When Kalpesh Kapadia came to the United States 24 years ago to study, he couldn't get credit.
"I came from a privileged background in India, both financially and academically," he said. "I was a top student, and when I came here, I had to start all over again."
His classmates all had student loans. Kapadia, who is now the chief executive of Deserve, which he founded, had to go to his parents. And he couldn't get a credit card for two years.
Rohit Mittal, the CEO of Stilt, and his co-founder Priyank Singh both had trouble getting credit as international students in the U.S.
"I was not able to rent an apartment or get a loan to complete my education," Mittal said. "My co-founder, who was highly paid at Microsoft, wasn't able to buy a car in Seattle. He had to save money for months to make a 50% down payment. Most of our non-U.S.-citizen friends were in the same boat. We lived with these challenges for years and many continue to do so every day."
Jason Gross, founder and CEO of Petal, was affected by the experience of one of his co-founders, Berk Ustin, who came from Turkey to the U.S. in 2005 for undergraduate studies at the University of California, Berkeley. He was turned down when he applied for a credit card, even though he had worked at Amazon and a hedge fund.
Kapadia, Mittal and Gross have founded fintechs that use machine learning to help provide credit to immigrants who have no credit history in this country and often don't even have a Social Security number. Instead of relying on a credit report and a FICO score, they analyze other data elements such as bank account cash flow and bill payment.
Longtime industry observer John Ulzheimer applauds these efforts. "In my mind, as long as the data is predictive and cannot be gamed, then more power to the companies that design and develop the tools and the companies that use them," he said. "I don't believe we'll be replacing a traditional credit report anytime soon as that data is still, by far, the most commonly used data for risk assessment."
The problem
According to Kapadia, there are 1.2 million foreign students in America. That's one out of six.
"The population has grown over the last 24 years, but the services they are offered have declined," he said.
After the 9/11 terrorist attacks, the U.S. government stopped issuing Social Security numbers to international students unless they had a job. "Without those Social Security numbers, you do not exist in the financial system," Kapadia said. "You're invisible."
In 2009, the Credit Card Accountability Responsibility and Disclosure Act was passed, limiting banks' ability to market credit products to students. As a result, half of students in the 18-to-24 age bracket have no credit card or credit history. "It's a Catch-22 situation," Kapadia said. "People who don't have credit, can't get credit. And if they can't get credit, they can't build their credit score."
The offerings
Deserve provides credit cards to students and young professionals 18 to 29 years old with thin or no credit files. The interest rate is around 20%. Cardholders can qualify for escalating travel and entertainment rewards as their FICO score goes up.
"We want to be providing you with the best credit card for the stage of life you're in; we want to retain you as a customer, not lose you to another company," Kapadia said.
Deserve also offers its card as a white-label service for others. Sallie Mae is a client.
To comply with the CARD Act, Deserve does not market on campus.
"We do not give you a T-shirt or a USB drive to apply," Kapadia said. "We don't solicit applications. And when we talk to campuses, our focus is on financial literacy and credit education."
Student cardholders are also encouraged to make full, on-time payments.
Raj Date, former deputy director of the Consumer Financial Protection Bureau and current managing partner of the venture capital firm Fenway Summer, was an early adviser to and investor in the company.
Petal offers a similar credit card for which interest rates range from 14.5% to 25.5% and there are no fees. It also offers a line of credit of up to $10,000.
Stilt offers loans to immigrants, international students and visa holders. It doesn't require a Social Security number. Its rates range from 7.99% to 35.99% depending on state usury laws.
Where machine learning comes in
Deserve uses machine learning for identity verification, fraud detection and fair lending compliance.
The company has developed an analytics-based credit-decisioning framework. To verify identity, Deserve requires a government-issued ID and checks the DHS database to verify the applicant's immigration status in this country. It validates that the address is of an actual residence (P.O. boxes and commercial addresses are not accepted). The phone number and email address are also investigated.
Once the identity has been verified, the software looks at ability to pay, taking into consideration information in the credit application. Deserve does a credit report pull to make sure the applicant hasn't abused credit in the past - a person won't be approved if they have a blemish on their credit report. If they have no credit report, it's no problem.
Then the software does a cash-flow analysis on the applicant's bank account with the help of the data aggregator Plaid. It looks for telling behavior like late fees or overdrafts, and regular direct deposits, subscriptions and bills paid on time.
