Skip to main content

Lending in the time of Coronavirus

Using Open Banking to maintain affordable lending during Covid-19

by Phoebe Allen15th May 2020

As a result of recent global events, many people have experienced (or are anticipating) a total or partial loss of income, which is likely to significantly affect their ability to repay credit. This change in circumstances will take a long time to be reflected in a traditional credit file – up to two 45-day update cycles. In addition, totally new economic circumstances mean that existing credit scores are unlikely to be predictive of default: historic repayment is now less indicative of repayment in the near future.

To serve their customers fairly, lenders’ challenge is to continue lending where it is affordable whilst accurately detecting, and accounting for, recent financial distress. Open Banking data is key to this: as of today, is the applicant still working? Are they on full pay, furlough or have their hours been cut? What impact has this had on their spending, borrowing and repayments?

For the majority of our clients, it is not feasible to manually analyse this information for every applicant. It’s also challenging for clients to construct these features and rules since even fairly binary indicators (such as job loss) have many complexities and edge cases.

Credit Kudos has worked as quickly as possible to build a range of out-of-the-box features that can be used to create new policy rules - helping our clients to continue lending where it’s fair and affordable. These indicators needed to be simple enough for our clients to deploy decisioning updates quickly, but sophisticated enough to help them implement a unique, nuanced ruleset that suits their policies.

All of our Covid-19 response features are part of our core solutions, so they can be implemented by all of our clients with no extra cost and minimal integration effort.

Partial or full income shock

The new Recent Income Shock indicator helps lenders to automatically detect a full or partial loss of income in the last two months.

To detect an income shock, we look at each source of income in turn, analysing the transactions’ amounts and dates.

For detecting total loss of income, we compare the dates of recent transactions to the frequency of the payment. To detect a partial loss of income, we compare the amounts of recent transactions to those earlier in the year. The more consistent the payment is in general, the more sensitive we are able to be to a small shock. Small anomalies such as expense reimbursements are taken into account automatically.

Each source is given a flag for Recent Income Shock that returns full, partial, none or unknown. For sources that have experienced a recent shock, the indicator also surfaces its magnitude - in terms of its impact on their overall income.

Insights are also “rolled up” into one overall metric, meaning lenders can choose between a single easy-to-implement threshold or a more nuanced set of rules.

Lenders can easily build decision rules to:

  1. Progress applicants who have not experienced a recent income shock, enabling continued lending to to affordable applicants
  2. Rule out applicants who have experienced income shock, reducing the risk of unaffordable lending
  3. Progress, refer or rule out applicants based on the magnitude and impact of a shock. For example, a lender may be able to still affordably lend to an applicant who has been furloughed, or an applicant that has had their hours cut but has an additional source.

Employment Industry & Employer

Our clients wanted to be able to automatically identify applicants employed in a number of “key worker” industries so that they could be fast-tracked or funnelled through a specific decision policy. Doing this automatically - avoiding requirements for payslips to be submitted - was also important for applicants, who are at present are likely working long hours in very difficult circumstances.

The Employment Industry label combines multiple dimensions of Lotus, our classification engine. For income sources - detected by our machine-learning Income Classifier - we harness Lotus’s Named Entity Recognition technology to extract the employer name from the payment and look it up in our dictionary of categorised employers and employer keywords.

We do this for the following key industries:

  • healthcare
  • transport
  • social care
  • government
  • police
  • utilities
  • fire and rescue
  • public services
  • justice system
  • logistics
  • armed forces
  • education and schools
  • prison service
  • security
  • pharmacies
  • retail essential
  • food essential
  • journalism

In some cases, we will also label the Employer, such as “NHS”.

This approach gives us very high precision and predictability, since transactions are mapped to a human-verified entity. Credit Kudos uses advanced Natural Language Processing technology to extract an “entity” from a transaction, which has excellent coverage - much more so than a rules-based approach. It also allows us to focus efforts on the most common industries in the list: healthcare, education and government.

Recent Borrowing

Lenders can now get information on Recent Borrowing the number and total value of loans the applicant has taken out in the last 30 days and 3 months.

This provides an important updated view of an applicant’s credit commitments, which would take 30-90 days to surface on a traditional credit report.

This is particularly useful at a time in which many people are having to use different types of credit to what they’ve used in the past: historical data is less relevant and recent data is critical for an accurate picture.

If you’d like to find out more information on how our latest features help lenders continue to lend where it is affordable, please reach out to our team here.