48,000
accounts not processed each month
Results
80%
accuracy in identifying self-curing accounts
Benefit
The solution minimised wasted effort on accounts that would pay anyway. It also reduced the frustration and embarrassment for ‘good’ paying customers who then didn’t have a negative customer experience. Equally importantly it allowed for non-paying customers to be prioritised.
Context
Each month, this bank has a set of rules to identify which credit card accounts should be moved into an outbound collections process. Once in this process, the customer is contacted (called, SMSed, emailed) to request payment. About half of these clients intended on paying and only missed the payment deadline for administration reasons. As a result the collections team has a dilemma – start collecting on day one and risk irritating good customers and wasting resources on customers that intend on paying or wait a few days and delay collections processes for customers that need to be followed up on.
Methodology
Using data available at the time of moving an account into collections, predict who will pay on their own within five days (known as self-curing). This allowed the collections team to differentiate their process to immediately start collecting from customers that the AI solution identified as non-payers and to leave the others to pay in their own time. The data that was used included account holder, credit bureau and product information. It should be noted that credit card transaction data was not available for this solution. It is expected that this transactional data would substantially enhance accuracy of the solution.