Actual Performance of Customer Risk Segments


Like many other lending institutions in emerging markets, this East African commercial bank focuses on underserved, mass market customers. To ensure it could maintain the quality of its portfolio, the financial institution used the same high-touch credit assessment process for every loan application. 

The institution used First Access to build an algorithm that would classify customer risk using only information available at the time of each loan application. The institution chose to divide loan applications into five risk segments.

A blind test would then help the institution determine the success of the segments and which action to apply to each segment going forward.

In the blind test, First Access scored all matured micro-loans disbursed over the previous six months, separating customers into five risk segments. The PAR1 values for each segment were distinct and increased from one segment to the next, as expected:

  Segment A: PAR1 of 1.00%
  Segment B: PAR1 of 3.53%
  Segment C: PAR1 of 9.97%
  Segment D: PAR1 of 22.42%
  Segment E: PAR1 of 26.78%

Since the algorithm was effective at segmenting customers according to risk level in the blind test, the bank could streamline its assessment process accordingly. The bank first chose to offer same-day approval for all repeat customers in Segments A and B, roughly 17% of all loan applications. Since the standard process took five to six days, this group of customers was approved 80% faster, which prompted positive feedback from customers and staff. As expected, these customers also performed substantially better than the rest of the portfolio. The institution has since increased the number of loan applications scored and its risk appetite, enabling same-day approvals for more loan applicants.

Overall, our work with this East African Bank has resulted in an approximately 20% reduction in underwriting time across its entire portfolio.