Online lending platform provider
The San Mateo, California, company builds
The company's leaders recently predicted they will report positive earnings before interest, taxes, depreciation and amortization in the fourth quarter. In the days following that September announcement, Upstart's stock price nearly doubled.
"Great quarter, great results. Very happy to see that," analyst Dan Dolev at Mizuho said during the company's second-quarter earnings call.
Not all analysts are bullish on the company, however.
"I remain worried about Upstart's credit underwriting, both from a macro consumer perspective and from what Upstart is doing to achieve the stabilization we've seen in recent transaction volume," said Vincent Caintic, managing director at BTIG, in an email interview.
The two-year winter
Upstart's Chief Financial Officer Sanjay Datta said last month that the company has had a "two-year nuclear winter."
This is due to a confluence of factors, including bank liquidity challenges, Girouard said in an American Banker podcast that will be published Sept. 24.
"In the first half of 2023 in particular, banks were just seeing loss of deposits," he said. "There was a lot of fear and doubt about lending. So a lot of banks pulled back, a lot of credit unions as well."
Another factor was consumer overspending, Girouard said.
"The American consumer became overextended as the stimulus that was provided en masse during COVID was withdrawn," he said. "In effect, there was this large massive layoff that happened across the landscape because the government suddenly stopped sending out checks to people. In the post COVID era, consumers have been over their skis on spending, and that's made it a riskier environment."
Higher risks led to higher interest rates, which led to lower loan volume on Upstart's platform, he said.
To get through this difficult time, Upstart reduced spending and
If efforts to make AI-based loan decision models less discriminatory degrade accuracy, that is problematic for the entire industry.
"We have fewer staff," Girouard said. The company currently has 1,200 employees. "We've had a few projects that we've decided to postpone or cancel. Those aren't uncommon things that any responsible company would do in these kinds of times."
Model improvements
The company also invested in improvements to its loan decision platform, Girouard said. Engineers have improved the ability to calibrate the models to changes in the macro economy, he said. They have also continued to work on "separation," or understanding the relative risks of different applicants.
"From a technical perspective, we can only wish a couple years ago that we had the tools that we have today, and we feel much better prepared," Girouard said. "You can't predict the macro, but you can certainly handle it with finer precision. That's what we hope to do with our lending partners, help them manage their lending responsibly through any kind of macro climate."
In one example of a model improvement, Upstart recently launched a new algorithm, Model 18, that incorporates the annual percentage rates of loans.
"APR is normally part of the output of a model – what APR should I charge this customer based on all these factors we know?" Girouard said. "But the interesting thing is that the APR itself affects the performance of the loan."
If customers are charged higher APRs, their monthly payments are going to go up, therefore their likelihood of default also rises.
But there's also adverse selection: A person likely to accept a 15% loan is typically less creditworthy than someone who's willing to accept an 8% loan.
"The best borrowers have lots of choices, and if you have a much higher APR for any particular borrower, you're likely to be adversely selected, and that results in underperformance," Girouard said. By taking APR into account in the credit decision, Model 18 has been boosting credit performance, he said.
"We're getting to a place where we feel like, if you're not using AI to do lending, you're going to be left behind at some point because it's just invariably more efficient, more accurate, better for all," Girouard said.
Caintic noted that Upstart is moving into smaller dollar loans, with the average loan size now $7,700 versus more than $11,000 two years ago. "Generally these smaller loans are to lower credit quality customers and have higher frequency of loss," he said.
AI models have other advantages. It's easier to recalibrate an AI-based model than a more traditional credit risk model based on linear regression, Girouard said.
"If you think of artificial intelligence as just a lot more data and more sophisticated math to understand and interpret that data, then invariably it's going to be an enormous advantage, particularly in volatile times," Girouard said.
Proving the efficacy of an AI-based model has to be done over a long period of time, he acknowledged. But "the sort of accuracy and precision with which you can make decisions based on what's going on day by day, loan by loan, is really brought to the front by AI," Girouard said.
The company has gotten some of the investors in its loans, including banks and credit unions, to commit to using the platform long-term, Girouard said.
"Over half of our capital in our platform today is in the form of longer-term commitments," Girouard said. "And that's a big leap forward for us."
In the bigger consumer credit picture, U.S. household debt is high; it rose to $17.80 trillion in the second quarter, according to the
Caintic, the BTIG analyst, said overall consumer credit quality is deteriorating.
"So the macro is going to get worse for Upstart, which will make underwriting an even further challenge."
Girouard admitted he worries about this. "We're very cautious about the state of the consumer, the stability of their income, their expenses, including, of course, debt."
But Upstart's loan decision models take consumers' debt into account before approving loans, he said.
"Any lender is going to look at their book and say, wow, looks like we're underperforming in this segment or that segment, we're going to make some tweaks and some adjustments," Girouard said. "This is just doing it much, much faster with much, much more precision. And ultimately we believe that will lead long-term to much better performance."
