How Square's Sub-4% Loss Rates Prove the Case for AI Underwriting for Merchant Cash Advance

Square's Loss Rate Benchmark Is a Wake-Up Call for Independent Funders

Block CEO Jack Dorsey's Q2 2026 shareholder letter contained a line that should keep every independent MCA funder awake at night: "Square Loans are underwritten on data banks can't see and serve sellers banks won't. Our loan cohorts have had loss rates of less than 4% through every cycle we've seen." That is not marketing language. It is a structural claim about the power of AI underwriting for merchant cash advance built on top of proprietary transaction data.

Square processes every card swipe, every invoice, every refund flowing through its ecosystem. By the time a seller requests capital, Square already knows daily revenue trends, seasonal patterns, refund ratios, and customer concentration. The underwriting decision is not a guess. It is a calculation derived from months of continuous observation. Traditional funders, by contrast, receive four bank statements in a PDF, often photographed on a phone, and try to reconstruct the same picture manually.

The question is not whether Square's approach is better. The results speak for themselves. The question is what independent funders can do about it. The answer starts with closing the data gap, and that means rethinking how merchant documents get collected, parsed, and turned into underwriting signals.

Why Proprietary Data Creates a Structural Underwriting Advantage

The Platform Lender's Data Moat

Square, Shopify, and PayPal share a common advantage that no ISO broker or independent funder can easily replicate: they sit between the merchant and the merchant's customers. Every transaction that flows through the platform generates a data point. Over time, those data points form a living financial profile that updates in real time.

Shopify Capital, for instance, originated $1.4 billion in small business loans and MCAs in Q2 2026. As covered in our analysis of Shopify's MCA-to-loan transition in Canada, Shopify can underwrite with confidence because it already knows a merchant's gross merchandise volume, average order value, and fulfillment speed before a single document is uploaded. The bank statement is almost redundant.

For independent funders, though, the bank statement is everything. It is the primary window into a merchant's financial health. And therein lies both the challenge and the opportunity.

How AI Closes the Data Gap for Independent Funders

Independent funders will never have Square's transaction feed. But they can extract far more intelligence from the documents they do receive. The difference between a funder who glances at monthly totals and one who uses AI to decompose every deposit, flag every NSF, categorize every recurring debit, and detect anomalous patterns is the difference between guessing and underwriting.

Modern AI-powered bank statement analysis does several things that manual review cannot match at scale. First, it normalizes data across hundreds of bank formats, from TD to Chase to regional credit unions, so that revenue figures are always comparable. Second, it identifies deposit velocity patterns that reveal whether a merchant's revenue is stable, growing, or artificially inflated by a recent lump-sum transfer. Third, it flags red flags like round-number deposits, sudden balance spikes before statement dates, and unexplained transfers between related accounts.

These are the same categories of signals that Square derives from its proprietary data. The difference is that Square gets them passively, while independent funders must extract them actively. The funders who automate that extraction process win. The ones who rely on a human analyst scanning PDFs page by page fall behind, not because the analyst lacks skill, but because the volume overwhelms the process.

Speed and Accuracy Are Not Competing Goals

A common objection to AI underwriting is that speed comes at the cost of accuracy. The data does not support this. Square's sub-4% loss rate is achieved at massive scale, with thousands of underwriting decisions made algorithmically every week. The speed is the point. Faster decisions mean merchants get funded before they seek capital elsewhere, and the lender captures deals that slower competitors lose.

For independent funders, the speed bottleneck is rarely the underwriting decision itself. It is the intake process. Getting four months of bank statements from a merchant who is juggling a business, collecting a government ID, obtaining a signed application, and assembling all of it into a clean file for review takes days when done manually. As we explored in our piece on how AI document extraction speeds up MCA underwriting, automating that intake step compresses the timeline from days to hours without sacrificing data quality.

Let's Submit addresses this directly. When a merchant receives a secure upload link by text, they can drop bank statements, IDs, and signed applications from their phone in minutes. AI extraction then parses the statements automatically, pulling revenue, deposits, daily balances, and NSF counts into a structured format ready for review. The funder's team never touches the raw documents until the data is already organized.

What Independent Funders Can Actually Replicate

It is tempting to dismiss Square's performance as unreachable for smaller shops. After all, Square has billions of dollars in engineering resources and a captive merchant base. But the core principles behind its success are replicable if you break them down.

The first principle is continuous data observation. Square watches merchants in real time. Independent funders obviously cannot do this pre-funding, but they can do it post-funding through ACH monitoring and renewal-stage bank statement collection. Building a longitudinal view of merchant performance across multiple funding cycles gives repeat funders a data advantage that approaches what platform lenders enjoy.

