Underwriting the Invisible Borrower: A Practitioner's Guide to MSME Credit Risk in India

How India's 64 million small businesses fall through the cracks of a lending system built for someone else — and how the data, the rails, and the judgment are finally coming together to reach them.

India has a credit paradox. There is no shortage of capital — banks and NBFCs are flush, and the government has spent a decade building world-class financial rails. And yet, of the country's roughly 64 million MSMEs, only about 14% have access to formal credit. Estimates of the unmet demand run from ₹25–30 lakh crore (the addressable near-term gap) to as high as ₹80 lakh crore in total. Nearly half of all MSME credit demand simply goes unserved.

I've spent two decades on both sides of this problem — first building credit-risk and trade-data software for Fortune 500 lenders in the US, then building an instant B2B credit-rating platform for India's MSMEs. And the single most important thing I've learned is this: the MSME credit gap is not a capital gap. It's a risk-assessment gap. We don't lack money to lend. We lack the ability to see the businesses that need it most.

This is a guide to that problem — and to the practical craft of solving it.

Why the borrower is invisible

The lending playbook most Indian institutions inherited was built for a borrower who barely exists in the MSME segment. That playbook assumes a deep, clean credit bureau file, years of audited financial statements, and a long trail of formal tradelines. It works beautifully where the data is already there. Point the same model at a kirana supplier in a tier-3 town, a two-person fabrication unit, or a small D2C brand, and it returns the only honest answer it can: insufficient data. The system reads that as decline.

The borrower isn't risky. The borrower is invisible. Three structural features create that invisibility:

• Informality. Much of the business runs on cash and undocumented transactions, so the financials that exist often understate the true scale of the enterprise.
• Thin or absent bureau files. A first-time borrower has no credit history, and a model trained to weigh history heavily has nothing to weigh.
• A collateral-first mindset. Most micro and small enterprises don't have clean, mortgageable assets — and where they do, weak land records make enforcement slow and uncertain.

The result is a system that is not hostile to small businesses so much as blind to them.

The data revolution: the borrower is becoming visible

The good news is that the raw material for seeing these borrowers now exists — and it isn't the credit bureau. It's a new layer of consented, structured, digital data:

• GST filings reveal real turnover, seasonality, and buyer concentration.
• Bank transaction data, shared through the Account Aggregator framework, shows actual cash flow — the truest signal of a small business's health.
• Income-tax filings corroborate declared income.
• UPI and payment histories capture the daily rhythm of a business most financial statements never see.
• Utility payments, e-commerce sales, and supplier-payment behaviour fill in the picture for businesses that have no bureau file at all.

For a thin-file borrower, this isn't supplementary signal. It's the entire signal. The question stops being “what does the bureau say?” and becomes “what does this business's actual cash behaviour tell me?”

The rails: Account Aggregator and ULI

Two pieces of digital public infrastructure are turning that scattered data into something a lender can actually use.

The Account Aggregator (AA) framework lets a borrower consent, in seconds, to share verified financial data directly from the source. Adoption has been remarkable: by the end of 2025, over 250 million users had linked accounts, and by early 2026 more than 2.8 billion financial accounts were enabled to share data across 17 licensed aggregators. That said — and this matters for anyone building on it — only around 38% of borrowers were AA-enabled as of late 2025, meaning the majority still require old-fashioned PDF bank-statement analysis. The rail is real, but it isn't yet universal.

The Unified Lending Interface (ULI), the RBI's “UPI moment for credit,” goes a step further. It gives any lender a single, standardised pipe to a borrower's data — GST, bank statements via AA, land records, and more — through common APIs, eliminating the one-to-one integrations smaller lenders could never afford. It compresses a decision that took weeks into minutes.

But here's the honest part, and it's the part the hype skips. ULI's early numbers are modest: on the order of ₹27,000 crore disbursed across roughly 600,000 loans in its first stretch, of which MSME loans were about ₹14,500 crore. It has widened to a dozen loan journeys, but scale-up is being held back by exactly the frictions you'd expect — poorly digitised land records and slow adoption by the large banks that hold most of the balance sheet.

The lesson: infrastructure solves the plumbing, not the judgment. Rails make the data available. They do not decide who to lend to.

How to actually underwrite the invisible borrower

This is where craft replaces plumbing. A practitioner underwriting a thin-file MSME works from a different playbook than a bureau-driven one:

1. Underwrite the cash flow, not the balance sheet. For an informal business, the bank statement is more honest than the audited accounts. Assess the pattern and stability of inflows, the buffer between inflows and obligations, and the seasonality. A business with modest but reliable cash flow is often a better risk than one with impressive but lumpy books.

2. Treat alternative data as primary, not decorative. GST turnover trends, supplier-payment punctuality, utility regularity, and platform sales are not “nice to have.” For a thin-file borrower they are the credit file. The skill is combining weak-but-independent signals into a calibrated view.

3. Score with calibrated confidence, not false precision. Public and alternative datasets are incomplete and messy. A model that pretends otherwise is more dangerous than no model. The value is in knowing how sure you are — and pricing or sizing the loan accordingly. A confident “small yes” beats a falsely precise “big yes.”

4. Make speed part of underwriting. A six-week decision on a working-capital need isn't a slow yes — it's a no the borrower can't wait for. For micro-tickets, an instant, good-enough decision that you monitor closely is worth more than a perfect one that arrives too late.

5. Shift the risk work from origination to monitoring. The thin-file borrower's story keeps being written after you lend. Continuous cash-flow monitoring and early-warning signals — a sudden drop in GST filings, receivables stretching, a key buyer going quiet — let you manage a portfolio you could never have underwritten to certainty up front. In this segment, early warning is the real underwriting.

The pitfalls that sink lenders here

Every one of these tools has a failure mode. The disciplined lender guards against them:

• Confidently wrong models. Point a model at a borrower it has never really seen and it doesn't say “I don't know” — it interpolates, smoothly, into territory where it has no business having an opinion. Design for humility.
• Models that learn the past and break on the future. A model trained on ten good years has never seen the eleventh. Regime changes are exactly when historical patterns betray you — and exactly when lending decisions matter most.
• Data quality and coverage gaps. With most borrowers still on PDF statements and land records patchy, the pipeline is only as good as its weakest input. Build for messy, partial data, because that's what you'll get.
• Consent and trust. These rails run on borrowers agreeing to share deeply personal financial data. The lenders who win will use that data narrowly and transparently. Trust is the moat.

What the winners will do

Put it together and the shape of the opportunity is clear. The lenders who crack India's MSME market won't be the ones with the most capital or even the best access to data — because once ULI matures, everyone will pull the same GST and bank feeds. Access raises the floor; it doesn't create an edge.

The edge will belong to whoever pairs three things: the rails (AA and ULI, to see the borrower), the model (alternative-data scoring with calibrated confidence, to price the risk), and the judgment (human accountability, early-warning discipline, and trust, to make the call and stand behind it).

For years, the honest answer to “why can't this business get a loan?” was “because we can't see it clearly enough to price the risk.” India has finally handed lenders the lens. Whether that closes the credit gap now depends on something no rail can provide — the quality of the judgment on the other end.

The invisible borrower is becoming visible. The next decade of Indian credit will be built by whoever learns to underwrite them well.

Figures are current as of early–mid 2026 and drawn from RBI, MicroSave, CGAP, and the Department of Financial Services; confirm against primary sources before citing.