I've spent decades in business lending and credit analytics, and if there's one thing that hasn't changed since I started, it's this: small businesses are still harder to underwrite than they should be.
For the most part, they're thinly capitalized, their financials are often a patchwork of QuickBooks exports, Excel spreadsheets and bank statements. The owner's personal credit history frequently tells you more than the business itself does. That gap between what lenders need to know and what they can actually see is what make this industry genuinely interesting and challenging after all these years.
What's changed is the toolkit that bankers have access to. We're no longer limited to a credit file and a pile of tax returns. We have a growing ecosystem of data sources, each built to answer a different question about a business, and each earning its place through years of validation. That's the story I want to tell here, not a story about one tool replacing another, but about a set of complementary lenses that, together, give lenders a far more complete picture than any single one could offer alone.
The commercial bureaus and scoring models that built this industry, the business and personal credit files maintained by the major bureaus, and the small business scoring models built on top of them, have earned their place for good reason. They draw on years of trade line history, payment behavior, and public record data to produce risk rankings that have been tested across enormous volumes of originations, in many cases for decades. Lenders trust them because they're stable, well-documented, and deeply embedded in underwriting infrastructure, from portfolio risk management to the scoring thresholds used across SBA lending programs. When you need a consistent, defensible way to rank risk across a large book of business, these traditional models remain some of the best tools ever built for that job. Nothing about the rise of alternative data changes that. If anything, the newer sources work best when they're layered on top of a foundation this solid.
Even with tools like FICO® SBSS, access to capital remains a challenge. The Federal Reserve's 2026 Small Business Credit Survey found that only about half of small businesses had their funding needs fully met, while roughly one-third faced a funding gap despite applying for financing.
That's not a story about bad borrowers. It's a story about lenders working with an incomplete view of risk. Traditional credit data often misses the real financial picture: cash flow, payment behavior, seasonality, and revenue trends, especially for newer and smaller businesses.
When lenders rely solely on bureau-based data, creditworthy businesses can be overlooked. The opportunity is to combine traditional credit models with cash-flow analytics and alternative data to make better decisions, expand access to capital, and manage risk more effectively
This is where a newer generation of commercial credit models is adding value. Firms like Lumos Data aren't replacing traditional credit scores — they're strengthening them. By combining consumer credit data with decades of historical performance data, along with macroeconomic, industry, and business-specific variables, lenders gain a broader understanding of commercial credit risk. That richer assessment improves Probability of Default (PD) and Expected Loss (EL) predictions, providing a clearer picture of each applicant and supporting more informed underwriting decisions.
FinRegLab's independent study with NYU Stern researchers, which analyzed more than 38,000 small business loans, found that cash-flow variables derived from electronic bank account data provide a stronger and more accurate basis for predicting loan performance than personal credit scores alone, with the effect especially pronounced for newer businesses and owners with lower credit scores. Importantly, the strongest results came not from swapping one data source for another but from combining cash-flow data with traditional credit bureau inputs in a single underwriting model.
That's the whole point. The two data types aren't in competition. Just like I have been saying for years in the traditional models, that you have to blend the consumer and business data to get the right view, you have to blend traditional models and the alternative data to get the view that modern risk management demands. They're measuring different things, and the institution that deploys a model that uses both outperforms a model that uses either alone.
This matters because the stakes for getting it right are real. Small businesses make up 45.9% of private sector employment and 43.5% of GDP in this country, and access to credit is directly tied to whether they survive their fragile early years. A better underwriting decision today is a business that makes payroll, keeps its lease, and is still operating in five years.
None of this means throwing out what works. It means recognizing that bureau scores, cash-flow analytics, and other alternative signals each answer a piece of the same question from a different angle, and that the lenders who blend them thoughtfully are the ones best positioned to say yes to more good businesses without taking on bad risk. That's not a technology trend, it's just better underwriting.
I've watched this industry evolve from paper files to bureau scores to machine learning models, and the throughline has always been the same: better information leads to better decisions, and better decisions mean more small businesses get the capital they've earned.
The next chapter of this industry belongs to the lenders willing to build that multi-dimensional view, not by chasing the newest data source, but by putting the right combination of tools to work for the businesses in front of them. That's a future worth getting excited about, and it's one this whole ecosystem, old players and new, is building together.