Posted by Dr. Andrew Jennings, FICO Chief Research Officer and Head of FICO Labs
In my paper on the new imperatives for analytics, I wrote a lot about the importance of the decision model. Decision modeling is a fundamental technique for understanding and improving decision strategies. It models the relationships between the data you have, the decisions you need to make, the likely outcomes of those decisions and the ultimate goal, such as profitability.
All decision makers have a view of the way a particular business situation operates whether they recognize it formally or not. The great advantage of the decision model concept is that it makes that view explicit and ties it to an objective like maximizing sales or profit. It also makes clear the decision variables—like credit limit and interest rate or channel and price—and the constraints that need to be met, such as expected losses not exceeding some predetermined value.
At the core of this model are predictions about how customers will react to potential actions. Generally these are the models that determine the effectiveness of the decisions that get created. However, it is very common for decision makers and analysts to create their decisions with a complete absence of any formal understanding of these action-reaction relationships. Not surprisingly, this leads to lost opportunity.
Decision models are often used in optimization, but they have other values as well:
- First, adapting the parameters and constraints of the model for some new circumstance forms a sound way of creating new decision logic. You are addressing the change at the structural level, not editing some derived construct.
- Second, because it makes explicit the underlying relationships, it becomes much clearer what information should be tracked to understand if a strategy is playing out as expected. You can track results against initial expectations, rather than simply following a long list of metrics.
- Third, because we have an explicit objective, the decision model forms an excellent basis for the comparison and simulation of one possible strategy against another.