Two-dimensional score matrices have been around for a long time. Typically, credit bureau scores and an internal (often custom) score are used. These two-dimensional score matrices are used in marketing, origination, or account management to make decisions, with other variables or policy rules.

Combining two scores into a risk tier is straight forward:

1. Determine score breaks for each score.
2. Cross two scores into a matrix, using breaks chosen in step one.
3. Measure performance outcomes you care about (ex: bad rate).
4. Group the cells of the matrix into risk tiers based on similar bad rates.

Sound’s simple, right? See a demo of the FICO decisioning platform for how to solve these kinds of decisioning challenges. Now, let’s examine the pros and cons of this approach:

The Good

Simplification is good.

1. Provides a single measure for risk. Everyone understands “risk tier 1” vs. “risk tier 5”.
2. Feeds into other measurements. Some institutions start with bad rates and through an analytic process, determine forecasted customer values, risk and other measures (responsiveness, take-up, revolving behavior, etc.)

The Bad

What could go wrong?

1. The scores are binned into ranges. This gives less granularity when making decisions, removing some predictiveness of the scores.
2. Over time, risk tiers may become inaccurate. As scores degrade and the make-up of your portfolio changes, bad rates associated with the tiers may change. You’ll need to re-validate risk tiers.
3. If risk tiers are for adverse action, reasons might not be obvious, since it’s more difficult to determine which score led to the action.

The Ugly

What is the downside?

1. Risk tiers are dependent on scores, redeveloped over time. An embedded score means you need to redevelop your risk tiers every time you rebuild scores. For a new external score (upgraded FICO® Score), you’ll need to re-calculate, gain approval and maintain risk tiers.
2. The ranges may not be optimal for every segment of your portfolio.  Specific segments might benefit from different ranges or may not have the same distribution within them; thus, the bad rates may differ.

Is there a way to develop and analyze tiers, but ensure the right way to measure risk when developing a strategy? In this blog, you can see how to develop and use risk tiers. For new blog posts, follow the FICO Community blogs.