Writer: Eric Bank
Editor: Lillian Guevara-Castro
Reviewer: Adam West
Key Takeaways
- The traditional credit score doesn't account for the financial habits of 90% of Americans with lower credit ratings, said Rob Barnhart, President of Snap Finance.
- Alternative data and responsible AI can fill the gap.
- The pace at which AI-enabled underwriting will reach mainstream depends largely on clarity around governance and privacy principles, said Barnhart.
When Snap Finance President Rob Barnhart entered the subprime lending market, he realized that something was lacking.
“We don’t use Experian, Equifax, or TransUnion as a part of our decision,” he said. “We don’t think they bring any value to the decisioning part of the process.” It is a radical approach, to say the least, and a far cry from how credit agencies are traditionally treated by lending institutions.
Recent federal data show that more than one in four U.S. adults have subprime and/or limited credit histories. Traditional scores can penalize people who pay rent and utilities on time — they lack credit cards and long-term loans.
Barnhart sees those gaps as a signal that the system needs to evolve. “About 35% of consumers have credit scores below 670, and they rely on financing,” he said. “That’s a real need that won’t be met through traditional bureau information.”
The impetus for change lines up with a growing financial movement. It factors in cash flow data and real-world payment behavior. FICO and Experian have both added buy now, pay later and bank-linked information to new scoring models. This is a trend that says Barnhart’s approach is gaining ground.
Alternative Data and Behavioral Underwriting
It’s a hybrid approach between connectivity and cash flow analysis. Says Barnhart: “We are looking at alternative credit bureau data, validated phone numbers, bank-related scores, and cash flow data. It offers a better perspective on people who are not connected to the traditional credit market.”
A consistent direct deposit may convey reliability — more than just a credit score will. That thinking mirrors the work of other non-prime innovators.
Petal Credit and Prism Data are examples. They use open-banking data to help assess changes in income as well as spending patterns. The Urban Institute says that adding utility and rent payments can lift scores without an increase in default risk. This is especially true for thin-file borrowers.
For Barnhart, the goal is to broaden the definition of creditworthiness. “You have to look at the story behind the score,” he said in unpublished remarks. “When lenders see the full picture, they can make better decisions for more people.”
Responsible AI and Transparency
The AI-driven Snap uses thousands of variables. Before they can go live, these variables must run the gamut of rigorous reviews to make sure they are legal and fair.
“We make sure those variables would withstand heavy scrutiny from external regulators,” Barnhart said. Bias-related data, such as demographics, are excluded entirely.
Snap relies on SHAP values to help borrowers understand their results. This is an explainable AI method that shows which factors most influenced a decision. That approach lines up with the Consumer Financial Protection Bureau’s recent guidance encouraging “adverse action clarity” in automated lending.
Transparency, Barnhart argues, is not optional. “You should be able to explain to people why they got the score they did and how to improve over time.”
Snap is compliant with PCI, SOC 2, and HIPAA standards. This protects sensitive data. Barnhart says privacy and security are “non-negotiable.” In addition, he says that ethical AI only works when consumers trust how their information is used.
Testing and Trust in a Changing Market
Barnhart’s analytics career has taught him to trust data over intuition. He says testing a 5% cash back offer boosted sales enough to outweigh lower margins. That’s a reminder that customer behavior can flout expectations.
He also said a Snap Finance experiment challenged another industry assumption. The company tested 12-month versus 18-month lease terms. The longer option was “much better for consumers.”
Higher approval amounts turned out to not hurt performance. “The unwritten rule was that 12 months was too long,” Barnhart said. “But the data showed us otherwise. The only way to find out who’s creditworthy is to test broadly. Models show you what actually works.”
That philosophy now guides Snap Finance’s approach to credit inclusion. The company seeks to reach borrowers that traditional systems overlook. AI is becoming standard in lending. Barnhart believes success will depend on balance. That means testing new signals without sacrificing fairness or privacy.
“The future of credit is not the score. It’s the story behind it,” Barnhart said.
