Lenders evaluating an embedded finance deal tend to make the same assumption. Letting a marketplace or software platform originate the loan, they think, means handing over underwriting too. That assumption is only partly correct. At Finezza, this is the gap we see trip up most embedded lending partnerships: what actually shifts, and what a lender assumes shifts, aren’t the same thing.
Key Takeaways
- Embedded finance relocates the underwriting data source from periodic bureau pulls to the anchor platform’s live transaction history.
- RBI’s Digital Lending Directions, 2025 classify anchor platforms as Lending Service Providers (LSPs) and bar them from making the credit decision.
- Regulated Entities (REs), meaning banks and NBFCs, stay fully liable for an LSP’s conduct even when the anchor originates the borrower relationship.
- Default Loss Guarantee (DLG) arrangements between lenders and anchors are capped at 5% of the disbursed loan portfolio.
- Lenders need origination systems that can score borrowers continuously from streaming anchor data, not only at the point of application.
- Pre-approved, anchor-triggered limits require the underwriting engine to run ahead of the borrower’s request, while the sanction decision stays inside the lender’s own system.
What Moves When Embedded Finance Sits Inside an Anchor’s Platform
The anchor, a business marketplace, a subscription software platform, or a direct-to-consumer aggregator, sits on live transaction data that a lender’s own origination flow never sees at application time: order volumes, payment cycle length, seller ratings, gross merchandise value (GMV) trend, and return rates.
For a seller with three years of consistent order flow but a thin credit file, that transaction trail often predicts repayment better than a bureau score pulled once, at the moment of application.
This is the part that moves inside the anchor’s platform: the data capture surface, not the decision built on top of it. Lenders already working with alternative underwriting signals will recognise the logic. It’s the same reasoning behind looking beyond a single bureau score for borrowers whose formal credit history undersells their actual repayment capacity.
Picture a B2B marketplace that bakes a seller credit line into its checkout flow. A seller with six months of trading history and a thin bureau file would ordinarily get declined or pushed into manual review. The marketplace, though, already tracks order fulfilment rates, return patterns against category peers, and whether GMV is climbing or flattening quarter over quarter. None of that shows up on a bureau report. The lender’s underwriting logic doesn’t change here: the inputs feeding it do, and that’s a narrower shift than most partnership conversations assume.
How Embedded Finance Turns Underwriting Into a Continuous Process
Most loan origination systems were built around a single moment: a borrower applies, submits bank statements and tax filings, and the system pulls bureau data to score them. Embedded finance breaks that moment apart.
Instead of one intake event, the anchor’s application programming interface (API) pushes transaction and behavioural data continuously. The lender’s decision engine has to score against it before the borrower ever opens a loan application, not after.
It’s a similar shift to what’s already happening with Account Aggregator-verified cash flow data in MSME lending, where structured financial data now arrives in minutes instead of through document uploads.
A pre-integrated lending stack that already talks to bureaus, payment gateways and e-sign providers over API makes this easier to build. A lender still running on batch uploads and manual reconciliation will find that the embedded finance infrastructure a workable anchor partnership demands looks nothing like their current intake process.
What RBI Won’t Let Move: The Credit Decision
This is where the assumption breaks down. Under RBI’s Digital Lending Directions, 2025, a Lending Service Provider (LSP), which is what an anchor platform becomes the moment it starts sourcing borrowers for a lender, can support underwriting and pricing on the lender’s behalf. What it cannot do is make the credit decision.
That stays with the Regulated Entity (RE), the bank or NBFC funding the loan. RBI states plainly that an outsourcing arrangement with an LSP doesn’t reduce the RE’s own obligations, or its liability for what the LSP does on its behalf.
Risk-sharing arrangements exist too. A Default Loss Guarantee (DLG), where the anchor or LSP absorbs part of the default risk, is permitted, but it’s capped at 5% of the disbursed loan portfolio. Even a generous DLG arrangement covers a sliver of exposure; the lender’s own underwriting still carries the rest. The Directions also require any DLG to be invoked within 120 days of default, so it functions as a backstop against portfolio-level losses rather than a per-loan safety net a credit team can lean on account by account.
Anyone who has built dual underwriting workflows for co-lending will recognise the shape of this problem: shared origination, but a decision that has to sit clearly with one accountable party.
Building an Underwriting Stack That Can Sit Inside an Anchor’s Platform
For a lender weighing one of these partnerships, the practical questions aren’t about the embedded finance model itself; that part is well understood by now. They’re about whether the origination system can pull an anchor’s data stream through an API, score it against internal rules and bureau data together, and still route the final sanction through the lender’s own approval workflow, not the anchor’s checkout screen.
A lender treating embedded finance purely as a distribution channel for lending usually finds out the hard way that anchor data quality varies between platforms. Credit scoring models built only for bureau and bank statement inputs don’t generalise well to transaction-pattern data without retraining.
Onboarding an anchor properly usually means a sandbox period where the lender’s team maps the anchor’s raw event fields (order status changes, refund triggers, payout timestamps) to whatever the scoring model actually consumes, rather than assuming a clean one-to-one match. Two marketplaces selling into the same category often structure their transaction logs differently, and a mapping that works for one rarely transfers cleanly to the next without adjustment.
It helps to pair anchor signals with the data points a bureau score misses, rather than swapping one for the other.
Get the architecture right, and the payoff is concrete: a seller on the anchor’s platform sees a pre-approved limit before they’ve asked for one, and lenders get a real way to cut loan turnaround time on segments that were previously too thin-file to underwrite profitably.
Frequently Asked Question
1. What is embedded finance in the context of lending?
Embedded finance places a lending product inside a non-financial platform, such as a marketplace or software tool, at the exact point where a business or customer needs credit.
2. Who owns the underwriting decision in an embedded finance partnership?
The Regulated Entity, meaning the bank or NBFC funding the loan, owns the underwriting decision. The anchor platform can supply data but cannot make the sanction call under RBI’s digital lending rules.
3. Can an anchor platform share default risk with the lender?
Yes, through a Default Loss Guarantee arrangement, but RBI caps this cover at 5% of the disbursed loan portfolio, so it doesn’t replace the lender’s own underwriting.
4. What data do anchor platforms typically feed into underwriting?
Order volumes, payment cycle history, seller ratings, GMV trends and return rates- signals that often predict repayment behaviour better than a single bureau score for thin-file borrowers.
5. What changes technically for a lender’s loan origination system?
It needs API connectivity to the anchor’s data streams and a decision engine that can score borrowers continuously, not only when a loan application gets submitted.
Conclusion
The practical takeaway is to treat these as two separate questions: what can move closer to the anchor’s transaction data, and what has to stay inside the lender’s own system regardless of who originates it. Get that split wrong, and the partnership either stalls at compliance review or clears review and creates the liability gap RBI’s directions were written to close.
Finezza’s low-code loan origination system is built API-first to connect with external data sources alongside bureau and bank statement inputs, while keeping the sanction decision inside the lender’s system. That distinction, what the platform ingests versus what it decides, matters before signing an anchor agreement. If you’re weighing an anchor partnership and want to see how the underwriting layer would work against your portfolio, book a demo and we’ll walk through it.
Leave a Reply