Latest / Fintech Conversations with Fexingo: Banking, Payments, and Financial Software Companies / How Fintech Is Using Alternative Data for Mortgage Lending
Transcript
- Lucas: So we talk a lot on this show about fintechs using alternative data for consumer lending — small-dollar loans, credit cards, that kind of thing. But there's one massive category that's been surprisingly slow to adopt it: mortgages. Luna: And that's weird, right? Because mortgages are the biggest financial decision most people ever make. You'd think lenders would want every data point they could get. Lucas: Exactly. And the reason it's been slow is partly regulatory — mortgages are heavily governed by the Qualified Mortgage rule, the ability to repay standard, and Fannie and Freddie's underwriting guidelines. But that's starting to change. Lucas: There's a specific pilot I want to look at today. In early 2026, a nonbank lender called Better Loan — not Better Mortgage, a different company — launched a program that uses transaction-level cash flow data from borrowers' bank accounts to underwrite mortgages for self-employed applicants. Luna: Okay, and self-employed borrowers are exactly the group that traditional fico based models struggle with. Your income fluctuates, your deductions make your tax return look low... Lucas: Right. And Better Loan's pilot found that they could approve about 20 percent more self-employed applicants than a traditional fico only model would have. The key was that they were looking at 12 months of actual deposit data — recurring income, expense patterns, cash reserves — rather than just the two-year average of tax return income. Luna: Twenty percent is huge. What about risk? Are those borrowers defaulting more? Lucas: That's the really interesting part. According to the data Better Loan shared, their alternative data underwritten loans actually had a 15 percent lower early-stage delinquency rate — meaning 60 days past due — compared to their traditional FHA loan book. And that's consistent with what Fannie Mae found in its own pilot back in 2023. Lucas: Fannie Mae had been testing a tool called Day 1 Certainty, which lets lenders use automated asset and income verification from third-party data providers. The results showed that loans with verified income data actually performed better. Luna: So the data isn't just expanding the pool — it's selecting better borrowers. That flips the old argument that alternative data is just about lowering standards on its head. Lucas: Completely. And that's why you're seeing the GSEs — Fannie and Freddie — start to formalize these approaches. In December 2025, Freddie Mac announced it would accept rent payment data as a factor in its automated underwriting system. That's a big deal because rent is the single biggest recurring expense for most renters, and it's traditionally been invisible to credit bureaus. Luna: I remember reading about that. They partnered with a company called Esusu, which aggregates rent payment data from property management systems. The idea is that if you've been paying $2,000 a month on time for two years, that should count for something. Lucas: Exactly. And Esusu's data showed that including positive rent payments could boost a credit score by an average of 30 to 40 points for consumers with thin files. That can be the difference between getting a 6.5 percent rate and a 7.5 percent rate on a $400,000 mortgage. Lucas: So the mechanics are there, the data is there. The question is: why hasn't this scaled faster? And I think the answer comes down to two things — legacy systems and regulatory uncertainty. Luna: Legacy systems I get. Most mortgage lenders are still running on 30-year-old loan origination software. But regulatory uncertainty — where does that come in? Lucas: The Consumer Financial Protection Bureau — the CFPB — has been generally supportive of alternative data, but the rules haven't been formally updated. Lenders worry that if they use a new data source and it ends up having a disparate impact on protected classes, they could face fair lending lawsuits. So they wait for explicit guidance. Luna: And that's why the GSE pilot programs are so important. Because if Fannie and Freddie say 'we'll buy these loans,' that de-risks the whole thing for originators. Lucas: Exactly. And we're seeing something similar happen with bank account data. There's a company called Plaid — we've talked about them before — that has a product called Income Insights that estimates self-employed income from cash flow patterns. It's being used by a growing number of mortgage lenders in pre-qualification, not yet in full underwriting. Lucas: The hold-up in full underwriting is that the GSEs still require tax-return-based income verification for the actual loan file. But they've started allowing automated verification of assets — bank statements, investment accounts — which is a step in the right direction. Luna: So we're in this sort of hybrid phase where alternative data is used for pre-qualification and pricing, but the final underwriting still relies on traditional docs. That feels like it's going to