The Cost Stack Underneath Embedded Finance That Most Founders Never Model

What looks like a clean user experience on the front end is sitting on top of a cost stack with at least three distinct layers: customer acquisition, credit loss, and operational overhead.

Most founders building embedded finance products have modelled one of these carefully and estimated the other two. That is usually where the problem starts. The pay-later button, the instant credit approval, the embedded loan surfacing inside a supply chain platform — each looks like a distribution miracle until you model what it actually costs to acquire, underwrite, and operate at scale.

Founders rarely do this in full before the first tranche of capital is deployed.

Three Models, Three Cost Structures

Embedded finance reaches borrowers through three distinct distribution models. Most founders treat them as variations of the same business, but they are not.

50. Direct Digital Distribution: The lender acquires borrowers through owned or paid digital channels and underwrites using bureau data and alternate signals. The customer relationship, acquisition cost, compliance overhead, and credit risk are fully owned.
50. Supply Chain Partnership: Credit is extended to participants in an existing supply chain, with the anchor entity providing distribution and often the underlying transaction data. The borrower is known before the credit conversation begins.
50. Platform Partnership: A non-financial platform embeds credit into its user journey and passes the borrower to the lender at the point of need. The platform owns the customer relationship; the lender provides the credit capacity.

The unit economics of these three models diverge sharply at every layer of the cost stack.

Customer Acquisition: Where the Divergence Starts

The cost stack begins with customer acquisition, and this is where the three models first separate. What the lender pays to put a borrower on the books depends entirely on who owns the relationship at the point of origin. 

50. The Direct Digital Acquisition Cost

Direct digital acquisition is expensive, and it is getting more expensive. Every lender competes for the same digital real estate, the same bureau-verified borrower, the same search intent.

Acquisition costs in this model are fully borne by the lender, and they compound quickly once you include KYC verification, onboarding drop-off, and the cost of borrowers who complete the application but never draw down. The fully loaded cost of a direct digital acquisition is consistently higher than the headline CAC figure suggests.

50. The Supply Chain Acquisition Cost

Supply chain distribution eliminates most of this cost. The anchor entity already has the relationship. The borrower is known, the transaction history is live, and the distribution happens inside a workflow the borrower uses regardless of whether credit is available.

Acquisition cost here is largely replaced by partnership cost: the commercial arrangement with the anchor, the integration overhead, and the ongoing data access fee. The headline number is lower. The structure of the cost is different.

50. The Platform Acquisition Cost

Platform partnerships sit between the two. The platform brings the borrower to the point of need. The lender pays for access through a revenue share, an integration fee, or a first-loss arrangement. The acquisition cost is variable rather than fixed, which changes the risk profile significantly.

Credit Cost: The Layer Most Models Underestimate

Acquisition cost is visible on the P&L from day one. Credit cost accumulates quietly and surfaces later.

The direct digital borrower is underwritten with less contextual information than the supply chain or platform borrower. Bureau scores and alternate data signals are useful, but they are retrospective. They tell you what the borrower has done. They do not tell you what the borrower is doing right now, inside a transaction they need credit to complete.

The supply chain borrower is different in a structurally important way. The lender has access to live transaction data: purchase orders, payment histories, inventory cycles, receivable flows. The underwriting is prospective, grounded in the economic activity the credit is financing. Default rates in well-structured supply chain lending programmes are materially lower than in comparable direct digital portfolios, because the credit decision is better informed.

Platform-embedded credit sits closer to the supply chain model than the direct digital model, provided the platform shares meaningful transaction data with the lender. When it does, the credit decision improves. When it does not, the lender is effectively doing direct digital underwriting with platform distribution, which is the worst of both models.

Pavitra Pradip Walvekar, investor, entrepreneur, and builder of lending infrastructure from the inside, puts it directly: the credit cost differential between a well-structured supply chain or platform programme and a direct digital book is not marginal. It is the difference between a portfolio that compounds and one that requires constant capital top-up to stay performing. 

Operational Cost: What Regulatory Tightening Has Changed

Recent RBI directions have changed how fintech and NBFC-led lending models account for cost, capital, and risk. From January 1, 2026, the Co-Lending Arrangements Directions, 2025 require each regulated entity to retain at least 10% of every co-lending exposure. This limits full capital recycling and changes the economics of co-lending structures.

The Digital Lending Directions, 2025 also keep Default Loss Guarantee cover capped at 5%, while adding stricter due diligence, disclosure, and portfolio-level monitoring requirements for LSP-linked lending models.

In February 2026, the NBFC IRACP Amendment Directions clarified how DLG can be factored into Expected Credit Loss calculations, provided it forms part of the loan contract and is not separately recognised. While this may ease provisioning pressure, it adds compliance work through dynamic ECL tracking, DLG invocation monitoring, and enhanced disclosures.

For founders, these changes mean regulatory compliance is no longer a back-office concern. It directly affects unit economics, cost of funds, provisioning models, and operating margins.

Why the Platform Model Wins Under Regulatory Pressure

The three models do not respond equally to regulatory tightening. Every cycle rewards depth over breadth, and integration over acquisition.

50. Direct Digital: The Cost Multiplier

The direct digital model absorbs regulatory cost linearly. More LSPs mean more due diligence. More borrowers mean more compliance overhead per unit. The model does not have a structural mechanism to reduce this cost as it scales.

50. Supply Chain: The Efficiency Advantage

The supply chain model absorbs regulatory cost more efficiently. The anchor entity is a single compliance relationship rather than many. The transaction data that improves underwriting also simplifies the audit trail. The co-lending retention requirement affects this model too, but the tighter credit cost and lower default rate create headroom that the direct digital book does not have.

50. Platform: Where Compliance Concentrates

The platform model, when structured correctly, is the most capital-efficient under the new framework. The platform brings contextual data, pre-qualified distribution, and a commercial relationship that can absorb some of the LSP overhead through the partnership structure itself. The compliance requirement does not disappear, but it concentrates rather than multiplies.

The founders who chose the platform model for scale reasons ended up being right for structural reasons. The two are not the same argument. The platform model is the cheapest to operate under the current regulatory environment.

Model it Before You Scale it

Embedded finance will keep growing. The distribution models will multiply. The regulatory framework will keep tightening. None of that changes the underlying problem: most founders enter this space having modelled the product and not the business.

The cost stack is not hidden. It is just not where the attention goes. Acquisition sits in the marketing budget. Credit loss surfaces in the quarterly review. Compliance overhead accumulates in the background. By the time all three are visible together, the unit economics are already baked in.

Model the stack before you scale it. That is the only sequence that works.