Debtor level collection operations using Bayesian dynamic programming

After a borrower defaults on their repayment obligations, collectors of unsecured consumer credit debt have a number of actions (e.g., telephone calls, formal letters,) they can take to secure some repayment of the debt. If these actions fail, collectors could seek legal proceedings. The operations management challenge in this setting is to decide which of these actions to take, how long to take them, and in what sequence to take them. Ideally, this collection policy should depend on how the defaulter has been performing during the collection process so far. In particular, it should take into account how many payments the defaulter has made under the current action, compared with how long that action has been tried. Other potential considerations aside, the objective of a collections policy typically is to maximize the recovery rate, i.e., the percentage of the defaulted debt that is recovered in the collections process. In this paper, we use a Bayesian Markov Decision Process (MDP) model to find an optimal policy of what action to undertake in the next period given the current information on the individual debtor's repayment performance thus far. The proposed model will be empirically validated with data provided by a European bank's in-house collections department. The model will be able to use by banks to decide their debt collection strategy.