Machine learning techniques such as artificial neural networks and support vector machines are by now widely known within the tech world. In contrast, other means of data analysis like subgroup discovery and local modeling mostly remain niche methods. However, they actually have a high value proposition in diverse scenarios from business to science and thus deserve more attention. Subgroup discovery in particular is a very flexible framework that allows to easily express and effectively answer practical analysis questions. In this post, I give an intuitive introduction to this framework assuming only a mild familiarity with machine learning and data analysis. In particular, I’d like to approach the topic from a somewhat unconventional but hopefully insightful angle: the idea that in data analysis it is often surprisingly powerful to selectively say “I don’t know”.