Embedded Lending: How Platforms Use Transaction Data to Underwrite Loans at Checkout

Traditional small business lending relies on credit bureau scores, tax returns, and months of back-and-forth with loan officers. Embedded lending flips this model: platforms that already process a merchant's daily transactions use that data to underwrite loans in seconds, disbursing capital without the merchant ever leaving the dashboard. This approach, known as contextual underwriting, replaces static credit assessments with real-time signals like sales velocity, refund rates, and customer retention patterns.

The results are significant. Shopify Capital originated $1.4 billion in merchant cash advances and business loans in Q1 2026 alone, a 71% increase from Q1 2025. PayPal surpassed $30 billion in cumulative small business lending by March 2025. These are not banks: they are software platforms that happen to have better underwriting data than any bank ever will.

What Is Contextual Underwriting?

Contextual underwriting is a credit decision framework that evaluates borrowers using platform-native operational data rather than (or in addition to) traditional credit scores. Instead of pulling a FICO score and requesting two years of tax filings, the platform already observes the merchant's business in real time.

The signals that power contextual underwriting vary by platform type, but they share a common trait: they are generated organically through the merchant's normal use of the platform, not collected through a separate application process.

Platform-Specific Underwriting Signals

• Inventory velocity: how quickly products sell through, indicating demand health
• Daily gross merchandise volume (GMV): revenue consistency and growth trajectory
• Refund and chargeback rates: a proxy for product quality and customer satisfaction
• Payment discipline: whether the merchant pays suppliers and platform fees on time
• Repeat purchase rates: customer loyalty as a predictor of future cash flows
• Fulfillment reliability: shipping speed, order accuracy, and dispute resolution patterns
• Seasonal patterns: historical revenue cycles that inform repayment scheduling

Why platform data beats bureau data: A FICO score tells you whether someone paid their credit card bill last month. Platform transaction data tells you how many units they sold today, whether their customers are returning, and whether their revenue is accelerating or decelerating. For working capital decisions, the latter is far more predictive.

Leading Embedded Lending Platforms

The largest embedded finance lending operations are run by platforms that initially had nothing to do with financial services. They built lending as an extension of their core business, leveraging proprietary data advantages that traditional lenders cannot replicate.

Shopify Capital

Shopify Capital launched in 2016 and has become one of the largest platform lenders globally. The program uses merchant store data to pre-qualify sellers for loans and merchant cash advances without requiring a credit check or formal application. In Q1 2026, Shopify originated $1.4 billion in financing, up from $821 million in Q1 2025. Full-year 2025 originations reached $4.2 billion.

In November 2025, Shopify launched Capital Flex, a revolving credit product that gives merchants continuous access to funds with a dynamic limit that adjusts based on real-time business metrics. This represents an evolution from one-time loan offers to persistent credit facilities powered by live data.

PayPal Working Capital

PayPal's lending arm surpassed $30 billion in cumulative originations by March 2025, having extended more than 1.4 million loans and cash advances to over 420,000 business accounts since 2013. Underwriting is based on PayPal payment history: merchants who process payments through PayPal receive pre-approved offers calculated from their transaction volume and account standing.

Repayment is automatic: PayPal deducts a fixed percentage of each sale until the advance is repaid. This structure aligns repayment with revenue, reducing default risk during slow periods.

Amazon Lending

Amazon Lending, launched in 2011, has extended over $8 billion in financing to marketplace sellers. The program is invitation-only: Amazon uses its proprietary seller performance data to identify qualified merchants and present pre-approved offers. Interest rates range from 3% to 16.9% APR, with loan terms typically between 3 and 12 months.

Stripe Capital

Stripe Capital originated 81,000 merchant cash advances and business loans in 2025, with estimated funding volume between $800 million and $1.2 billion, representing 45% year-over-year growth. Stripe's data shows that businesses accepting Capital offers grew 27 percentage points faster over the following year than comparable businesses that declined.

Vertical-Specific Entrants

The embedded lending model has expanded beyond general e-commerce and payment processors. Toast Capital underwrites restaurant loans using point-of-sale transaction data that traditional banks never see: ticket sizes, table turnover rates, and seasonal dining patterns. Toast accumulated approximately $1 billion in loan balances by mid-2024 and generated $5 billion in total financial services revenue in 2025.

