Credit Karsten Moran for The New York Times
Miguel Jimenez, a 20-year-old amateur musician in Appleton, Wis., had his eye on a classic Rickenbacker bass guitar last year. He searched the Internet and found Reverb, an online marketplace for musical instruments, which had the bass guitar he wanted. But the price tag was $ 1,600.
“It was way more than my credit card ever would have allowed me to spend,” recalled Mr. Jimenez, a sales associate at a local retail store, whose bank credit card has a $ 500 limit.
Yet that apparent hurdle soon fell away. Reverb offered an alternative, Affirm, a lending start-up that provides installment financing. Mr. Jimenez applied in a few minutes on his smartphone, his purchase was then approved almost instantly, and he got his guitar. He chose a 12-month plan, with monthly payments of $ 150.
Why would Affirm, a San Francisco start-up, make the seemingly snap judgment that Mr. Jimenez was a solid credit risk for a purchase more than three times the limit on his MasterCard?
“I wouldn’t know,” replied Max Levchin, Affirm’s chief executive. “Our math model says he’s O.K. Probabilistically, he’s good for the money.”
The new wave of digital lending start-ups all say they are bringing the new math of data science to analyzing whether borrowers will pay back their loans. So far, the upstart consumer bankers offer mainly alternatives to credit cards, unsecured personal loans of all kinds, and refinancing of student-loan debt.
The traditional approach to risk-scoring relies on a person’s credit history, as distilled in FICO scores. The newcomers crunch all kinds of additional data including social network profiles, bill payment histories, public records, online communications, even how applicants fill out forms on the web.
It’s a digital-age spin on banking’s basic rule: know your customer. The more detailed, data-driven portrait of borrowers, the start-ups insist, will allow them to offer credit to more consumers at lower costs — and still make a tidy profit. That, they say, is especially true for young people with scant credit histories — the tech-savvy millennials whose money-handling habits are already shaking up the banking industry, as I wrote about this week.
Affirm is one of the most ambitious new lenders, in terms of relying purely on algorithms and automation to make its lending decisions. Many of the other start-ups are very data-driven but people are used to ask applicants follow-up questions or to make the final call on loans.
Not so at Affirm. A person taps in his or her name, cellphone number, birth date and the last four digits of the person’s Social Security number. Affirm sends the applicant a text message of a short string of numbers, which the person types in to authenticate that he or she is the person at that cellphone number. Then, Affirm’s algorithms approve or reject the loan and purchase within seconds.
“It’s not magic, it’s math,” said Mr. Levchin, a co-founder of PayPal.
Affirm’s task is made easier because, while it is a credit card alternative, it is not a general-purpose credit card. Affirm’s installment financing is for the purchase of a particular good at a particular merchant. The more than 100 merchants using Affirm so far are heavily weighted to specialized retailers of home furnishings, clothing, jewelry, exercise equipment, bicycles and musical instruments. Their offerings tend to be pricey. And initially, Affirm has been most used by people in their 20s and 30s.
All that raises the question of whether Affirm’s credit-analyzing models are clever or whether borrowers with certain demographics buying certain goods are simply good bets? To paraphrase Mr. Levchin, It’s not math, it’s the market.
As a start-up, Affirm does not publicly disclose its financial results. But its early performance and prospects have impressed investors. Affirm has raised $ 325 million in debt and equity financing.
And Affirm’s internal figures indicate its installment loans cost less than credit-card loans. For borrowers with prime credit ratings, Affirm’s interest on a 12-month loan is 8 percent, compared with 16 percent in interest and fees on a credit card loan, according to the company’s analysis.
For subprime borrowers, the Affirm cost is 13 percent, compared with 19 percent for credit cards. For “entry” borrowers with scant credit history, the Affirm rate is 16 percent, compared with 24 percent for credit cards. A separate analysis by Affirm shows that its loan approval rates are also substantially higher across all three groups of borrowers.
The effectiveness of the new lending math practiced by Affirm and others is one issue. But the other issue is fairness. That challenge is underlined in a new report by the Federal Trade Commission, “Big Data: A Tool for Inclusion or Exclusion?” The commission report pointed to the consumer benefits that can flow from the technology, but also addressed concerns that “companies could use big data to exclude low-income and underserved communities from credit and employment opportunities.”
Affirm, for one, acknowledges that fairness has to part of its model. Mr. Levchin is on an advisory committee to the government’s Consumer Financial Protection Bureau, and Affirm’s general counsel is Manuel Alvarez, a former lawyer for the consumer protection bureau. The danger to avoid, Mr. Alvarez said, is “you seem to have a neutral tool, but its impact is not.”