Twitter veteran launches credit score that digs deep into cash flow
The New York fintech startup Harvest began offering a new credit score Thursday that uses artificial intelligence to take into account cash flow, spending habits and financial discipline alongside traditional scoring data to help lenders determine if a potential borrower is creditworthy.
It's also designed to help consumers understand their own financial health and creditworthiness. Several New York credit unions and a large U.S. bank have signed up to use the new score, the company says.
The score, which is called Pro Index, is part of a growing trend toward the use of alternative data in credit decisions. Fintech lenders like Upstart, Petal and Kabbage already use alternative data in their AI-based underwriting models to analyze cash flow, spending patterns and even information about potential borrowers' education and profession to figure out whether they have the means and willingness to repay a loan or pay off a credit card.
Harvest’s Pro Index is one mechanism banks could use to compete with such lenders and to reach people who don’t have a credit score. Other vendors, including Upstart, Enova, Zest and Kabbage offer credit-modeling technology that banks can use to bring alternative data into loan decisions.
Creating a new score
Nami Baral, founder and CEO of Harvest, started her career as an investment banker, then was head of revenue partnerships at Twitter for four years. She left Twitter to start Harvest in 2017 with the initial idea of helping improve the net worth of the average American.
“I was thinking about it from the asset perspective: Can we help them improve their net worth by maybe getting them more savings tools or getting them better investment tools,” Baral said.
But when she started building the minimum viable product and canvassing Americans about their needs, she saw bigger problems. So many people live paycheck to paycheck, lacking the $400 they would need to deal with an emergency; and the average consumer has $30,000 of debt, excluding mortgages.
“I saw a big mismatch there, and I decided to start Harvest to build automated tools to help consumers reduce their liabilities and thereby improve net worth by attacking the liability side of the equation,” she said.
Harvest got into the Rise accelerator run by Barclays and Techstars and finished its first product there, an automated bank-fee-negotiation tool.
“The average American has all sorts of debt — student loans, mortgages, auto loans, personal loans — but also pays a lot of money in bank fees every year,” Baral said. “I saw an opportunity to get money back for them." For instance, the tool negotiates overdraft fees and annual percentage rates.
More than 100,000 customers now use this tool and have gotten $2 million in fee refunds, she said. The software also watches these consumers’ debt load and provides suggestions on how to better manage it.
The economic troubles brought on by the pandemic got Baral thinking it might be time for her startup to pivot again.
“We realized that the traditional credit score, if it is taken in isolation, tends to not really do a lot of justice to customers in times of uncertainty like this,” Baral said. “And we realized that if you can combine the signaling provided by your credit score, which is a lagging indicator, with something that is a leading indicator of somebody's financial wellness and discipline, then we can bridge that gap in terms of the customer getting a score that they can actively try to improve, while giving lenders insights and information that the traditional credit score is not able to provide.”
Baral acknowledged that the FICO score is ingrained in most banks’ underwriting models.
“But every lender you talk to will tell you there are definitely drawbacks or gaps that the traditional credit score is not able to fill,” Baral said. “For example, it doesn't work for unbanked or underbanked customers. It does not work for thin-file or recovering-file customers. It does not really show how a customer is trending.”
For instance, the creditworthiness of a consumer who has a 720 FICO score but recently dropped from 780 is different from a person whose score has risen from 680 to 720, she said.
“We realized that there needs to be a better financial scoring model that does incorporate what the credit score is able to provide, but then also augments it with other kinds of information so that for lenders, it's easier to make their credit decisions. And for customers, the credit constraints that they may be experiencing on the traditional credit scores are somewhat mitigated,” Baral said.
So her team developed the Pro Index, which is derived from a combination of a traditional credit score and a cash flow data analysis. Harvest works with multiple data aggregators to obtain the cash flow data, with users’ permission.
The Pro Index analyzes cash flow data for signs about how consumers spend their income: Are they disciplined? How much goes to essentials and nonessentials? It also evaluates how consumers use debt.
Consumers who spend 100% of their income on essentials, for instance, may not have additional funds to pay off debt or start saving and won’t be penalized for that. But consumers who spend on unessentials and take on more debt might see their Pro Index drop.