Deserve comes up with a composite score that takes into account identity and ability to pay. This score tends to correlate strongly to the person's eventual FICO score, Kapadia said.
Petal also has the ability to do cash-flow underwriting.
"We're looking at the whole picture. We're looking at cash-flow data and traditional bureau data to arrive at what we call the full digital financial record of the applicant," Gross said.
"Cash flow allows us to approve people even if they have no credit information at all."
Machine learning lets the company look at thousands of data points that have complicated relationships to one another, Gross said.
For instance, like most creditors, Petal considers income and monthly expenses, along with trends and volatility in that information.
"We like to sum it up as measuring the amount that somebody makes, saves and spends on a monthly basis," Gross said. "We like to apply something that I call the kitchen-table test - which is, if a family sitting around their kitchen table is trying to decide whether or not they can afford a big purchase, maybe it's a new car or a vacation, what are the things that that family is talking about? They're not talking about their history of using credit or their credit utilization. They're talking about: How much money do we have in savings? What's my next paycheck? What bills do we have coming up? Because we know that, at the end of the day, that's what really describes whether you can afford the new thing."
Some traditional banks shy away from using machine learning in their underwriting. They cite concern about what regulators will think and about the risks of replacing underwriting systems they've used for decades.
Gross and Kapadia both say that rather than having a black box for decision-making - which is regulators' top concern about the use of artificial intelligence in underwriting - their systems are transparent and auditable.
And in the case of Petal, its bank partner, WebBank, is regularly examined by regulators who are fully aware of the underwriting model.
Stilt uses machine learning techniques to make sense of financial and nonfinancial data. "We have categorized and ranked 100,000 universities across the world along with millions of employers, jobs and positions," Mittal said. "We also use transaction-level data from their bank accounts to understand daily, weekly, and monthly spending patterns. We can combine deeper characteristics of applicants that are highly correlated with on-time repayment behavior. This also helps reduce fraud."
Outcomes
Deserve has been at this for three years and has 50,000 cardholders from 164 countries. They're mostly students, spread across 2,400 colleges in all 50 states.
The delinquency rate on the card is between 2% and 3%, with delinquency defined as 90 days past due. The annualized charge-off rate is between 5% and 7%.
Gross, who launched Petal with WebBank a year ago, wouldn't specify the number of cardholders, but said he hopes to have 100,000 by the end of the year. The company now has 70 employees and it received $30 million in Series B funding in January led by Peter Thiel's Valar Ventures.
Users of Petal tend to be people who, because of the CARD Act and the financial crisis, are getting their first credit card post-college. "They are starting their credit journeys a little bit later in life," Gross said. "That's one of the reasons why a product like the Petal credit card is so important today. You have people applying to their first credit card when they're in their 20s and they have a real financial need and the only products available to them are the cards that used to go to students. There's a gap in terms of being able to meet the needs of today's new-to-credit consumer."
Petal is also attracting a lot of international students and professionals, he said.
He declined to discuss delinquencies and charge-offs, saying it is too early. "We are optimistic based off the repayment behavior and the utilization that we're seeing in the portfolio today," he said. "We've seen good performance, even from the credit-invisible consumer in the portfolio. We will obviously be learning more as time goes on."
Stilt did not disclose how many borrowers it has either. It's lending in 15 states. More than 90% of its borrowers are between 24 to 35 years of age and their average income is $80,000. They hail from more than 100 countries, primarily India, Mexico, China, Nigeria, Canada and the U.K.
Lessons learned
Stilt set out with a goal of helping international students get better access to credit. One thing the company has learned in the past three years, according to Mittal, is that the problem is much bigger and acute than he initially imagined.
"The U.S. credit system is fundamentally stacked against new immigrants who move here to build a new life," Mittal said. "It takes them years to build a financial standing. They also turn out to be one of the better asset classes in personal loans. They are motivated, high-quality individuals who understand credit and do whatever it takes to pay their dues."
Jason Gross at Petal said he has found that this type of product is harder than it looks.
"We've learned how operationally complex and difficult it is to change the way things are done in this industry," he said. "We've recruited a world-class team to rebuild the credit card from the ground up, and changing decades of industry practice requires significant investments of time and resources. It's hard work, but necessary if you want to create a best-in-class product."
