An organizational learning model of the software development process

Summary form only given. The subject of this paper is a quantitative model for incremental change effects in large software development projects. In particular, we would like to measure and predict two sources for incremental change in software. The first is bug reports submitted by users and developers. While most bugs are relatively mundane, some of these bugs can prompt major redevelopment. The second source is enhancement or feature requests. As with bugs, such requests can drive the development team to make changes (major and minor) to the software, possibly at great expense and delaying time to the next software release. Here we take an organizational learning approach to modeling the software development process. The rationale for applying a learning model is fairly straightforward: the team works learns how to construct better software by solving design and programming problems. By studying the rate at which an organization learns, we might be able to predict its performance for new development projects. The fundamental idea presented here is that any learning process is governed by two factors: the set of potential learning opportunities available at any given time t; and the rate ρ(t) at which opportunities are discovered and exploited by the team at time t. The model has been validated using bugzilla data from a number of open source development projects