Service sector firms are heavily reliant on the use of Enterprise Resource Planning (ERP) systems to manage business processes. But how all of these are used in a day-to-day basis across teams and offices is still something of a black box. In this paper, we investigate real-world usage patterns of an ERP using the SALT dataset, which consists of millions of records from a service-oriented ERP system. We correlate data in great detail and apply clustering methods to identify hidden behavior patterns that look at how sales offices functions and how often the different document types are operated, and other data trends over time. We find that a small number of sales offices carry a large majority of the ERP activity, some document types and shipping terms rule the system, and the ERP activity follows weekly and season trends. Second, we cluster the usage of ERP using machine learning to better understand and improve ERP configuration and use. Service organizations can benefit from these results to optimize efficiency and personalize their ERP systems according to user behavior.