Alternative data can help lenders say yes to borrowers with little credit history, but only if it sits beside the basics. Three respondents explain which signals they would trust and how to keep the risk in check.

Alternative data is moving from experiment to mainstream lending. Rent, utility bills and bank account cash flow now feed some credit scores and many small business models. Still, the question for lenders is how much weight these signals deserve. Many borrowers remain hard to score. In CFPB data from 2010, about 11% of US adults had no record at the three largest credit bureaus. Another 8.3% had files too thin to score, the Congressional Research Service notes. For how lenders reach bank data with consent, see our explainer on open banking.

We asked how lenders and risk teams use alternative data to judge borrowers with little or no credit history, and how they keep the risk in check. Three people answered in this round, and we used all three. None works in credit risk at a lender, so we’ve set their advice against the research and the rules.

Where new credit signals now sit

Regulators have opened the door wider. In July 2025, the Federal Housing Finance Agency let mortgage lenders use VantageScore 4.0 instead of Classic FICO. That score draws on rent, utility and telecom payments, plus 24 months of trended data, according to the same CRS report. In addition, the agency adopted FICO 10T in April 2026. As a result, alternative data now reaches US mortgage decisions, not just fintech apps.

The UK is fixing gaps in existing files first. In February 2026, the FCA proposed that lenders sharing credit data with one designated credit reference agency share it with all of them, according to its press release. That would fill some gaps before alternative data even comes into play. Meanwhile, the CFPB is rewriting the US open banking rule, which governs how lenders pull bank data with consent. It sent a revised proposal for review in August 2026, Consumer Finance Monitor reported.

Use alternative data to cut uncertainty, not replace history

Chirag Kulkarni is founder and CEO of Taco, one of several internet businesses he has co-founded.

“Alternative data shouldn’t replace credit history. Its value is in reducing uncertainty when an applicant has little recorded history but shows steady business activity. Even so, those signals shouldn’t distract from the basic question of whether they can repay.

Risk teams should separate early signs of growth from evidence that supports a lending decision. Customer interest or digital engagement may suggest a business is growing, but neither proves it can repay. Test each input against the industry, the business’s age, seasonality and wider economic conditions, and review the downside cases before limits go up. Keep decisions measured, and adjust them as real payment behaviour gives you stronger evidence.”

Chirag Kulkarni, Founder and CEO, Taco

Research on cash flow lending backs his caution. A National Bureau of Economic Research working paper found that adding bank-statement cash flow to a FICO-based model improved default prediction, especially for owners under 35. Lenders that leant heavily on cash flow approved 2.4 percentage points more applicants under 40. Those young owners’ businesses also survived no worse. So alternative data adds most where a credit file says least, which fits his point. In practice, that means treating softer alternative data, such as web traffic or online engagement, as context, while cash flow and repayment carry the decision.

However, the paper is not peer-reviewed, and it draws on three fintech lenders. We could not confirm what Taco does today. He also gives no lending example of his own, so his answer is general guidance rather than a case study.

Build limits from cash flow and repayment

Ahmed [SURNAME TO CONFIRM] is a financial author at CoinTime.

“I’d use transaction behaviour and cash-flow data to build a fuller picture of borrowers who don’t have traditional credit records. For individuals, that could mean recurring income, account activity, regular bill payments and past repayment behaviour. For SMEs, I’d look at how often they make sales, how steady their revenue is, how they pay suppliers, and their operating cash flow.

The risk side comes from combining those signals with identity checks, affordability checks, fraud detection and cautious starting limits. Borrowers can then earn higher limits as they build a reliable repayment record. I’d also track default rates by customer segment and review the model regularly to catch inaccurate or discriminatory decisions. The aim is to lend based on how people really handle money, rather than shutting them out because they don’t have a credit score.”

