Speed of response following a missed repayment is essential as is being able to segment or triage the risk to the bank to help deploy resources effectively.
Fundamental to an effective response is understanding whether this is a first-time situation, an occasional situation, or a persistently delinquent customer.
Banks can use a collection ‘data lake’ that provides a rapid analysis and response without the delays associated with processing and extracting highly verified data from a highly structured and regulated data warehouse.
We discuss in more detail ‘data lakes’ and how they can be used for this sort of purpose in Operations II. Essentially, the data lake is populated with data from source systems without having to go through the structured and slower data warehouse route. The downside of any data lake is that the source data must be of sufficient quality.
Once a customer defaults for the first time, their data is loaded into the data lake repository and collection record created. The data will include account balances, limits, days past repayment due dates, the current credit score or behavioural score, and other data concerning the debt.
The customer is scored to determine the priority and action. If it is the first time that the customer’s repayment is overdue, it will be a ‘nudge’ to remind them. Details of the action and its result (for example, payment) are captured by the collections data lake.
If the delinquency continues, the response to previous attempts to rectify the position are added to the algorithm to determine the next action.
And so on until the customer returns to normal repayment patterns or the debt is recovered.
After a period of say 24 months, customers who have not defaulted again are removed from the collections data lake. Should they subsequently default, the process begins again.
The data used to populate the data lake comes mainly from two systems:
It is crucial that historic data of previous collections activity for a customer is stored and used to inform current collections tactics and actions, rather than starting from scratch every time.
Collections Data Requirements
Minimum data should include:
It should be noted that financial obligations relate not only to loans, overdrafts, and credit cards, but also to services granted and not already paid for upfront by the customer.
Because the bank’s collection team is finite, and the number of customers defaulting can rise quickly (as happened when Covid-19 restrictions were introduced), customers are segmented according to historic and current data:
The bank must optimise the channel used, the channel timing, and messaging most suitable to individual customers to achieve these objectives whilst also knowing who will accept and make the repayment and who will not respond and roll-forward to the next stage.
In later-stage collections, the focus is more on recovering the debt and less on retaining the customer.
Data-driven analysis also helps improve collection performance by constantly re-assessing collection process efficiency and any associated human intervention process.
Active credit management requires daily reporting such as:
As mentioned previously, the collection method chosen must take into consideration the cost of such an exercise. Banks or financial institutions normally do not have portfolios large enough to support a long collection process. The cheapest and most effective way to collect after a certain period, normally less than 60 days, is to pass overdue customers to a collection agency.