In Latin America, Mercado Libre's Mercado Crédito extended over $3.3 billion in loans in 2023, targeting small businesses and consumers across Brazil, Argentina, and Mexico using marketplace transaction data. The platform's closed-loop model spanning payments, credit, and insurance gives it underwriting visibility that standalone lenders in the region cannot match.

Platform Lending Compared

*Shopify, PayPal, and Square combined for $6.7 billion in SMB loan originations in 2025 within the broader $1.4 trillion SMB lending market.

Default Rates and Approval Rates Compared

The core promise of contextual underwriting is better risk assessment. If platforms genuinely understand their merchants' businesses in real time, default rates should be lower than traditional underwriting approaches that rely on stale financial statements and credit scores.

The data supports this claim, though with important caveats. Platform-data-driven SMB lending produces annualized default rates of roughly 3% to 6%, compared to 8% to 14% for traditional "cold underwritten" SMB loans where the lender has no prior relationship with the borrower. The average business loan default rate across all lender types in the United States sits at approximately 7.5%. As of March 2026, the trailing twelve-month default rate for SBA 7(a) loans reached 4.8%, its highest level since 2013.

The data advantage compounds over time: Platform lenders improve their models with each loan cycle. A merchant who successfully repays a first advance generates additional behavioral data that refines the risk model for their next offer. Shopify reports that repeat borrowers represent a growing share of originations, with repeat borrower performance improving quarter over quarter.

The Technology Stack Behind Embedded Lending

Embedded lending requires tight integration between the platform's core infrastructure and lending-specific components. The technology stack typically spans four layers.

Data Ingestion and Feature Engineering

The platform's transaction processing system feeds real-time and historical data into a feature store. Raw events (individual sales, refunds, disputes) are transformed into underwriting features: 30-day rolling revenue, refund rate percentiles, customer retention cohorts, and growth acceleration metrics. This is where payment gateway data becomes credit intelligence.

Risk Modeling and Decisioning

Machine learning models evaluate the feature set against historical loan performance to produce a risk score, offer amount, and pricing. These models retrain continuously as new repayment data arrives. The decisioning engine runs automatically, generating pre-qualified offers that appear in the merchant's dashboard without any action on their part.

Loan Origination and Servicing

Once a merchant accepts an offer, the origination system executes the loan agreement, handles regulatory disclosures, and triggers disbursement. Servicing is largely automated: repayment deductions occur at the transaction level, and the platform monitors the merchant's ongoing performance for early warning signals.

Disbursement and Settlement

This is where traditional infrastructure creates friction. Most embedded lenders disburse via ACH transfer, which takes one to three business days. A merchant who needs working capital today receives an instant approval but then waits for the banking system to move the money. Some platforms offer same-day or next-day funding at a premium, but the underlying payment rails remain the bottleneck.

Regulatory Landscape for Platform Lenders

Embedded lending introduces regulatory complexity that scales with the platform's level of involvement in the credit decision. The fundamental question: when does offering loans through your platform require a banking or lending license?

Three Regulatory Models

Platforms typically choose one of three approaches to navigate lending regulations, each with different compliance burdens and economics.

1. Partner with a licensed lender: The platform provides the distribution channel and data; a licensed bank or lender originates the loans. This is the most common approach for early-stage programs. The platform avoids the need for state-by-state lending licenses but gives up margin and some control over underwriting.
2. Use a bank partner with platform-controlled underwriting: The platform licenses an embedded lending technology provider and separately contracts with an originating bank partner. This gives the platform more control over underwriting policy and unit economics but requires internal compliance, risk, and operations capability.
3. Obtain lending licenses directly: The platform acquires state-by-state lending licenses, builds underwriting in-house, and originates on its own balance sheet or through warehouse credit lines. This is what Shopify, Square, and Amazon have done at scale. The regulatory, capital, and operational burden is substantial.

Key Compliance Requirements

Regardless of the model chosen, platforms offering embedded lending must navigate several regulatory requirements. FCRA compliance governs how consumer and business credit data is used in underwriting decisions. State-level lending licenses impose jurisdiction-specific rules on interest rates, disclosure requirements, and collections practices. KYC and AML requirements apply to loan disbursement, particularly for larger amounts. Truth in Lending Act (TILA) disclosures are required for many loan products, though merchant cash advances have historically occupied a gray area.

The regulatory picture is evolving. As platform-originated lending volumes grow (Bain & Company projected US B2B embedded lending to reach $50 to $75 billion by 2026), regulators are paying closer attention to whether these programs require the same oversight as traditional lenders.

