Continue Reading
The FICO score was engineered in 1989. The underlying methodology has been refined over the decades, but its core inputs — credit card utilization, length of credit history, account mix — were designed for a world where a person’s financial life fit neatly inside the banking system. For roughly 45 million Americans, it doesn’t.
These are the credit invisibles and the near-primes: people with thin files, recent immigrants, young adults, gig workers with irregular income, and anyone who runs their financial life largely in cash or through non-traditional accounts. A FICO score tells you nothing useful about them. That’s not a small problem. It’s a $150B+ credit gap that traditional lenders have historically just stepped around.
The FICO framework rewards longevity and product diversity inside the credit system. Hold a credit card for fifteen years without missing a payment and your score reflects it. But that same framework penalizes people who are simply outside the system — not because they’re bad risks, but because they haven’t participated in the specific financial products FICO was designed to track.
A 28-year-old nurse who has paid her rent on time for six years, never missed a utility bill, and carries a positive checking account balance every month is, by any reasonable measure, creditworthy. Her FICO score may tell a different story because those payments aren’t reported to the bureaus. The rent she pays reliably every month is invisible to the model making the decision about whether to extend her a car loan.
The result is a systematic mispricing of risk. Traditional lenders underestimate the creditworthiness of a large population segment and either decline them outright or charge them rates that reflect uncertainty rather than actual risk. The arbitrage opportunity for lenders willing to look at broader data sets is significant.
Several alternative data categories have emerged as meaningful predictors of repayment behavior:
Rent payment history. Monthly rent is typically the largest recurring payment obligation in a consumer’s life. On-time rent payment over extended periods correlates strongly with responsible financial behavior. Inclusion of rent data in credit models has been shown to improve thin-file consumers’ scores substantially in pilot programs. The barrier has been data collection — rent payments generally require the landlord to report them, and most don’t. Platforms that connect directly to property management software or bank accounts are solving this at scale.
Utility and telecom payment history. Similar to rent, electricity, gas, and mobile phone payments are recurring obligations that thin-file consumers handle every month. The infrastructure for collecting this data at the bureau level has existed in nascent form for years but uptake among lenders has been slow. That’s starting to change as newer underwriting platforms make the integration straightforward.
Cash flow analysis. Open banking APIs have made real-time transaction data accessible with consumer consent. A 90-day view of a consumer’s checking account tells you income stability, expense patterns, average balance, and how they handle overdraft situations. This data is arguably more predictive of repayment capacity than any bureau score, particularly for gig and contract workers whose income looks irregular by traditional measures but is actually quite consistent when viewed at the right granularity.
Subscription payment behavior. Streaming services, gym memberships, software subscriptions — these are small-dollar recurring obligations that surface behavioral patterns. Consistent payment across multiple subscriptions over time correlates with responsible financial management. Several underwriting models now incorporate this as a positive signal for thin-file applicants.
We look at this space through two lenses: data aggregators and underwriting model builders.
Data aggregators sit between the raw source — property management systems, utility providers, bank account transaction feeds — and the lenders who need the data in a usable format. The moat is coverage, speed, and data quality. Getting a landlord to integrate takes time. Building a database of reliably formatted rent payment records for millions of units takes longer. The companies that have done this work have meaningful lead positions.
Underwriting model builders are the companies using this data to make better credit decisions, either as lenders themselves or as model-as-a-service providers to banks and credit unions. The interesting dynamic here is the feedback loop: every loan they underwrite using alternative data generates repayment data that improves the next generation of the model. Over time, a lender using alternative data accumulates a training set that no traditional bureau can replicate.
The lender who first underwrite the credit invisible population at scale isn’t doing charitable work. They’re capturing a customer segment that proves to be less risky than the bureau scores suggested — and building loyalty that traditional lenders spent decades ignoring.
Alternative data in credit underwriting operates under FCRA and ECOA, which means lenders using non-traditional data sources need to be able to demonstrate that their models don’t produce disparate impact on protected classes. This is technically challenging because some proxies for creditworthiness can correlate with demographic characteristics in ways that create regulatory exposure.
The companies doing this correctly are investing in model fairness testing as a core compliance function, not as an afterthought. Bureau-reported data has been tested against disparate impact standards for decades. Alternative data models are newer and require rigorous fairness analysis to ensure that expanding credit access to underserved populations doesn’t inadvertently encode different forms of discrimination.
This regulatory complexity is itself a competitive advantage for companies that handle it well. It raises the barrier to entry for less sophisticated players and creates a meaningful moat for teams that combine underwriting expertise with compliance depth.
The credit bureau model has been disrupted at the edges but not at the center. The major bureaus have responded by launching their own alternative data products — Experian Boost, TransUnion’s cash flow inclusion — but these are opt-in consumer programs, not systematic inclusions. They improve the scores of consumers who take the time to enroll, but they don’t change the default underwriting approach.
The more significant shift is happening at the lender level. Fintechs that built alternative data underwriting from the ground up have accumulated several years of loss data that validates their models. They’re now in a position to prove to institutional capital markets that their loan books perform as expected. That proof point unlocks securitization, which unlocks scale.
The 45 million credit invisibles are not going away. As data infrastructure matures and model validation accumulates, the lenders willing to serve them on evidence-based terms rather than default bureau assumptions will capture a large and underpriced market.
Working on credit infrastructure or alternative data? We’d like to hear from you.