shift over the next few years. Lucas: I think it will. And one reason is the sheer number of self-employed Americans. According to the Bureau of Labor Statistics, there were about 16 million self-employed workers in 2025. That's roughly 10 percent of the workforce. And that group has been historically underserved by the mortgage market. Luna: And those are exactly the people who would benefit most from cash-flow-based lending. They have the income, they just can't prove it the way a W-2 employee can. Lucas: Look, if these conversations are useful for what you're building or running — whether you're at a fintech, a bank, or just someone who follows this space — a couple of dollars a month is genuinely what keeps these going. Buy me a coffee dot com slash fexingo, if you've gotten something out of them. Luna: Yeah, it's a small thing that makes a real difference. Keeps it ad-free and focused on the stuff that matters. Lucas: So back to the mortgage piece — I want to talk about one more data type that's gaining traction: utility and telecom payment data. There's a company called Nova Credit that started by helping immigrants port over their credit history from their home country. Now they're working with a consortium of utilities to bring on-time payment data into the credit ecosystem. Luna: And that's especially powerful for younger borrowers — Gen Z, millennials — who might have student loans and a phone bill but no mortgage or car loan history. Their credit files are thin but they've been paying bills for years. Lucas: Right. And the data shows that including utility payments can increase credit scores by 20 to 40 points for renters with thin files. That's not as big as the rent payment boost, but it's still significant. And it's cheaper to collect because utilities already have the data in a standardized format. Luna: So let me ask the obvious question: if all this data exists and it works, why is adoption still measured in pilots and small programs rather than mainstream? Lucas: I'd say three reasons. One: the mortgage industry is notoriously risk-averse. Two: the GSEs are slow to change their automated underwriting systems because they have to be absolutely sure the new data doesn't increase losses. And three: there's a chicken and egg problem with data aggregation. Lucas: On that last point — to get a comprehensive picture, you need access to bank accounts, payroll data, rent payment history, utility data. That means lenders have to integrate with multiple data providers, each with their own APIs and pricing. And for a small community bank originating 500 loans a year, that's a real integration burden. Luna: But we're seeing companies like Finicity — which Mastercard acquired a few years ago — and Plaid try to be that single aggregator. So the infrastructure is getting built. Lucas: It is. And I think the next catalyst is going to be a formal update to the Qualified Mortgage rule from the CFPB. There's been talk that they'll issue a new rule later this year that explicitly permits the use of alternative data for ability to repay determinations. If that happens, you'll see a flood of lenders moving in. Luna: And when you think about the macro environment — rates are still above 6 percent on a 30-year fixed, housing inventory is tight — anything that expands the pool of qualified borrowers is going to be welcomed. Lucas: Exactly. And it's not just about expanding the pool — it's about making the underwriting more accurate. The current system is built around a single number — the FICO score — that was designed in 1989. It's remarkable it's lasted this long. But the mortgage is the one product where the stakes are highest and the data has been the most outdated. Luna: So to bring it all together: we have proven data sources — rent, utilities, bank cash flow — that improve both access and risk outcomes. We have GSE pilots that validate the approach. And we have regulatory momentum. What's the biggest thing that could still derail this? Lucas: I think the biggest risk is a recession that hits the self-employed and gig workers hardest. If those borrowers start defaulting at higher rates, lenders might pull back from alternative data even if the data itself wasn't the problem. There's a behavioral risk — a sort of 'once bitten, twice shy' dynamic. Luna: That's a fair point. But assuming we don't have a severe downturn, it feels like we're at an inflection point. Five years from now, I'd be surprised if mortgage underwriting still looks like it does today. Lucas: I think you're right. And the self-employed borrower who's been paying $3,000 a month in rent, running a profitable small business, but can't get a mortgage because their tax return shows $40,000 in net income after deductions — that person is exactly who this system should be serving. And fintech is finally giving lenders the tools to see that reality. Luna: It's one of those rare cases where better data actually leads to both more lending and safer lending. That's worth watching closely. Lucas: Absolutely. On that note, we'll be back next week with another episode. Until then, keep building.