Machine Learning (ML) methods (and Predictive Analytics, Artificial Learning, and so on …) are very popular but not often used in criminology. Perhaps partly because it has been argued that Big Data and ML make the scientific method obsolete, ML might be ignored altogether by criminologists who only think of it as a replacement of their current research practice. Because I think that these methods hold promising opportunities regardless of the current debate, I provide a gentle introduction to a few key concepts of Machine Learning. In particular, I discuss how a machine can ‘learn’ from data, and I discuss the type of research questions for which ML is useful as well as the bias-variance tradeoff. As a case in point of ML’s potential, I investigate which combinations of different types of businesses and facilities at very small spatial units are most predictive of crime. Previously having been studied using Conjunctive Analysis of Case Configurations, I show the suitability of Machine Learning techniques to study this particular question.