Ahmed [SURNAME TO CONFIRM], Financial Author, CoinTime

For SMEs, the alternative data he lists, such as supplier payments and sales frequency, mostly comes from bank feeds and accounting software. That makes consent-based data access central to his model. His step-up approach, starting small and raising limits as repayment builds, suits borrowers with no track record. It also limits losses while alternative data proves itself. His point on tracking defaults by segment matters too, because a model can look accurate overall and still fail one group. That is how unfair outcomes hide inside averages.

Still, he writes about finance rather than running a lending book. His answer describes what he would do, not a programme he runs, so we’ve edited his opening to reflect that.

Check the application against the documents

John Mac founded Exit Evidence, which advises Shopify founders preparing to sell, and he says up front that his firm does not lend.

“I’m John Mac, founder of Exit Evidence and OPENBATT. We’re not a lender, but we do assess founder-owned e-commerce businesses that aren’t yet ready for bank finance, using many of the same alternative data and risk checks.

The first thing I look for is the ‘lie’ rate, not sentiment. Anybody can lie on an application. We use natural language processing to pull data points from the founder’s own documents and compare them with what they said in the application. Say a founder claims a core strength is a diverse customer base with recurring revenue. A sample of their contracts shows most customers are on month-to-month plans with no long-term commitment, and one customer makes up 40% of revenue.

Or the founder says the business has a strong supply chain, and an analysis of customer reviews shows a recent spike in complaints about late deliveries and stock-outs. You can do the same with bank statements, card processing statements and sample contracts. If something doesn’t add up, it’s a red flag. I’d review those cases by hand rather than kick them out of the funnel automatically. NLP is a great flagging tool, not a credit decision tool.

Then there’s false positives. People say sentiment analysis is hard because customers only leave reviews when they’re angry. That’s true, but sentiment and topic are two different things. Reviews may be negative overall, but if the main topic is late deliveries, and those complaints have jumped sharply over the last quarter, that tells you something. So I look for a change in the sentiment or topic of a business’s communications. If messages from its suppliers show more complaints about late payments, for example, that’s a red flag for credit risk.

The same goes for contracts. NLP can make mistakes, but it’s very good at spotting changes in terms and conditions. I’d only treat a flag as significant if a second data point supports it, such as a change in supplier terms alongside a rise in chargebacks.”

John Mac, Founder, Exit Evidence

His method uses alternative data to test what a borrower claims, not just to score them. Bank statements and card processing records are alternative data in their own right, so the cross-check adds little cost. One customer providing 40% of revenue is a classic concentration risk, and contracts can reveal it when an application does not. Also, his insistence on human review fits UK data protection law. The Information Commissioner’s Office gives the automatic refusal of an online credit application as an example of a decision with significant effects, which brings extra safeguards.

However, Exit Evidence advises founders on selling their businesses and does not lend. His examples read as illustrations rather than named cases, and he gives no figures for how often flags prove right. Still, his two-signal rule is a sensible guard against false alarms from alternative data.

What works for lenders using alternative data

The three answers point to a clear order of work. First, decide what each new signal tells you, and keep the ability to repay at the centre, as Kulkarni argues. Next, test each input against industry, business age and seasonality before it shapes a limit. After that, cross-check what applicants claim against their contracts, bank statements and reviews, as Mac suggests, and review mismatches by hand.

Then start new borrowers on cautious limits, and raise them as repayment history builds. Finally, track default rates by segment and review the model often, because alternative data can drift as conditions change. Banks already using AI in credit work, such as TD’s credit memo tool, face the same need to show their reasoning.

What to watch next

Rules on data access will shape how far alternative data can go. Watch for the CFPB’s revised open banking proposal, since it could change how lenders pull bank data. Also watch whether the FCA confirms its credit data sharing rules after its consultation closed in May. In addition, mortgage lenders’ uptake of VantageScore 4.0 will show whether alternative data such as rent and utility payments changes real approvals.

Above all, alternative data works best as a second opinion, not a replacement. Lenders that pair it with affordability checks and cautious starting limits can approve more thin-file borrowers without taking blind risks.

FintechBits covers payments, banking and fintech developments for readers in the US and UK. This article is for information only and is not financial advice. Views expressed are those of the FintechBits editorial team.