Part 04 — Digital & Embedded Lending
Operating Models
Digital lending is not one model. In Indian small and medium enterprise (SME) credit, the same borrower journey can hide four very different risk and control structures: a regulated entity lending directly, a marketplace passing the lead, a lending service provider (LSP) doing substantial work for a lender, or a lending-as-a-service (LaaS) stack that combines software, sourcing, underwriting operations and sometimes an in-house non-banking financial company (NBFC). The operating model matters because the Reserve Bank of India (RBI) looks through the user interface to the regulated entity (RE) that books the loan. Under the Reserve Bank of India (Digital Lending) Directions, 2025, dated May 8, 2025, an LSP is an agent of an RE for functions such as customer acquisition, underwriting support, servicing, monitoring or recovery; outsourcing does not dilute the RE’s responsibility.
Direct Digital NBFC
Section titled “Direct Digital NBFC”A direct digital NBFC owns the app or web journey, the credit policy, the sanction, the loan account and collections. Examples in the broader SME market include digital-first lenders and NBFC arms that offer unsecured business loans, merchant cash advances, invoice discounting or short-term working-capital loans. In this model, the borrower may see only one brand, but the legally important point is that the NBFC is the lender of record.
The operating economics are closest to a traditional balance-sheet lender. The NBFC pays for acquisition through performance marketing, branch or partner sourcing, but keeps the net interest margin after cost of funds, expected credit loss, collection cost and operating cost. Typical unsecured micro and small enterprise (MSE) tickets are ₹1 lakh to ₹50 lakh for digital-originated term loans, with tenors of 3 to 36 months. Published market pages from lenders show similar boundaries: FlexiLoans advertises partner-originated collateral-free business loans of ₹1 lakh to ₹50 lakh with 48-72 hour approvals, while Indifi lists lender products up to ₹100 lakh with annual percentage rates (APR) varying by lender and product on its lending partners page.
Credit decisioning is not “the algorithm approved it.” In a serious NBFC, the model is a scorecard and rule engine wrapped by policy: negative industry filters, geography limits, bureau thresholds, bank-statement surrogate income, goods and services tax (GST) turnover, existing obligations, fraud checks and deviation approval. The underwriting file still has to support the RE’s assessment of creditworthiness because the RBI’s 2025 digital lending directions require the RE to obtain and keep borrower economic-profile information before lending.
Collections also remain the NBFC’s accountability. It may use call-centre or field agencies, but the customer must be told who is authorised to contact them when a recovery agent is assigned. Direct digital lenders therefore need a full stack: digital lending app (DLA), loan origination system (LOS), loan management system (LMS), bureau reporting, National Automated Clearing House (NACH) or e-mandate integrations, collections workflow, grievance handling, information-security controls and the public website disclosures required for DLAs, LSPs and products.
Marketplace Or Aggregator
Section titled “Marketplace Or Aggregator”A marketplace or aggregator is primarily a distribution and comparison layer. It may collect borrower intent, basic eligibility data and consent, then route the application to one or more REs. The customer relationship is shared: the platform owns the discovery moment, but the lender owns sanction, disbursement, servicing and grievance responsibility once a loan is issued. If the platform merely introduces leads and has no role in loan servicing or recovery, the model resembles a digital direct selling agent (DSA). If it collects data, performs lender matching, displays offers and supports fulfilment, it moves into LSP territory.
The important 2025 change is the treatment of multi-lender LSP journeys. Where an LSP has arrangements with multiple REs, the RBI requires a digital view of matching offers with lender name, amount, tenor, APR, monthly repayment obligation and penal charges, and prohibits dark patterns that push a borrower to a preferred lender. This directly affects SME loan marketplaces because the old “we will call you with the best offer” funnel is no longer enough for a compliant in-app loan comparison journey.
Marketplaces usually earn referral fees, processing-fee share or success fees. That income is more volatile than net interest margin but lighter on capital. The hard part is conversion quality. Lenders discount leads if GST turnover, bank statements, bureau data or industry classification arrive late or inconsistent. A mature marketplace therefore invests in pre-screening and consent plumbing: Permanent Account Number (PAN), Udyam, GST, bank statement or Account Aggregator (AA), bureau consent, constitution documents and proprietor/director know-your-customer (KYC).
