Interpretability is the top priority — your audience needs to understand *why* the model makes each prediction - You're building a preliminary model to explore which features matter - The dataset is small and a complex model would overfit - You need to explain the model to non-technical stakeholders