The introduction of IFRS 9 will mean focus is placed on pre-delinquency — treatment of accounts before they exceed limits or miss payments — to prevent a move from Stage 1 to Stage 2, which in turn can prevent an increase in provisioning levels. (Treating these accounts is also known as pre-collections or early warning.) This approach could help reduce the number of accounts falling into the formal Persistent Debt process when they reach 18 months. This pre-emptive intervention could also have a positive impact on the customer experience when they are facing financial issues.
Early warning identification of credit-stressed customers is not new, but it’s fair to say it has its challenges. Identified too early, it becomes harder to manage customer satisfaction if taking action, as the customer may not yet realise they are truly credit-stressed. But if credit problems are identified too late and followed by a line decrease, this has been shown to accelerate customers into a delinquent state without any measurable reduction of exposure at default. Those that get the combination of early signals of risk at precisely the right time and can deploy the right treatment benefit greatly.
Issuers have historically targeted a sample segment and followed a test-and-learn approach to establish what works in pre-delinquency treatment and what does not.
First Signs of Trouble
There are a host of metrics an issuer could use when trying to detect the accounts that are most likely to fall into financial difficulties, including more trend-focused data, transactional data and bureau information, with the latter providing a wider customer view. If accounts fall into the pre-delinquency subpopulation, reviewing their other products held is also key — issuers need to be aware of the potential knock-on effect for other business lines and the broader relationship.
Example metrics include the following, although the full list FICO recommends is more extensive:
- Change in % payments to balance over 3 or 6 months to determine if payment amounts are decreasing.
- Reaching highest balance levels.
- Overlimit for the first time or first time in X months.
- First use of cash or increase in usage.
- Recent limit decrease.
- Recent bounced/late payments.
- Recent direct debit cancellation.
- Sudden use of card after period of inactivity.
- < 18 months in persistent debt.
The chosen predictive metrics need constant review as they will vary in line with any economic changes.
Consistency with other account management treatment needs to be considered to prevent sending mixed messages and confusing the customer. For example, you don’t want to offer an increase or allow overlimit spending while also making contact relating to potential problems.
Another question issuers face is where this process sits in their business: customer service or collections. It generally resides with a customer service team, but when it’s with collections the early-stage collectors seem best placed to carry out the calls, and they need strong negotiation skills.
Actions can be a mixture of automated and customer calls, with an emphasis on the latter and in advance issuers need to be clear as to what they can offer. Examples include:
- Payment reminders including payment methods available.
- Account closures.
- Limit decreases or move of balances to other products. Decreases could be automated based on trigger events or happen during a customer call.
- Setting up a direct debit or a change in the amount collected via this method.
- Taking a payment.
- Credit education.
Managing Pre-Delinquency Calls
The impact on the area responsible for this has to be taken in accounts. We have seen call times last from 5 to 25 minutes, when the expectation was that the calls would be on the shorter end of this range. The team responsible need to be aware that any automated actions may result in an increase in inbound calls.
The tone is very important, as customer responses have often been mixed. The account in question is in order and you do not want to damage the relationship. However, we have found more recently that most customers are comfortable with the issuer trying to responsibly prevent future debt problems.
A bonus of having a conversation with a customer is the opportunity to make sure you have all the correct contact details in place. This type of exercise also highlights the scale of missing phone numbers. Contact preferences could also be taken at this point.
As agent calls are costly, the use of these for the right people is key. Customer education can be achieved via cheaper digital channels, which still allow for a conversation to take place if needed and allow customers to source additional information if they wish.
Testing the Approach
If this is a new initiative, we recommend testing the strategy on a proportion of those who would qualify to gauge the impact and benefit from the learnings.
Due to the difficulty in identifying the right accounts to contact, issuers may consider a data-driven approach and incorporate into their strategies models predicting the propensity for consumers to roll from current to 1 cycle. The data-driven approach may save several iterations of judgemental strategies and accelerate the learning process. Once there is enough data on the impact of the actions taken, you can apply analytic optimization to ensure the best action is taken for each consumer at the right time, through the most appropriate or preferred channel.
To discuss this topic in more detail, please contact the Fair Isaac Advisors team. We can provide further recommendations around metrics, approaches and implementation for this type of strategy. You can reach me at firstname.lastname@example.org.