Asset Publisher

Breadcrumb

How a Credit Union Leverages a Propensity to Pay Model with Delinquent Loan Collection



Propensity to Pay Model For Delinquent Loan Collections

Solution: Analytics, Predictive Analytics

Industry: Finance & Banking

Project: Propensity to Pay Model For Delinquent Loan Collections
Client: Credit Union

A major credit union located in the Midwest offers a variety of different loans and lines of credit to its qualified customers.

Challenge: Current Loan Collection Strategy Prompts Evaluation

As some loan accounts become delinquent, the credit union was executing a mass phone based contact strategy to collect past due amounts.  While direct phone calls were somewhat effective, the cost of calling every delinquent account became an expensive effort for the credit union to maintain and resulted in a lower ROI.  The credit union theorized that not all accounts required a live phone call for collection to occur and that perhaps a a more targeted collection strategy, supported by data science, could be used to prioritize its collection efforts.

Based on its exisiting relationship and its extensive data science experience, Dunn Solutions was contacted to help the company identify its riskiest deliquent loan accounts, so that a more effective collection strategy could be implemented.

Solution: Dunn Solutions' Data Scientists Develop Propensity to Pay Model to Segment Accounts

Dunn Solutions' data scientists started by gathering historical customer account data and recognized that 87% of delinquent accounts made a payment within a 30-day window.  The objective was to develop a propensity to pay model which would identify those accounts that most required a phone call contact in order to generate a payment within a 30-day window.

Dunn Solutions' data scientists initially evaluated over 150 different features to be included in its propensity to pay model. This list eventually was narrowed down to 19 key variables that were considered to be the best to predict future payment behavior. 

Dunn Solutions' data scientists used those variables to run hundreds of iterations of the propensity to pay model and tested the relationships between the variables to fine tune the most effective way to rank and segment customer accounts based on an associated score.  The scores were to be used by the credit union to test viable collection strategies for each divided group of accounts.

Result: Propensity to Pay Model Leads to Targeted Loan Collection

Based on the model, it was determined that significant cost savings could be realized by applying more expensive collection methods only to its riskiest customers. Dunn Solutions is working with the credit union to A/B test different collection methods for each group of accounts. Further evaluation and modeling will be conducted as the credit union rolls out a more optimized strategy to improve collection effectiveness.