Why economic volatility doesn't worry this fintech

Krubiner-Gal-Pagaya
If there's an economic downturn, Pagaya founder and CEO Gal Krubiner says there will be demand for deeper payment data and AI for credit decisions.
Pagaya

With many companies tightening their belts out of recession concerns, New York-based fintech Pagaya is instead looking at growth — a strategy that echoes its approach to digging through data to find the opportunity that others may overlook.

Pagaya went public on June 22 via a merger with EJF Acquisition Corp., a special purpose acquisition company, or SPAC. Pagaya hopes to expand its services, which include using artificial intelligence and payment data to help financial institutions manage credit. 

The company burst into the financial news on Monday after its stock, which had slid more than 60% since its listing, suddenly jumped more than 300% on Friday, raising speculation of a short squeeze, or a sudden exit of short sellers to cover their positions, similar to what happened with GameStop in 2021. 

Pagaya would not comment on its stock price, which on Monday had given back most of Friday's gains. Gal Krubiner, co-founder and CEO of Pagaya, said that as the economy struggles, banks will tighten credit, which Krubiner contends creates an opening for Pagaya's model that analyzes thousands of data points to generate tailored recommendations for its clients' credit decisions, ostensibly to identify borrowers who don't meet a FICO score cutoff but are still a good fit for a loan. 

"AI can observe more data and provide more relevance for each consumer," Krubiner said.  

Pagaya has built a data set of more than 50 million consumers that feeds expenditures into the company's AI engine, according to Krubiner.

Fears of a recession abound, but payment and other financial technology companies that can address pressing business challenges will draw funding, according to venture capital investors.

July 11
Mendoza-Adrian

Pagaya was founded in 2016 with three people and now employs more than 800. It had about $475 million in revenue in 2021, which was up about 380% from 2020.  

The company refers to itself as a B2B2C firm, using payment records and other data sources to inform lending terms. Its partners include Ally, SoFi and Visa. In the run-up to going public, Pagaya hired Ashok Vaswani, former CEO of Barclays U.K., to be its president. Vaswani is responsible for leading all commercial, risk, regulatory, compliance and legal efforts with bank partners. 

Pagaya is including buy now/pay later payments as data points for its AI engine, and is also adding more recurring payments beyond monthly utilities bills and rent. 

"There is more available than just the payment at the end of the month. That is one reason why we partnered with Visa," Krubiner said.

Fintech lenders typically make more use of AI and machine learning than banks do, though banks have closed the gap in recent years. Using AI based on a potential borrower's payment trends is not widely used for credit decisioning, though there is a lot of interest in the concept, according to Craig Le Clair, vice president and principal analyst at Forrester Research. 

"What you usually have now is a human being that is populating third-party data, or maybe they have an internal app that is doing scoring so there's a loose amalgamation," Le Clair said. 

Newer AI engines for consumer lending look at factors such as email or social media-related activity in addition to credit performance. If email addresses or social media accounts remain the same for a long period of time, that's a good sign, according to Le Clair, stressing this type of data is combined with other types of data, such as a pattern of payments of bills, or types of items purchased on an e-commerce site or a store. 

A sample pattern could be an increasing size of payments for luxury goods, or a decline of luxury goods payments toward more basic-needs purchases that suggest concerns of economic stress, according to Le Clair. 

This data is then run through an AI program with data from hundreds of other borrowers with similar profiles to produce a likely outcome of extending credit to that applicant, according to Le Clair. This can produce a risk profile that's different from what would be suggested by a credit score that comes from credit card or loan payments fed to a reporting bureau.  

"In retail banking, there's a lot of good credit opportunities that are being missed because the loan applicant does not have a record of being in debt and making payments on that debt," said Le Clair. 

As another example, a potential borrower's payments for an expensive customized car wash could indicate financial health for consumer credit, according to Stewart Watterson, a strategic advisor in Aite-Novarica's retail banking and payments practice. 

If a trail of payments for a single borrower suddenly stopped, there would be enough examples of other people who stopped these types of payments to make a predictive assessment, according to Watterson.  

"Traditional credit decisioning tells you that something bad has already happened," Watterson said. "By using payment data and AI you can get a sense of a propensity that's forward looking." 

Fighting bias

The benefit of AI — that it can understand trends humans can't see — has the downside of making it hard for humans to explain a particular decision, particularly to an applicant who has been rejected. 

"The more you use the complexity of machine learning, there are fewer people in an organization who understand how the decision is made who can explain it," Le Clair said.

There is also the issue of bias in machine learning, which can worsen over time as the AI system gathers more information that's funneled through the original program, Le Clair said. 

"Our history is biased," Le Clair said, adding about 90% of programmers were white men up until about two years ago when the diversity of university graduates in AI-adjacent disciplines began to improve. "That past of mostly white men creates a current challenge that's hard to suss out."

Rohit Chopra, director of the Consumer Financial Protection Bureau, has warned about bias in AI-informed lending. The algorithms can never be free of bias and could result in unfair credit determinations, according to Chopra. 

AI firms such as Zest, which has partnered with financial institutions, contend machine learning can be programmed to broaden data sourcing in a manner that reduces bias in lending decisions. 

Pagaya said it is currently using AI technology to underwrite loans that come through its partners, adding its technology is compliant with all fair-lending rules and regulations, and the company ensures that by doing independent validation with reputable third parties. 

For reprint and licensing requests for this article, click here.
Payments
MORE FROM AMERICAN BANKER