Analytics Meets The Business Process


Ultimately we need to use analytics to enhance business processes. These in turn are designed to optimize the efficiency and efficacy of operations within the organization.. While analytics may seem like a world away from the design and management of business processes, it will rapidly become a core enabler of improvements in operational activity.

The business process management (BPM) movement gained significant momentum during the last decade and settled on an approach encapsulated in BPMN (Business Process Model and Notation). Processes are represented by diagrams, and a variety of methods are used to make sure the whole thing hangs together, and no undesirable side effects have been created. Embedded in many processes are decision points, although traditional BPM does not support the definition of such things. Clearly there is a need for such decisions to be coordinated (a person cannot be offered payment alternatives if an application for credit is rejected in the first place), and BPM does indeed provide a framework for this. However the internals of the decisions are not represented.

Analytics on the other hand is primarily concerned with building models which can be used in a production environment. By analytics I include predictive analytics (employing data mining technologies), prescriptive analytics (the optimization of resource usage), business rules engines (for deterministic business rule deployment), and business intelligence to analyze historical performance (usually termed descriptive analytics). Since the resulting models must eventually be embedded into a production environment, it makes sense that their specification should be added to that of business processes. To this end a new standard is being proposed, which is known DMN – Decision Model and Notation. It bridges two worlds – that of the business process, and that of the analytics model (which may be specified in PMML – Predictive Model Markup Language). It does this by allowing specification of a decision in a form easily understood by both business people and the technical people who build the models. Roughly speaking it allows specification of the inputs and outputs a decision requires.

There has recently been an advisory from the Object Management Group (OMG) that users of BPMN should start to consider using DMN to incorporate decisions into business processes. This is all good news, since creating decision models is something of a Wild West Frontier at the moment, and the whole thing needs to be integrated with existing business process modeling. There will of course be resistance, but the only alternative will be decision model anarchy, where decisions are uncoordinated and unknown to business units that are actually employing them. It is also a meeting of two worlds – the disciplined, structured world of business process management, and the exploratory, innovative world of decision modeling. The former is largely deterministic, and the latter probabilistic. DMN plans to bridge this gap – we wish it luck.