This introduction to the special issue on Business Process Intelligence (BPI) discusses the relation between data and processes. The recent attention for Big Data illustrates that organizations are aware of the potential of the torrents of data generated by today's information systems. However, at the same time, organizations are struggling to extract value from this overload of data. Clearly, there is a need for data scientists able to transform event data into actionable information. To do this, it is crucial to take a process perspective. The ultimate goal of BPI is not to improve information systems or the recording of data; instead the focus should be in improving the process. For example, we may want to aim at reducing costs, minimizing response times, and ensuring compliance. This requires a "confrontation" between process models and event data. Recent advances in process mining allow us to automatically learn process models showing the bottlenecks from "raw" event data. Moreover, given a normative model, we can use conformance checking to quantify and understand deviations. Automatically learned models may also be used for prediction and recommendation. BPI is rapidly developing as a field linking data science to business process management. This article aims to provide an overview thereby paving the way for the other contributions in this special issue.