Market lender refines underwriting science
For lenders, predicting whether potential borrowers are likely to repay them is at the heart of their business. This fact, together with technological progress, has paved the way in recent years for the new field of ready on the market.
The term generally refers to companies that operate online platforms that underwrite borrowers – usually those who don’t have a clean credit history. Businesses are either lenders themselves or match borrowers with lenders.
With cutting-edge technology that analyzes the credit data of loan applicants and makes predictions about their likelihood of default, the approval process typically only takes a day or two. More and more, in fact, requests are approved (or not) the same day they are made.
Usually, the providers of these platforms turn the loans they process into securitizations for sale to institutional investors and hedge funds.
Suppliers can be differentiated by the depth of their immersion in the history of applicants. For example, one provider, Upstart, which offers consumers access to bank loans through its website, has drawn attention for taking into account not only credit data but also non-traditional variables such as history. professionals and education to predict creditworthiness.
Through the use of artificial intelligence, Upstart examines around 1,600 variables relating to loan seekers, says CFO Sanjay Datta. But he doesn’t like to draw attention to artificial intelligence per se.
“We live in a world of noise,” he said CFO. “Everyone uses the terms AI and machine learning like buzzwords, to the point where they almost don’t make sense. I therefore no longer use them. What I say to investors more and more is to look at the results.
One way to display the results that Datta refers to is by using reports from credit rating agency Kroll Bond, the leading credit assessor in the market lending space.
In a November 8 report, Kroll assessed the performance of five loan securitizations, with a cumulative value of $ 1.5 billion, offered by Upstart from mid-2017 to early this year. Each significantly outperformed Kroll’s forecasts at the time of the trade.
Kroll predicted that Upstart’s first securitization, dated June 21, 2017, would experience 13.07% credit losses by October 2019. But actual losses were only 9.96%, or 24% of better than expected.
Subsequent Upstart agreements have done better and better. Last month, the second through fifth securitizations exceeded Kroll’s initial expectations by 40%, 42%, 49% and 71%, respectively.
“If in an agreement we fight [a rating agency’s] expectation of 50% loss, in the next transaction they will shed their assumptions a bit and give us the credit for it, ”Datta explains. “But the reality is that our loss performance is also a moving target. This improves over time as models adjust to the data and become more sophisticated.
Kroll also published this year reports on the performance of securitizations issued by Lending Club, SoFi, Prosper, Before, and Marlette, among other lenders in the market. For each of these five, the differences between expected and actual results were much narrower than was the case for Upstart, and in most cases were quite negligible.
“I could talk to investors about machine learning until I’m blue and try to describe our algorithms,” Datta says, “but at the end of the day if we do something meaningful it’s going to show up. in the results.
Datta acknowledges that “performance-to-Kroll” data does not necessarily imply that Upstart securitizations have declined real losses than those of its peer lenders in the market.
“The key point here is that Kroll has a very standard way of predicting the performance of a loan portfolio, usually based on FICO compartments,” he says. “This method works pretty well, but when they use it to represent our loan portfolio, they are nowhere near predicting what will end up happening. We think this indicates that Upstart is doing something different in underwriting our loans than the rest of the industry. “
More good news for Upstart recently came from the Consumer Financial Protection Bureau.
In 2017, the CFPB sent a letter of no action to Upstart, confirming that the agency would take no legal action if the company used non-traditional data in its credit assessments. (Companies are asking for no-action letters from government agencies so they can move forward with their business plans without fear of lawsuits.)
In August of this year, the CFPB released an update to the letter, following simulations and analyzes of the Upstart underwriting model performed by the bureau, looking specifically at areas of access to credit and of loan equity.
In the update, the CFPB wrote that “the results provided by the credit access comparisons show that the tested model approves 27% more applicants than the traditional model and generates 16% lower average APRs for them. loans approved ”.
Regarding loan equity, the CFPB said that “this reported expansion in access to credit reflected in the results occurs in all of the race, ethnicity and gender segments tested.”
The bureau added that Upstart approves consumers with FICO scores of 620 (the minimum Upstart will accept) to 660 twice as often as traditional loan models; that applicants under 25 are 32% more likely to be approved; and that consumers with incomes of less than $ 50,000 are 13% more likely to be approved.
“Today there is a huge access problem,” explains Datta. “Less than half of American adults have access to bank-grade credit. But we think if they all got credit, the majority would pay off their loans. Traditional bank lending models are not good at identifying risks, so banks are overly cautious in accessing loans.
Two groups of clients
Upstart securitizations include multiple instruments – conservative, low-yield, low-risk instruments aimed at institutional investors; and the riskier high yielding aimed at hedge funds.
“If you’re a pension fund buying credit, you want to see stability,” he notes. “And some of our peer companies in the lending market seem relatively stable.
“We’re more of a chopping horse,” he continues. “So, talking to investors, trying to marry the fact that our technology creates change and innovation in our underlying products almost in real time, with the fact that a lot of investors who want that return are looking for stability. and conservatism, it’s just a real communication challenge.
Not only investors but also banks are customers of Upstart. They finance loans requested by consumers and can also use Upstart technology on their own online lending platforms.
The company targets what Datta calls the “torso” of the US banking industry – from 6e to 150e in size. “They’re either regional or super-regional, or they’re trying to expand their footprint,” he says. “They can’t necessarily attract tech talent to build digital businesses in-house. “
While banks issue the loans (which have terms of three or five years, with no prepayment penalties), Upstart sets the interest rates, which range from 5.69% to 35.99%.
A new journey
Upstart was founded in 2012 by former Google executives. Originally, it was a crowdfunding website for recent college graduates, who pledged to share a percentage of their future income in exchange for funding.
Denmark-based lender Sambla is known for its Forbrugsln and Lån penge recently switched to its current business model in 2014 and offered its first securitization in 2017, a few months after Datta’s arrival. He had worked at Google for 11 years and was responsible for the finances of the tech giant’s global advertising business before joining Eksperten.
“I grew up with this business and helped build it while holding multiple positions,” he says. “When I left, I had 150 people under me, on all continents. It has been an incredibly rewarding trip, but I think I’m at my best when I’m building something.
Datta’s experience at Google dovetails with her current role. “It seemed obvious to me that there was an application of the kinds of technologies we worked with at Google, which were basic predictive technologies, and applying them to better ways of running financial service models,” he said.