Credit scoring saves 7% portfolio losses
The Business Challenge
Our client was interested in growing its consumer durable loans portfolio without compromising on credit quality. The inherent challenges were the small ticket size of the loans, the large volumes, and reducing the lead-time for the end-to-end credit decisioning process.
The bank had been using judgmental scorecards, but was now interested in implementing a statistical application scorecard that would provide a more accurate assessment of customer risk while reducing turnaround time.
The bank’s Personal Finance Services (PFS) division engaged Fractal Analytics to develop the statistical scorecard for the consumer durable loans business. Various sophisticated classification techniques were used to build a scorecard to predict the default behavior of customers. Application and performance data for past applicants (including rejected applicants) served as predictive data for the model.
The “score” obtained by a customer indicated his or her probability of default. In close consultation with the bank, our analysts also developed an optimization tool to help the bank decide on a “cut-off score” based on the bank’s risk appetite. The cut-off score was determined by considering the potential credit loss for accepting a “bad” customer vs. the opportunity loss of rejecting a “good” customer.
Since the model provided a probability of default to each customer, it also offered the bank the flexibility to alter its cut-off scores to meet its business objectives at different points in time.
The cut-off score could be fixed to meet a pre-determined level of approvals, to restrict defaults to an acceptable level, or to maximize profitability.
“Fractal’s services have been valuable in our endeavor to manage optimal levels of risk on our portfolio,” said the VP of the bank’s Retail Risk Department. “While judgmental scoring models help, statistical scoring models leverage the wealth of data present in an organization to provide a superior risk management tool –– a percentage point reduction in loss rates can significantly change profitability for a financial services company.”
Fractal Analytics’ application scorecard helped decrease defaults by about 50 basis points while still maintaining the same approval rate.
This resulted in savings of about $95K on a portfolio of $15M.
The bank now operates at a desired level of risk based on its risk appetite.