Is your credit bureau score as predictive as it can be?

Does your underwriting criteria diminish the predictive power of the credit bureau score?  In many developing countries, best practices for scorecard implementation including the process of validating credit bureau scores against actual performance are not common place.  Applied Business Intelligence Group recently undertook a custom modeling project for one of the largest financial institutions in Mexico. This particular institution was relatively conservative and their underwriting criteria successfully eliminated many potential "BADs" but at the same time reduced the predictive power of the credit bureau score by 60%.

Why does this phenomenon happen?  Credit bureau scorecards are developed across a diverse universe of credits issued by multiple financial institutions. A lender’s credit underwriting criteria define a subset of this diverse universe, which will result in a target population that is either more or less risky than the  universe as a whole depending on the lender’s risk appetite.  It is not surprising that the predictive power is diminished within the lenders target market due to differences in the between the universe and target population.

What was the solution?  ABIG has developed enhanced credit bureau based scorecards that typically increases the predictive power by 50% over the credit bureau score. Using our DataDynamics product, ABIG was able to effectively develop over a thousand credit bureau characteristics that were compared with actual performance to identify the characteristics that best predict ”GOOD” or “BAD” performance within the target population.  The chart above demonstrates that the credit bureau score is under predicting the credit risk in the lower score ranges while over estimating credit risk in the higher score ranges.  Also, the chart illustrates the improved predictive power of the enhanced credit bureau scorecards across all score ranges.