LSP-Fronted Lending
Section titled “LSP-Fronted Lending”In the LSP-fronted model, the platform is the visible front end and the RE is the lender behind it. This is common in payment-processor merchant loans, business-to-business (B2B) marketplace credit, software-as-a-service (SaaS) seller finance and supply-chain finance. The LSP may provide embedded user experience, alternate data, pre-approved offer generation, document capture, repayment reminders and first-level customer support. The RE supplies the regulated balance sheet, approves credit policy, books the loan and reports to credit information companies (CICs).
The customer experience can feel like “my platform gave me a loan”, but regulatory attribution must be explicit. The Key Facts Statement (KFS), sanction letter and loan agreement must identify the lender and costs. Disbursement must go from the RE to the borrower’s bank account, except for permitted cases such as specific end-use disbursal to an end-beneficiary or co-lending flows. Repayments must come directly to the RE’s account, not to a third-party pool account of the LSP. Those fund-flow rules are central to regulatory guardrails applied.
In B2B procurement finance, the LSP-fronted model often looks cleaner because a specific-end-use loan can be disbursed directly to the supplier; see B2B Commerce & Procurement Finance.
Credit risk sits with the RE unless there is a permitted default loss guarantee (DLG). The 2025 directions allow DLG only within the specified framework: the provider must be eligible, the cover must be contractually defined, and the cap is 5 percent of the disbursed amount of the fixed DLG portfolio. That cap changes platform economics. A sourcing platform cannot promise first-loss protection of 10-20 percent to make a weak book bankable; it has to improve actual underwriting, pricing and collections.
Lending-As-A-Service
Section titled “Lending-As-A-Service”LaaS is a deeper model than lead generation. A LaaS provider supplies reusable credit infrastructure: APIs, rule engine, product configuration, lender routing, co-lending orchestration, dashboards, portfolio monitoring, reconciliation and sometimes its own NBFC. Glaas describes itself as embedded credit infrastructure with API/software development kit (SDK), sandbox, underwriting engine and in-house NBFC capability through Gromor Finance on its official site. U GRO Capital describes itself as a DataTech NBFC with embedded financing and co-lending partners; public reporting and company pages refer to GRO Score, GRO Xstream and embedded financing, not “GLASS” (UGRO embedded financing).
The customer owner in LaaS is negotiated. A platform may keep the user interface and brand, the LaaS provider may operate the credit journey, and the RE may own the legal borrower relationship. Credit decisioning may be “joint” in workflow terms, but the RE cannot outsource accountability. A bank or NBFC can use platform data and a LaaS score, yet the sanction must be under its policy or an approved co-lending arrangement.
The economics can combine software fees, per-application fees, success fees, servicing fees, excess spread where allowed, co-lending income and DLG fees. The risk is complexity: three ledgers may exist for one borrower relationship - platform order ledger, lender loan ledger and LaaS reconciliation ledger. Failures show up as customer disputes, unposted repayments, incorrect CIC reporting or KFS mismatches.
Operating Model Comparison
Section titled “Operating Model Comparison”| Model | Customer owner | Credit decision | Balance sheet | Collections | Main revenue |
|---|---|---|---|---|---|
| Direct digital NBFC | NBFC | NBFC policy and underwriters | NBFC | NBFC or its agents | Interest, fees |
| Marketplace | Platform until lender handoff | Lender | Lender | Lender | Referral/success fee |
| LSP-fronted | Platform front end, RE legal lender | RE, using LSP data/services | RE; DLG only if compliant | RE accountable; LSP may assist | Service fee, sourcing fee, DLG fee |
| LaaS | Shared by platform/LaaS/RE | RE or co-lenders using platform stack | REs, sometimes LaaS group’s NBFC | Contracted but RE accountable | Platform fee, servicing fee, co-lending economics |