Limitations of Contextual Underwriting

Contextual underwriting is powerful, but it has structural limitations that traditional credit assessment does not.

• Platform lock-in: merchants can only borrow based on their activity on a single platform; off-platform revenue is invisible to the model
• Survivorship bias: underwriting models are trained on merchants who already succeeded enough to receive offers, potentially missing creditworthy businesses in earlier stages
• Cyclical blindness: platforms with limited operating history may not have data spanning a full economic downturn, leaving models untested in recessionary conditions
• Concentration risk: if the platform itself experiences disruption (policy changes, fee increases, algorithm shifts), loan portfolios can deteriorate simultaneously across many borrowers
• Cost of capital: platform lenders often carry higher funding costs than banks, which translates to higher borrower rates despite lower default rates

The Disbursement Problem

Even the fastest underwriting engine is only as fast as the money it moves. A merchant receiving an instant approval still faces a one-to-three day wait for ACH settlement in the United States. For a restaurant owner who needs to pay a supplier tomorrow or a marketplace seller who needs inventory before a flash sale, that delay is material.

Some platforms mitigate this by pre-funding from their own balance sheets and settling with the bank later. Others charge a premium for faster access. But the root cause is the same: ACH and wire transfer rails were not designed for instant, 24/7 settlement. Even FedNow, which supports real-time payments, has limited adoption and transaction size caps that restrict its use for business lending disbursement.

Stablecoin Rails as a Settlement Layer

Stablecoins offer a potential solution to the disbursement bottleneck. A dollar-denominated stablecoin settles in minutes regardless of time of day, day of week, or banking holidays. For embedded lending, this means the entire workflow from underwriting decision to funds in the merchant's account could happen in under five minutes.

This is already happening at scale in some corridors. Brookings Institution research documents how stablecoin rails are enabling faster financial services distribution in emerging markets. In September 2026, Tether and Fasanara Capital launched StableFund, a $400 million credit vehicle using stablecoin settlement infrastructure to fund SME and consumer lending across more than 60 countries.

For cross-border embedded lending, the advantage is even more pronounced. A platform like Mercado Libre lending to sellers across Brazil, Argentina, and Mexico currently navigates multiple national settlement systems with different operating hours and currencies. Stablecoin disbursement on a network like Spark could unify that settlement layer: one rail, instant finality, and dollar-denominated value that merchants can hold or convert locally.

The Convergence Ahead

Embedded lending is converging with several adjacent trends that will reshape how platform capital flows to merchants.

AI-Enhanced Underwriting

Machine learning models are incorporating increasingly diverse signals beyond transaction data. Natural language processing analyzes customer reviews and support tickets for early warning signs. Computer vision evaluates product listing quality. The underwriting feature space is expanding far beyond what any loan officer could evaluate manually.

Programmable Repayment

Programmable payments enable repayment terms that adapt dynamically to business conditions. Instead of a fixed percentage deduction, smart contract logic could adjust repayment rates based on real-time revenue thresholds: higher deductions during strong months, lower during slow periods, with hard floors to protect the lender.

Open Banking and Data Portability

Open banking frameworks are beginning to address the platform lock-in problem. If merchants can authorize sharing their transaction data across platforms, embedded lenders could underwrite based on a merchant's full business picture rather than a single-platform view. The banking-as-a-service infrastructure supporting this is maturing rapidly.

Instant Settlement Lending

The combination of contextual underwriting and stablecoin-based disbursement points toward a future where the entire lending cycle collapses into minutes. A merchant on a platform processes a sale, the underwriting model updates in real time, a new credit offer appears, and upon acceptance, stablecoin funds settle instantly. No application forms, no waiting for ACH, no business hours restrictions.

Bitcoin Layer 2 networks like Spark are well-positioned for this use case. With support for stablecoins like USDB, instant settlement, and transaction fees measured in fractions of a cent, the disbursement layer no longer needs to be the bottleneck. Developers building embedded lending products can explore the Spark SDK and documentation for integration options, and users looking for wallets that support stablecoin-based financial services can try General Bread, a Spark-powered wallet.

For a broader view of how embedded financial services are reshaping platform economics, see our research on buy now, pay later economics and the composable payment infrastructure stack.

This article is for educational purposes only. It does not constitute financial or investment advice. Bitcoin and Layer 2 protocols involve technical and financial risk. Always do your own research and understand the tradeoffs before using any protocol.