Underwriting the Invisible Borrower

India has a ₹25–30 lakh crore MSME credit gap.

After 20 years inside credit risk, I'm convinced we've been calling it the wrong name.

It's not a capital gap. Banks and NBFCs have money to lend. It's a risk-assessment gap — we don't know how to underwrite the businesses that need the money most.

Here's the uncomfortable math. India has 6.19 crore registered MSMEs. Over 99% of them are micro enterprises. Credit penetration in the sector sits around 14% — against 37% in China and 50% in the US.

Why the gap? Because the lending playbook we imported assumes a borrower who doesn't exist here.

I spent 15 years building credit and trade-data software for Fortune 500 lenders in the US — at eCredit and then Cortera. That world runs on a deep, clean credit bureau, audited financials, and decades of trade-line history. The model works because the data is already there.

Now point that same model at a kirana supplier in a tier-3 town, a small fabricator, a two-person D2C brand. The financials are informal. The bureau file is thin or empty. So the model returns the only honest answer it can: “insufficient data” — which the system reads as “decline.”

The borrower isn't risky. The borrower is invisible.

That's the problem I've spent the last few years building for: an instant B2B credit rating and portfolio-analysis platform for Indian MSMEs, built on public and alternative data rather than a credit file that was never going to exist.

A few things I've learned the hard way:

1. Alternative data isn't a nice-to-have — it's the only data. GST filings, electricity usage, supplier payment behaviour, banking transaction patterns. For a thin-file borrower, this isn't supplementary signal. It's the entire signal.

2. Be honest about the limits. Public and alternative datasets are incomplete and messy. A model that pretends otherwise is more dangerous than no model. The value is in calibrated confidence, not false precision.

3. Speed is underwriting. A 6-week credit decision isn't a slow “yes” — it's a “no” the borrower can't wait for. Instant, good-enough scoring beats perfect-but-too-late every time.

The lenders who crack the Indian MSME market won't be the ones with the most capital. They'll be the ones who solve the data problem — who get comfortable underwriting the invisible borrower.

That's where the next decade of credit innovation in India gets built.

What's the most useful alternative-data signal you've seen for thin-file lending? I'd genuinely like to compare notes.