Now that my bait-and-switch title has lured you here, I would like to call your attention to the art and science of decision modeling. Before you can make a data-driven decision you need to specify both the decision and the data, and doing that is what I’m referring to as decision modeling. Specifying the decision includes both the decision process (when is the decision needed and what steps must be taken to arrive at the decision) and the decision logic (what criteria must be met to arrive at a particular decision). Specifying the data includes the input data (what information must be available before you can make the decision), and the output data (what information is required to communicate the decision to whatever process, system, or person is going to act on it).
A bunch of clever folks (i.e. decision management tool vendors like FICO) are going nuts realizing they are missing out on lucrative markets because almost nobody knows how to do decision modeling. To address this pitiful waste they came up with a decision modeling standard called DMN (Decision Model and Notation: http://www.omg.org/spec/DMN/). DMN defines a diagrammatic approach to decision modeling. A number of tools that are based on it are available, including FICO’s DMN Modeler, shown below:
This shows a simplistic diagram modeling a cross-sell decision (i.e. a well-calculated approach to fleecing your customers without actually driving them into bankruptcy, which would negatively impact your future sales). DMN is relatively simple, in that there are only four types of nodes in a diagram:
- Input Data (the blue oval)
Denotes data used as an input by one or more decisions. This data represents the initial set of business terms on which the decision model will operate. Business terms are the subject of a future post (or book, doctoral dissertation, feature film, or opera libretto), so just think of them as the "things" you make your decisions about. In an actual decision service these things would map to the data that the service evaluates to make the decision.
- Decision (the pink rectangle)
Represents the act of determining an output value from a number of input values, using decision logic. Decisions take a set of business terms as input and return a business term representing the requested decision value.
- Knowledge Source (the one with the wavy line on the bottom)
An information node indicating the owner or author of the decision logic. This is useful for assigning blame when problems arise.
- Business Knowledge Model (the one with the clipped corners)
Decision logic (i.e. the business rules that specify how the decision is actually arrived at).
I will be discussing various aspects of decision modeling, using DMN, in upcoming posts. It is my hope (along with those of the aforementioned vendors) that you will come to see DMN as a valuable tool for working in the world of decision management. I make no guarantees that you will either get rich or save the world. However, if you are interested in seeing measurable improvement against your KPIs, then DMN is a good way to start!
You can try reading the DMN Specification linked to above if you are masochistic. Alternatively, here's a video about DMN Modeler: