Implementing an Experience Factory Based on Existing Organizational Knowledge

This paper describes the development of an experience factory in an Australian organization involved in the field of telecommunications. Faced with improvement needs, in 1998 the organization placed special attention on approaches to support improvement activities. Towards the end of 1998 a project was started between the company and the Centre for Advanced Empirical Software Research (CAESAR) to evaluate the experience factory/quality improvement Paradigm in the organization. The concept was to be evaluated as an approach to support local improvement activities and to realize benefits in the shortest possible time. The choice of experience factory was motivated by research projects such as PERFECT, and by the application of the concepts in the Software Engineering Laboratory and in Daimler Benz. The project plan was to develop the tools and techniques to improve the speed and quality of software development, and to enhance the transfer of process knowledge between projects and project groups. Another aspect of the project was that the organization recognized that there was considerable experience already existing and stored digitally within the organization. However access to that experience was problematic. The results of the work in this company showed that the collection and transfer of experience in the organization was more important, and had a greater improvement potential, than the establishment of a classic experience factory framework. This paper contrasts the classic experience factory framework with that used in this organization. The paper further analyses the reasons for the approach taken, the methods by which the approach was used, and the benefits of this approach over the classic experience factory approach for this organization given its particular improvement goals. The research indicates that implementation methods, goals, and tools for an experience factory will differ between organizations, and that further industrial validation of the methods presented in this paper is warranted.