Challenge: Debt Management Service (DMS), within the Fiscal Service at the Department of the Treasury, seeks to identify, collect, and resolve debts owed to government agencies, including those associated with state child support and delinquent student loans. DMS utilizes a number of passive and active collection activities to collect on delinquent debt. DMS aimed to apply optimal collection techniques to maximize collections given their limited resources.
Summit’s Approach: Summit integrated relational databases containing information on debts and collection activities. We then classified different clusters of debts based on debt and debtor characteristics. Implementing machine learning and regression techniques, we determined which debts would likely result in the highest collections using DMS' active collection tools and campaigns. The models also identified the optimal collection timing and holding period for conducting collection activities to maximize overall collections. The operational and financial impact of these changes in collection strategies were then estimated through the use of randomized control trials.
Result: The revised collection strategies led to an 7% increase in collections, increasing DMS’s revenue as well as reducing the costs of the DMS program to taxpayers.