The second principle is automated pattern recognition. Square's algorithms detect risk signals that human reviewers would miss or catch inconsistently. Independent funders can deploy the same class of technology on bank statements. Machine learning models trained specifically on MCA merchant financials can identify fabricated cash flow patterns, stacking indicators, and revenue seasonality with a consistency that no manual process can match. The key is using models purpose-built for lending documents, not general-purpose AI that treats a bank statement like any other PDF.

The third principle is friction-free merchant interaction. Square sellers never fill out a loan application. The offer appears in their dashboard, and they accept with a click. Independent funders will always require more from merchants, but the gap can be narrowed dramatically. Asynchronous document collection, where a merchant uploads everything from a single mobile-friendly link at their convenience, removes the back-and-forth emails and faxes that slow traditional intake. When the process feels easy, merchants complete it faster and abandon it less often.

Taken together, these three principles form a playbook. You do not need to be Square. You need to extract maximum intelligence from the data you can access, automate the repetitive parts of your workflow, and make it painless for merchants to give you what you need.

Why Institutional Capital Now Demands AI-Grade Underwriting

Loss rates matter beyond the P&L. They determine whether a funder can access institutional capital at attractive terms. When Square reports sub-4% losses through every cycle, it sets a benchmark that warehouse lenders and securitization buyers use to evaluate the rest of the market.

Funders seeking credit facilities or investment-grade ratings in 2026 face increasingly detailed questions about their underwriting methodology. Institutional partners want to see systematic, repeatable processes, not ad hoc human judgment. They want to know that the same merchant profile produces the same decision regardless of which analyst reviews it. AI-driven underwriting provides that consistency. Manual underwriting, by definition, does not.

This dynamic creates a feedback loop. Funders who adopt AI underwriting tools produce more consistent portfolios, which attract cheaper capital, which allows them to offer more competitive terms, which attracts better merchants, which further improves portfolio performance. Funders who resist automation find themselves locked out of the capital markets that would let them scale.

The Federal Reserve's Survey of Household Economics continues to show that small businesses increasingly seek non-bank financing. The opportunity is growing. The question is which funders have the infrastructure to capture it without blowing up their loss rates.

Frequently Asked Questions

What data does Square use for underwriting that banks cannot see?

Square underwrites on real-time transaction data flowing through its payment processing platform. This includes daily sales volume, average transaction size, refund rates, customer repeat purchase frequency, and seasonal revenue patterns. Traditional banks only see aggregated deposit totals on monthly statements, missing the granular transaction-level detail that reveals true merchant health. Independent MCA funders can approximate some of these signals through AI-powered bank statement analysis that decomposes deposit patterns and flags anomalies.

Can independent MCA funders achieve loss rates comparable to platform lenders?

Matching sub-4% loss rates requires systematic changes to the underwriting process, not just better analysts. Independent funders who combine AI document extraction, automated cash flow analysis, and structured data collection from merchants can significantly close the gap. The key is eliminating inconsistency in how bank statements are read and interpreted. Purpose-built AI models trained on lending documents deliver the repeatable accuracy that institutional capital partners demand.

How does AI-powered bank statement analysis work for MCA lending?

AI-powered bank statement analysis uses optical character recognition and machine learning to extract structured data from bank statement PDFs or images. The system identifies deposits, withdrawals, daily balances, NSF charges, and recurring debits across multiple months and bank formats. It then normalizes the data so that underwriters can compare merchants on a consistent basis. Advanced systems also flag fraud indicators like fabricated round-number deposits, unusual inter-account transfers, and balance manipulation near statement close dates.

Why does document collection speed affect MCA underwriting quality?

Slow document collection does not just delay funding; it degrades data quality. Merchants who are chased for documents over multiple days often submit incomplete files, outdated statements, or low-quality photos. By the time the file is complete, the merchant may have already accepted funding from a faster competitor. Async collection tools that let merchants upload everything from a single mobile link in one session produce cleaner, more complete files and reduce abandonment rates significantly.

Conclusion

Square's sub-4% loss rates are not magic. They are the result of underwriting on rich, continuous data with automated decision systems. Independent MCA funders cannot replicate Square's transaction feed, but they can replicate the principles: extract maximum intelligence from available data, automate repetitive analysis, and remove friction from merchant interactions.

Let's Submit is built for exactly this workflow. From async document collection via mobile upload links to AI-powered bank statement extraction that pulls revenue, balances, and red flags automatically, the platform compresses the gap between what platform lenders know and what independent funders can learn. Visit letssubmit.ca to see how automated intake and AI extraction fit into your underwriting process.