Data really can give you crystal balls

When it comes to needing foresight, you can either consult a mystic or consult the data. We opt for the data.

Share this:

By: Neil Edwards on 27th August 2021, 3 minute read

Let me take you back to 16 March 2020 and the beginning of lockdown.

We were all told to stay indoors. Shops, bars, restaurants and other businesses closed overnight.

Days later, the government stepped in with emergency measures to keep businesses afloat, announcing the furlough scheme, business rates relief and the early iterations of the Coronavirus Business Interruption Loan.

Case Study: Peer to Peer Business Lender

How we developed a unique targeting approach and improved lead volumes from direct marketing by more than 80%.

None of us really knew what to expect.

Predicting the future

Business lenders had a double predicament to consider. They were brutally exposed to their customers no longer being able to make repayments on the loans they had outstanding. Interest income wasn't being earned and expensive capital wasn't coming back.

One of our clients turned to us for some insight on their lending book.

How many requests for repayment holidays could they anticipate? What was the propensity for their customers to fail? What level of provision for bad debt should they be considering?

We were aware of some lenders doing broad brush assessments by sector: if a customer was in non-essential High Street retail they were bound to fail, if they were in video conferencing, they would be planning their retirement.

We knew it needed to be more nuanced than that. There would be good businesses in badly affected sectors with the resiliance to survive, and there would be weak businesses in apparantly bullet proof industries for which the pandemic would be the final straw.

We set about building a proprietary "propensity to fail" model for the client and ran it over the portfolio.

The model incorporated around 20 different factors, including hard data such as time in business, availability of free credit balances and exposure to other lenders, and softer data like website updates and social media activity. All combined to create a current rather than historic propensity to fail score.

The finding...

The upshot was that we told our client not to worry. Their book looked to be in very good shape and while they could anticipate requests for repayment holidays, in most cases they should agree these in the knowledge that it was the business owners being prudent rather than desperate. All businesses with a high propsensity to fail were flagged so that special consideration could be given if a request was received and possibly a provision raised.

...and the result

Roll forward to today and our client has recently told us that they have suffered zero losses, have no customers on repayment holidays and only 50 on revised repayment terms out of more than 40,000.

The credit for the quality of the book undoubtedly goes to the underwriting teams, but it is pleasing to see that our forecasts were correct and that we were able to give the business the information it needed to manage the situation at the time.

If you would like to know more about propensity modelling, be it propensity to fail, propensity to grow, or propensity to buy, please contact us. All modelling is bespoke to the client and a specific set of needs.

Share this:

Neil Edwards

Author

Neil Edwards

Neil is a Chartered Marketer and Fellow of the Chartered Institute of Marketing with many years' experience in marketing, brand and communications.

CEO / The Marketing Eye

Related Reading