How lenders are leveraging transaction data to mitigate material issues in credit models post-COVID
By: Michael Mifsud
Credit risk models are some of the most important within a financial services business, fundamental to both profits and long-term bank stability. They are also critical elements to customer experience and growth, ensuring that those who represent good credit risk receive finance swiftly (and from you, rather than a competitor).
In Australia, there were already rumblings as to the deficiencies of these models before the financial impost of the COVID-19 pandemic.
But as the pandemic passes, the benchmarks and heuristics that underpin a broad swathe of consumer credit models have become increasingly irrelevant, reliant as they are on data that in some cases can be more than a year old. In today’s economic climate, what was true of a customer even a month ago may not hold today.
Customers who were previously good credit risks may be dealing with markedly changed personal and financial circumstances for which even the most foresighted could not prepare. Repayments will be late or missed, and income and expenditure will vary substantially from historical levels.
With such material disruption, how can we understand which applicants are still employed and likely to stay employed? How can we differentiate those who can sustain lower levels of income from those who can’t as easily manage their expenditure or those who have deferred payments that will hit them hard in months to come?
Existing benchmarks for income and expenditure will not suffice. On the income side of the ledger, the proverbial ‘three recent payslips’ won’t necessarily identify where income has changed substantially over the last month or last fortnight, and how that sits in the broader context of a customer’s peer group. When it comes to expenses, defining a new ‘normal’ becomes even more complex. Consumer behaviour during the pandemic is not reliable, and behaviour in the wake of the pandemic is likely to be altered completely by the experience. We know from previous downturns that shifts in consumer expenditure last long after the recession has ended.
Keeping up with changes in a borrower’s monthly or fortnightly income vs. expenditure as they happen and making sense of what it means for their future creditworthiness is crucial.
The only way to accomplish this is by leveraging insights from high-velocity transaction data to give an accurate snapshot of each customer’s unique situation – a real-time “profit and loss statement” for each customer, that can distinguish customers who would otherwise look relatively similar under traditional approaches.
For example, using Q.Refinery to curate and categorise expenditure data, we’re able to capture minute details from transactions belonging to a customer who, according to traditional measures, appears creditworthy, and demonstrate that within the context of a generational downturn they present an unacceptable level of risk.
Likewise, as normalcy resumes, we know that a customer diligently paying their utilities, regularly buying their groceries and looking after their pets is a better credit risk than a customer eating and drinking out, riding frequent Ubers, and spending large amounts of money on fuel but little on car insurance.
We might offer guesses as to what each of these attributes means, but a data-driven approach does not seek to discern or discriminate. It simply identifies changes in personal and financial circumstances at a unique customer level and quantifies the implications for the applicant’s risk profile.
These red flags in transaction data are more critical now than ever. The pandemic means that banks who ignore critical signals in transaction data could incur significant costs from bad debts or materially underestimate the provisions that need to be held; potentially endangering tens of millions of dollars in profit depending on the size of the bank’s loan portfolio and security quality.
The argument for modelling credit risk using insights surfaced from transaction data was compelling even before the pandemic. The technology is well-established and proven, and the benefits, in terms of improved customer and business outcomes, speak for themselves. But the financial risk of not embracing this approach becomes significantly greater the farther we drift from any concept of normality.
How can we help?
Our product, Q.Refinery, transforms transaction data to an enterprise-grade structured data asset, giving lenders access to powerful insights based on up-to-the-minute transaction data. Our global analytics consulting team, with deep expertise in delivering value from banks through data, is already working with banks around the world to solve these problems, and we’re uniquely positioned to deploy Q.Refinery within your organisation to protect your customers and your bottom line through this upheaval.
if you’d like more information, please feel free to contact me, and I will be happy to connect you with the relevant lead in your market.
About the author
Michael Mifsud has over 20 years of analytics experience in the financial services industry driving profitable growth through better use of data. He has strong technical expertise in pricing strategy, workflow automation, and credit risk modelling across the credit lifecycle from marketing, underwriting and decisioning through to portfolio management, collections, and recoveries.