I posted same to LinkedIn DMN Group. Please help if you can.
I’m looking for a single example, or a through lead, of how it’s best to decompose and group decisions for a model like the (public) attached. It’s easy to claim it has to be BPMN driven, but it’s not worthy, since it’s one of too many discrete and independent cases. So, ought to fool the model assuming all inputs are readily available. All variables are Boolean or Enumerated.
I broke it down to the following decision boxes: “Open Case”, “Diagnose LAN”, “Diagnose WAN”, “Perform field action”, “Refer Internally”, Refer Externally", and “Close Case”. In spite of my model’s modularity, it requires bidirectional links among some of the decision boxes.
This is a novel DMN use case that would open up an arm in telecoms and ICT incident and problem management, including automation. Thanks in advance to any brainstorm.
What is the outputs you are trying to create?
The output of each box is the following decision box. Is this a proper treatment?
At case opening, the agent needs to decide whether to suspect LAN or WAN problem. At either of the, there may be a decision to reset or replace unit in field, or to refer the case internally (to one of four specialist support groups) or externally (to telco or power provider). Only when the agent, or any of the referees, returns with a positive resolve, the agent would decide to close case.
Of course, my logical decomposition doesn’t rule out the fact that there may be cross links and/or bidirectional links back to a decision box that’s been visited before. It’s a logic and not an activity flow (per how I chose to model it) and hence I believe that moving across until the close case output is produced is semantically OK. Do you agree?
@Technopsis why is this too complicated for a mix of BPMN with call activities that use DMN for certain decision points?
I recommend that you use a process which invokes different business rule tasks. It’s a common pattern, called decision flow.
DMN doesn’t allow you to model this kind of decision tree. You can only model single decisions which can use nested decisions.