Patterns continue to play an important role in software engineering in general, and in software development in particular. In this study, two preliminary surveys, focusing on the two fields of architecture and software engineering, were carried out to investigate the role and effect of patterns. The surveys indicate that while, patterns are unpopular within the architecture community and are criticised for stifling creativity, software patterns are popular within the software community and a high proportion of software development companies use them in their development practice. The results however show that in the vast majority of cases, pattern usage is limited to design-based problems, involving a single type of pattern (i.e. design patterns). The results further show that process-based patterns are seldom used in the software development industry, which prompted the topic of the main investigation of this research to evaluate the effect and utility of process patterns.
A controlled experimental research method was designed and used to evaluate the utility and value of process patterns in the management of software development projects. In this '2x2 factorial design' experiment, the subjects were divided in two groups of experimental and control, where the experimental groups were given a set of process patterns to use in their software development projects. Overall, there were over 750 subjects involved in this experiment and a total of 260 software development projects (individual and group projects) were investigated. Measurements of a number of appropriate software attributes were taken during the life of the projects though a devised goal-based measurement process. A further number of attributes were measured after the projects were completed. Using metrics, a number of software attributes across the four major phases of the development lifecycle (i.e. Requirement Analysis, Design, Implementation, and Delivery) were measured and statistically analysed. In addition to these specific measurement data, official marks awarded to the projects by the tutors were also used in the analysis. The objective was to determine if the experimental groups produced software projects that were of higher quality, in terms of the measured software attributes, than the control groups.
The experiment results show that, in the case of thirteen measured attributes, the treated groups scored significantly higher than the control groups. The improvements are across all the four major development phases, with at least two attribute in each phase, showing significant improvement. The experiment, therefore, confirms that the application of process patterns in software development projects, improves the quality of the projects in terms of a number of specific attributes such as productivity and defect density. The results further show that the treated subjects in the group projects performed significantly better than those in the individual projects. This, therefore, confirms that while the application of process patterns significantly improves the quality of both group and individual projects, the improvement is more prominent in the case of team projects. Process patterns are thus shown to be more effective on team projects in improving the quality of software development projects.

Thesis Submitted for the Degree of Doctor of Philosophy at Middlesex University.