Tala: How to Lend to People the Banks Can't See

Billions of people around the world can't get a loan — not because they'd never repay it, but because no bank has any way to know whether they would. No credit history, no formal payslip, no collateral. To a traditional lender, they're invisible. Tala's entire business is built on a single, powerful reframe of that problem, and it's a great one to reverse-engineer.

The insight: invisible isn't the same as risky

The conventional view treats "no credit history" as "too risky to lend to." Tala's founder, Shivani Siroya, started from a different premise after interviewing hundreds of people in emerging markets: these borrowers weren't uncreditworthy, they were unmeasured. Small business owners, market traders, people with real, regular cash flow — just none of it captured in the formats a bank understands.

Once you believe that, the problem changes shape completely. It's no longer "how do we lend to risky people?" It's "how do we measure creditworthiness for people the old system can't see?" That's a solvable engineering question rather than an unsolvable risk one.

The winning move: the phone is the credit file

Here's the part that makes the whole thing work. When a customer downloads the Tala app and opts in, the app looks at the data on their smartphone — with permission — to build a risk profile. The patterns turn out to be startlingly predictive. Do you call the same people consistently? Is your contact list orderly? Do you pay your bills on a regular cadence? Thousands of these small behavioural signals, taken together, predict repayment.

The smartphone, in other words, became the credit file for people who never had one. That's the core innovation: not the lending itself, but the alternative data model that made lending to the "invisible" possible at all. Traditional lenders had the capital; they didn't have the measurement. Tala built the measurement.

Distribution was baked in

Notice how neatly the model rides an existing wave. Tala didn't have to convince people to buy a device or learn a new behaviour — it rode the smartphone adoption already sweeping emerging markets. The same object that put the internet in someone's hand also carried the data that made them legible to a lender. The distribution channel and the underwriting engine were the same device.

That's a quiet lesson in timing. The idea couldn't have worked a decade earlier; the smartphones weren't there. Riding an inevitable trend, rather than fighting to create one, is often what separates a business that scales from one that grinds.

The flywheel: every repayment sharpens the model

Then there's the compounding advantage. Every loan Tala issues and every repayment it collects feeds back into its risk model, making the next prediction better. More borrowers means more data; more data means sharper underwriting; sharper underwriting means better loss rates and the ability to serve more borrowers profitably. The data moat widens with every cycle — and a new entrant starts from zero.

This is the kind of advantage that doesn't show up in a feature list but decides who wins over years. The product you can see is a loan app. The product that actually compounds is the model underneath it.

What I take from it

Reverse-engineered, Tala's success is a stack of good decisions on top of one great reframe. It saw that "invisible" borrowers were mismeasured, not unworthy. It found the credit signal hiding in a device those borrowers already owned. It rode smartphone adoption instead of fighting for distribution. And it built a data flywheel that gets stronger every day it operates.

The lesson I take: the biggest opportunities often hide behind a problem everyone has quietly accepted as unsolvable. "You can't lend to people with no credit history" was treated as a law of nature. It was really just a measurement gap — and measurement gaps are exactly the kind of thing a well-aimed startup can close.