K-shaped SMB lending: why cash flow underwriting needs context

TL;DR: In a Sept. 23 live session, Ocrolus SMB leaders David Snitkof and Patrick Shubert showed that 2026 SMB application data reveals a K-shaped market, with strong merchants improving and stressed merchants slowly degrading beneath stable aggregate numbers. Fuel exposure tiers, industry payroll benchmarks and deeper multi-lender stacking show why cash flow underwriting needs industry-level context.

Aggregate data on small business lending in 2026 suggests stability. Margins have recovered from a first-quarter fuel shock and the funding environment remains positive. One level down, the picture splits. In the Sept. 23 Ocrolus live session on the state of SMB lending, David Snitkof, general manager of SMB at Ocrolus, and Patrick Shubert, head of SMB data science, walked through application data showing a K-shaped market: strong merchants getting stronger while stressed merchants slowly degrade. For anyone practicing cash flow underwriting, a portfolio-level read hides the signals that matter most. As Snitkof put it, “There’s no single small business economy. There are many small business economies.”

Watch the full session with David Snitkof and Patrick Shubert.

The K-shaped SMB economy shows up in cash flow data

Ocrolus sees millions of SMB credit applications a year, with transactions tagged and businesses benchmarked by industry, which makes the split measurable.

Start with liquidity. When applicants are sorted into quartiles by nonsufficient funds (NSF) and overdraft activity, the strongest quartile has stayed stable while the weakest has seen those events climb further. Revenue follows a similar pattern: across two years of data for the industries Ocrolus examined, bottom-decile and top-decile monthly revenue appear to be diverging.

Industry cuts sharpen it. Finance and insurance topped the list of lagging industries by relative change in operating margin. Rates have risen over the past year, and cost-of-capital-sensitive businesses can’t always pass that through.

Transportation and warehousing, agriculture and forestry are pulling ahead, and Shubert noted the leaders are improving faster than the laggards are deteriorating. Stress in the bottom quartile is building slowly, the kind of drift an aggregate average smooths over.

Fuel exposure proved why segmentation beats intuition

Early in 2026, Ocrolus began segmenting applicants by fuel’s share of operating expenses: Tiers 1 through 4 plus a non-exposed group. When conflict in the Middle East triggered a fuel price shock in March, the tiers diverged sharply.

Intuition says trucking and transportation, the most fuel-exposed businesses, should have been hit hardest. Tier 2 took the largest hit instead, across access to credit, revenue, operating margin and performance on existing credit products. These are agriculture, construction and fuel station businesses where fuel is a large expense but pricing power is limited. Tier 1 operators can often pass costs to customers, and many prepurchase or hedge fuel.

“With the right amount of granularity and segmentation, you do find these kind of counterintuitive findings,” Shubert said. Fuel prices have stayed high, but margins have since recovered across segments as owners pulled other levers. Shubert said findings like this shape how lenders think about credit policy by sector and geography.

Industry benchmarks change what a healthy SMB looks like

Restaurants and construction firms are both overrepresented in SMB funding portfolios, and at the median their operating margins barely moved year over year. Underneath, they behave very differently.

The median restaurant applicant operates close to break-even, which makes the segment an early warning signal. NSF rates in accommodation and food service ticked up slightly, and restaurant margins are returning to baseline after the summer. Construction NSF rates sit below historical levels, and operating margins are lifting as the segment recovers from first-quarter fuel and shipping costs.

Revenue composition diverges too. Construction firms receive a meaningful share of their inflows as checks, while checks are rarely a major revenue source for restaurants. The same deposit pattern means something different depending on who’s applying.

Payroll is another example. More than half of the businesses Ocrolus sees are employers with payroll large enough to be a major operating expense. In health care, median payroll outflows run about 32% of revenue. For construction and restaurants, the figure is closer to 16% to 17%. Without that benchmark, Snitkof said, lenders can’t apply the proper context to an underwriting decision.

Snitkof said the thresholds for “what makes a good business or what makes a risky business” are “very different by industry.” Historically hard to automate, that nuance is now within reach thanks to advances in AI-driven cash flow analysis.

Stacking and fraud raise the stakes

The share of applicants with three or more financing counterparties on their bank statements has increased, while the share with no active financing has dipped. Merchants who stack are stacking deeper, and the healthiest borrowers appear to be waiting longer between positions.

Fraud compounds the risk. Generative AI has increased the number of fraud vectors and made convincing forged documents and synthetic identities more widely available. Shubert noted that some portfolios have tolerated a degree of misrepresentation because it could be offset through risk-based pricing, but fraudsters are now emboldened. His advice: practice good fraud hygiene and invest in document fraud detection, leaning “into the technology as much as the fraudsters have.”

The session’s takeaway fits in one word, repeated three times: context. Snitkof sees data, analytics and technology applied with that context as the path to broadening access to credit for small business owners. For more on the trends behind the session, download the latest State of SMB lending report.

Key takeaways

• SMB lending in 2026 is K-shaped: the strongest NSF quartile has held stable while the weakest has seen overdraft and NSF events rise.
• Finance and insurance led the lagging industries on relative operating margin change as rates rose, while transportation and warehousing, agriculture and forestry pulled ahead.
• The March fuel shock hit Tier 2 businesses (agriculture, construction, fuel stations) hardest, because they carry heavy fuel costs without the pricing power or hedging of Tier 1 transportation operators.
• Healthy looks different by industry: median payroll runs about 32% of revenue in health care versus 16% to 17% in construction and restaurants.
• More applicants now show three or more financing counterparties, and generative AI is making forged documents and synthetic identities easier to produce.

FAQs

What is a K-shaped economy in small business lending?

A K-shaped economy describes a market where one group of businesses keeps improving while another steadily deteriorates, even when aggregate numbers look stable. In 2026 SMB application data analyzed by Ocrolus, the strongest merchants show stable NSF and overdraft rates while the weakest show rising liquidity stress.

Why does industry context matter in cash flow underwriting?

The same cash flow metric means different things in different industries. Median payroll runs about 32% of revenue in health care and roughly 16% to 17% in construction and restaurants, and checks make up a far larger share of inflows for construction firms than for restaurants. Without industry benchmarks, lenders can’t apply the proper context to an underwriting decision.

How did 2026 fuel prices affect small business credit?

Ocrolus data shows the March 2026 fuel price shock hit Tier 2 fuel-exposed businesses hardest, including agriculture, construction and fuel stations. These businesses carry high fuel costs but have limited ability to pass them on. Margins have since recovered across segments even with fuel prices still elevated.

What is loan stacking in SMB lending?

Loan stacking occurs when a business carries multiple financing positions from different funders at the same time. Ocrolus application data shows the share of SMB applicants with three or more financing counterparties on their bank statements has increased, signaling deeper stacking among merchants already using credit.

How is generative AI changing fraud in SMB lending?

Generative AI has increased the number of fraud vectors and made convincing forged documents and synthetic identities more widely available. Ocrolus SMB leaders recommend consistent fraud hygiene and technology such as deepfake document detection to keep pace with fraudsters.