Final Thesis: Predicting Inner Source Success: A Machine Learning Approach Utilising Open Source Proxy Data

Abstract: Inner Source promises to improve collaboration and code reuse inside organisations, but many initiatives still struggle because managers lack predictive, data-driven support. Existing platforms, such as MECOIS, mainly offer descriptive views of past activity and do not help answer whether a repository is ready for Inner Source or how adoption is likely to develop. This thesis addresses that gap by designing, implementing, and evaluating a predictive module for the MECOIS analytics platform. The module uses development data from corporate Open Source projects on GitHub as a practical proxy for Inner Source environments. The thesis defines a collaboration-based success measure that captures the extent of integrated work contributed by developers outside the core team. Several Machine Learning models are applied to estimate the likelihood that a repository will reach a highly collaborative state and to identify the structural factors most strongly associated with success. The resulting predictions and explanations are embedded in an interactive MECOIS dashboard that allows managers to explore different ‘what-if’ scenarios. Although available Open Source data limits the evaluation and cannot capture all organisational and cultural influences, the work shows that predictive analytics for Inner Source is feasible and provides a foundation for more data-informed decisions about when and how to open internal repositories.

Keywords: None

PDF: Master Thesis

Reference: Julian Maibach. Predicting Inner Source Success: A Machine Learning Approach Utilising Open Source Proxy Data. Master Thesis. Friedrich-Alexander-Universität Erlangen-Nürnberg: 2026.