Designing a decision

Successful decision modeling projects begin by fleshing out all the elements that compose your decision: the diagram, the data model, and the decision logic.

Diagrams provide an abstract, high-level representation of how decisions and the data that is required to make these decisions are structured and related to each other. Creating diagrams is an iterative process where you decompose the decision that you want to make. There are two approaches to decomposing decisions:

Decision-driven design
The decision-driven approach works from an outcome that is positive to the business. You start with the decision that needs to be made and ask what data is needed for this decision. For example, a seller might want to determine the amount of discount they can offer a loyal customer. Starting from the decision "how much discount to offer", they then add node-by-node all of the data and other influencing decisions that they need to automate this decision.
Data-driven design
The data-driven approach starts with the data that you have and works out what part of the business is affected by this data. It asks what are the business decisions that the data influences. For example, a business selling fresh produce might have some data on the weather. Starting from the weather data, they might create a decision model to help them decide whether it is worth setting up their stand at an open market given fewer people are likely to go if it rains.

Both of these approaches are valid but the decision-driven approach is the recommended one: by breaking the end decision into smaller decisions, you end up with a simpler model that is easier to understand.

Diagrams provide an abstract, high-level representation of how decisions and the data that is required to make these decisions are structured and related to each other. They are composed of a set of nodes that are used as building blocks to represent decisions in a graphical way:

  • Decision nodes represent the end decision, that is the decision that you want to automate, and the subdecisions that the end decision depends on.
  • Data nodes represent the data that is needed to make a decision.
  • Function nodes encapsulate computations from other decision models.
  • Prediction nodes encapsulate predictions that you can call directly from your decision model.
Diagram with two decision nodes, one input data node, one prediction node and one function node
The decision logic is captured in the form of business rules and decision tables:
  • Business rules are if-then statements that are written with a syntax close to natural language that can be readily understood by business experts. If-then statements associate a condition (if) with an action (then). When the condition is met, the rule action is triggered. You create business rules by using a wizard that lets you select the criteria that you want the rule to cover. You write rules by using a guided editor that lets you assemble statements and add missing variables.
  • Decision tables represent decision logic as a table where each row corresponds to a business rule. You create decision tables by using a wizard that lets you select the conditions that you want to use in your rules. You use the special facilities that are provided by the decision table editor to work with decision tables.