How Smart Lending Dumbed Down

Editor’s note: Julia Kukiewicz edits UK-based, a price comparison service that provides guides, product reviews and comparison tables for consumers.

Between 2008 and 2010, Wonga was doing something mind-bendingly stupid: sending letters from fake law firms to borrowers behind on their loans. It sounds like a bad office prank.

They even took the firm names from apparently still current employees — except when you consider that they did it 45,000 times and that it will cost them £2.25 million to give each injured party just £50 in compensation, as they were ordered to a few weeks ago. It may even land them in court.

Wonga wasn’t alone in sending faux-legal threats; in the past few weeks it’s emerged that many big lenders and the Student Loans Company (SLC) have using similar tactics for years, but the payday lender was the only one to create their law firm out of thin air. In an under regulated market, Wonga clearly weren’t too concerned about having to explain themselves. But under regulation alone isn’t enough to explain their actions.

What does explain them is at the heart of their business and that of many other fin-tech lenders: prediction and collection.

Creative destruction

If ever there was a sector ripe for a little Joseph Schumpeter-style creative destruction, it was banking in 2008. Borrowing from mainstream lenders was, and remains, slow, confusing and difficult, particularly for people with poor credit histories.

Startups like Moven, LendUp and Zestcash in the U.S., Wonga in the UK and Lenddo and Kreditech elsewhere promised an alternative: smart lending delivered efficiently, with transparent charges.

Smart lending means big data. Really big. “All data is credit data,” the CEO of Zestcash said in 2012. “We just don’t know how to use it.” Wonga claims to use 8,000 data points to make a lending decision. Kreditech says it uses 15,000.

By data they mean all the traditional identity checks and indicators that a borrower will repay, plus everything they can trawl from the depths of the Internet: social media; Google searches; court records; and any indicators that can be picked up from the context of the application down to the time of day and whether the applicant uses Firefox or IE.

This extra data is the secret sauce of alternative lending. Some, like Philippines and Columbia-based Lenddo, emphasize that social bonds predict and encourage repayments. Others, like Zestcash, simply say they have a better mousetrap: a more reliable way to pick out the most responsible borrowers.

Every borrower is assessed by algorithm. As Wonga’s then chief executive put it in a Wired Money talk last year, “we’ve moved beyond anecdotal, judgment based decisions… we leave that to the machines.”

But somewhere along the way, Wonga’s predictions started turning out to be wrong.

The lender has always claimed to write off about 7 percent of their loans but in 2011, Companies House records show, they wrote off £76.8 million, or 41 p ercent, of their £185 million annual revenue.

A Competition Commission investigation into the payday sector earlier this year noted that default costs “make up nearly half of total industry operating costs, suggesting that differences in lenders’ ability to assess risk may have a significant impact on their ability to compete.”

Calling in debts   

Which brings us to the problem of calling in debts.

According to Bank of England figures, default rates were rising even before 2009. Over the past few years, all lenders have started to take collections more seriously. As credit reference agency employee put it in a 2012 paper, “the focus was always on lending. Now the focus is on collections and people are running very fast to do what they do better.”

Armed with the conviction that their data would predict defaults far better than the banks, Wonga may well have been underprepared for an environment where collection is as important as selection.

They may also have underestimated the negative effect of their outsider status. For borrowers with multiple debts, an outsider they have no other affiliations with, like a current account or mortgage, can fall low on the list of repayment priorities, even one charging such a high interest rate.

It’s telling that even big lenders felt the need to use letters from in-house legal teams to encourage borrowers to repay; they did it, they say, because letters with the bank’s own letterhead were ignored.

A lender that has overestimated its predictive powers and underestimated its need to call in bad debt is apt to do very stupid things. Like make up a law firm.

If other alternative lenders want to avoid doing the same they’ll need to stop treating big data like a crystal ball and start getting those creative minds to work on ways to collect repayments that don’t involve a subsidiary of Dewey, Cheatem & Howe.