Lending to small and medium businesses should be a great business. India has over 60 million MSMEs, most under-banked, hungry for working capital. Yet traditional lenders have historically either avoided the segment or priced it so high that formal credit never reached it. The problem was never that small businesses are inherently bad borrowers — it's that assessing them was expensive, slow, and unreliable. Understanding why explains why this decade's data infrastructure is changing everything.

The Four Structural Problems

1. Opacity

A salaried borrower's income is a payslip; an MSME's income is whatever its books say. Historically, those books were unaudited, partly cash-based, and optimised for tax minimisation rather than credit presentation. Declared profit and actual cash flow could differ wildly — in both directions.

2. Heterogeneity

Consumer lending scales because 100 million borrowers are statistically similar. SMBs are not: a kirana store, a SaaS startup, a textile trader, and a job-work machine shop have completely different cash-flow shapes, cycles, and risk drivers. One scorecard fits none of them, but building sector-specific models for each niche was uneconomical at branch-based underwriting costs.

3. Cost-to-serve

Traditional underwriting meant document collection, field verification, financial statement analysis, and committee approvals — days of human effort per file. On a ₹5 lakh loan, that process can consume more in cost than the interest earns. Small loans were economically impossible to underwrite the old way.

4. Thin formal footprints

Commercial bureau files existed only for businesses that had already borrowed. First-time borrowers — the majority of the segment — were invisible by definition, trapping them out of the system that would have proven their reliability.

The Consequences Everyone Felt

Lenders responded rationally: demand collateral, personal guarantees, minimum vintage criteria, or simply decline. MSMEs responded equally rationally: they turned to informal credit at punishing rates, delayed supplier payments, or grew slower than they could have. A structural credit gap of tens of billions of dollars persisted not from lack of willingness on either side, but from lack of information.

What Changed: India's Data Stack

Over the past several years, four public/regulated systems quietly solved the opacity problem:

  • GSTN: monthly, invoice-level revenue declarations, cross-verifiable through buyers' input tax credit claims.
  • MCA: entity identity, directors, charges, and compliance history for every registered company.
  • Account Aggregator framework (operated under RBI regulation with Sahamati as the ecosystem's non-profit facilitator): consented, standardised, machine-readable bank statements — no forged PDFs, no branch visits.
  • Digital bureaus: CIBIL and Experian commercial files deepening as formal borrowing spreads.

Together these turn what used to be a two-week document chase into a few API calls. And crucially, the data is behavioural — it shows what a business does, not what it declares.

How Modern Underwriting Actually Works Now

  1. Identity first: GSTIN/CIN-based KYB verifies the entity in seconds.
  2. Cash flow scoring: 12 months of AA-pulled bank statements parsed algorithmically — inflow trends, bounce rates, fixed commitments, counterparty concentration.
  3. Revenue corroboration: GST filings cross-checked against bank credits; divergence flags risk or fraud.
  4. Segment-aware models: because data is cheap, lenders can build separate models per sector instead of one blunt average.
  5. Continuous monitoring: the same rails allow post-disbursement tracking — early-warning signals replace annual reviews.

    6. The economics invert: cost-per-assessment drops from thousands of rupees to single digits, making ₹2 lakh loans viable where only ₹20 lakh loans used to be. Platforms like KredFlow apply exactly this stack to vendor financing — underwriting a buyer against its real transaction data in minutes so a SaaS contract can be paid upfront to the vendor without anyone waiting weeks for approval.

    What Still Remains Hard

    Honesty requires noting the residual gaps:

    • Very young businesses still lack history depth; models must lean on promoter profiles and transaction velocity.
    • Data quality varies: proprietorship boundaries blur business and household finances.
    • Model risk is real: behavioural data can misread deliberate manipulation, which is why fraud analytics run alongside credit scoring.
    • Consent friction: AA flows depend on borrowers granting access — education still matters.

    The Takeaway

    SMB credit assessment was hard because small businesses were illegible, not because they were unbankable. India built the reading apparatus — GSTN, MCA, Account Aggregator, deepening bureaus — and the segment is opening accordingly. For business owners, the practical move is to keep your digital trail clean and consistent; legibility is now literally the price of admission to affordable capital.