Splitting decisions into sub-decisions

The decision diagram to determine a specific customer's discount has many inputs. To avoid big decisions like these, the bulk of information can be divided into sub-decisions. Finding the correct level of abstraction is an important success factor for decision diagrams. When deciding whether to split a decision consider:

• Complexity
• Reusability
• Authority 

The decision diagram to determine the discount is now split up into one top decision, and three sub-decisions. This makes the decision diagram easier to understand, as the dependencies and input values are clear.

This also makes the decision logic in the decision tables easier to maintain.

When expressed as sub-decisions, each decision can be changed independently, without changing the other decisions.

Once a decision has more than seven inputs and/or sub-decisions, it is clear that the decision logic will most likely be very complex. To simplify that decision can be split into two or more decisions.

The decision Determine discount has six inputs. Because of the large number of inputs the decision diagram looks complex.

Sometimes sub-decisions can be reused in other decision models. Some examples of reusable sub-decisions are verify status, validate contract, and customer status.

The sub-decision Customer status is used in two decision diagrams. In the first, the sub-decision Customer status serves as an input for the decision Calculate discount. In the second decision diagram, the same sub-decision is an input for the decision Determine priority of shipment.

DMN knowledge sources are used to show the policies or regulations that influence and/or guide a decision. It could be that part of the decision is based on an internal policy, and another part depends on an external regulation. In this example, splitting up the decision into sub-decisions could help make the decisions inputs and their origins more transparent, and easier to manage.