Abstract
Most practitioners and researchers agree that when developing software, process affects product, and the usefulness of a process depends on the context. However, which processes are most useful for a specific company or project is generally unknown. When studying the relation between context, process and product, one challenge is that experiments often lack realism, which makes the transfer of results to industry difficult. In contrast, most of the important factors vary beyond the researcher’s control in case studies, which makes it difficult to identify cause and effect relationships. This paper reports a study where realism was combined with control over certain context and process factors. Four companies developed the same system, and the price varied by a factor of six. Certain patterns of relationships were expected (expensive company, low cost, schedule overrun); others were unexpected (cheap company, maintainable system because of small code). The community needs to identify the most important relationships among process, context and outcome.