“Your Pro Index improves if you are improving your credit score, improving your spending habits, or reducing your debt within the context of your income,” Baral said.
The Pro Index considers debt-to-income ratio, but it doesn’t penalize people who don’t want to take on much debt.
“If you are more financially healthy as a customer, then you are more creditworthy as a customer, too, because you'll be able to service your debt regardless of where your credit score might sit,” Baral said.
The Pro Index also incorporates real-time income verification, real-time employment verification, and real-time deposit and spending data. So along with using the score in their credit underwriting, banks could use Pro Index as a monitoring tool to reduce portfolio risk, and perhaps provide early help to customers whose financial wellness is declining. Harvest calls its product for banks “Ability-to-Pay as a Service.”
Going beyond FICO
Again, Harvest is not the first company to see a need for a broader look at consumer creditworthiness beyond the traditional credit score.
Most banks use FICO and VantageScore (the three credit bureaus’ joint alternative to FICO) scores in their loan and credit card decisions. Both sets of credit scores analyze consumers’ credit reports and take into account factors like the person’s payment history, length of credit history, amount owed and available credit.
People who don’t have credit for one reason or another — perhaps because they’re young and just starting out, they’re older and have paid off all their debt, they don’t like having debt. or they’ve had a temporary financial setback — do not generate the data needed to calculate a credit score.
There have been attempts over the past few years to help such “credit invisible” or thin-file consumers qualify for credit.
FICO, Experian and Finicity rolled out UltraFICO in 2019. This score uses Finicity’s account-data aggregation technology to tap into checking, savings and money market account information.
Also last year, Experian introduced Experian Boost, which adds recurring, but not traditionally reported, payments to a credit report, including utilities such as electricity, water, cellphone, internet and natural gas.
And TransUnion rolled out eCredable Lift, which similarly lets consumers report utility and phone payments into their credit reports.
FICO also recently introduced the UltraFICO, a new credit scoring model that incorporates bank account data such as savings balances, length of account history and frequency of transactions in coming up with a credit score.
But most banks still use the older FICO and VantageScore scores embedded in their entire underwriting systems .
The biggest challenge to any newcomer to this market is creating a reliable and predictable model that regulators can follow, said Steve Ely, CEO of eCredable, which furnishes phone and utility accounts directly to credit bureaus such as TransUnion. That way, the data is part of the traditional credit report and therefore can be incorporated in the traditional credit score.
“Disparate impact is a major red flag when it comes to scoring models,” Ely said. “Scoring models like FICO and VantageScore have plenty of evidence that they can stand up to regulatory scrutiny. If the lender is using something the regulator is not familiar with, they have to spend extra time explaining their underwriting approach. The time is an issue, but so is the risk of being wrong. It’s much easier to just let the regulator check the box and move on.”
However, regulators have signaled an increasing openness to the use of alternative data in credit decisions.
In December, the Federal Reserve Board, Federal Deposit Insurance Corp., Office of the Comptroller of the Currency, Consumer Financial Protection Bureau and the National Credit Union Administration issued a joint statement saying they recognized that the “use of alternative data may improve the speed and accuracy of credit decisions and may help firms evaluate the creditworthiness of consumers who currently may not obtain credit in the mainstream credit system.”
Banks are starting to open up to the idea, too.
“We see a range of activities in the market and among policymakers that point in the direction of opening up credit risk assessment to the use of broader data,” said Melissa Koide, CEO of FinRegLab. “Among the most promising of those initiatives are efforts to incorporate alternative financial data — in particular, transaction-account information that gives lenders a detailed and timely view of an applicant’s income and expenditures.”
FinRegLab’s prior research on the use of such cash flow data suggests it improves lenders’ ability to assess the creditworthiness of applicants who are on the margins or outside the traditional credit scoring system, she said.
“We are also seeing signs that the use of this data has helped lenders respond to economic uncertainty during the pandemic,” Koide said. “Given that, we expect continued efforts to use this data and to resolve policy questions about data portability and relevant consumer protections to accelerate.”