Towards a Rigorous Approach for Managing Process Evolution

Most software development projects are delivering late and over budget due to the fact that they are performing in an evolutionary and dynamic world. It has also recognized that unforeseen evolution is hard and costly but necessary to manage. This paper therefore describes the evolutionary nature of software systems and associated processes. In addition, we propose a framework for categorizing process evolution focusing on typical aspects such as origin, cause and type of process evolution, as well as how, when and by whom the change is performed. Evolution being classified by the framework is called evolution pattern. A set of typical project and product properties are identified to have major influence on evolution profile. An evolution pattern is recorded with detailed information on project context in which it occurs, product profile it develops and impacts on progress it implies. An approach is then presented guiding how to learn and reuse evolution experiences from previous and in future similar projects respectively. Together, they constitute a quantitative foundation for improving project estimates and process planning which is inspired by Quality Improvement Paradigm (QIP) and its Experience Factory [BCR94]. We are also conducting a case study at a banking software house to validate our hypotheses on improving estimates accuracy, productivity and product quality by application of the framework and approach. Realistic data on evolution, project and product profile are collected and systematically recorded for future usage in a local experience database.