Implementing Knowledge Management in Agile Projects by Pragmatic Modeling

Abstract

Background: Team knowledge is diluted and destroyed through domain evolution and staff turnover. A challenge to any project, agile projects are particularly vulnerably as they rely more on tacit knowledge than plan-based approaches. Increasing project sizes and durations deteriorate this situation. Introducing documentation and modeling to turn tacit into explicit knowledge as exercised in traditional approaches is perceived as costly, and impeding with agility. Objective: We want to improve agile practices for large, long-running projects by adopting and adapting long-standing modeling practices, challenging these practices in the process. We aim to establish a more pragmatic view of what should be considered a model, and how complex system models could be organized to better support their usage. Method: We propose several additions and changes to existing agile practices, and a new notion of model. We highlight how models are used in industry, and how existing modeling languages and tools might be improved to better support these usage modes. Results: We have successfully implemented our approach in a large project. A transfer to a smaller, more typical agile project in a different environment is under way. Conclusions: Modeling can be a valuable and appreciated addition to agile development projects. However, this requires a pragmatic approach beyond the conventional wisdom of MDE and academic modeling practices. A broader view on what models and modeling are is useful in practice, and offers relevant new research questions