Increasing competition forces business organizations to improve the efficiency of their operational business processes, certainly where costly physical resources are involved. By integrating real-time, IoTbased information from these resources into business processes, advanced real-time decision making can be realized to enable the required efficiency increase. There are various challenges though. Firstly, the resources can be large in number, heterogeneous in nature and owned by different business parties. Secondly, the data is typically heterogeneous in format and large in volume. Thirdly, business scenarios are diverse and evolve over time. Consequently, converting IoT data into usable information to drive business processes is not a trivial task. To address this, we propose the use of a novel combination of existing technologies in distributed analytics (DA) and business process management (BPM). To deal with the size, heterogeneity and ownership of data, we don't bring the data to the analytics, but bring the analytics in a distributed format to the data. We use parameterized micro-services that are packed into software containers to make them dynamically deployable from a service repository into the IoT edge. To deal with the number of IoT resources and the diversity of scenarios, we automate the deployment and management processes of the containerized microservices using a BPM engine. This engine interprets graphically specified process models that define the data flow between the DA modules and business decision making. Our approach leaves large amounts of raw data at its origin and is highly flexible in its data processing scheme. We show the feasibility of our approach in a proof-of-concept prototype implementation.