Building a bank for people who’ve never been to one.

Credit decisioning

I designed a credit model for retailers with little formal credit history. We weighted details like how many shutters a shop had, how many suppliers visited each day, and what paperwork the retailer had with their landlord. These weren’t guarantees a bank would normally accept, but in our experience they were useful predictors of repayment risk. We used them to place retailers into credit tiers and assess the likelihood of a loan becoming non-performing (NPL).

A week to sell the stock

BNPL sounds fancy. For a retailer, the idea was simple: get the goods for a week, sell them, and pay for them afterwards. We explained it using the idea of renting, which the market understood much more easily than invoice financing.

The retailer had less cash tied up in stock. The supplier got paid by the bank the same day. The bank and investors earned their share from financing the purchase.

With everyday goods like rice and lentils, we could ask a practical question: would this shop sell that sack by next week? That was something we could assess from how the shop actually traded.

Invoice discounting

Paying a supplier three days earlier gives them three more days to put that money back into the business. They can buy the next batch of stock and sell it, instead of waiting for a delivery person to bring the payment back.

For a supplier who turns stock over quickly, getting those days back repeatedly can mean as many as 50 extra buying-and-selling cycles over a year. It depends on their cash cycle, but the value is straightforward: the same money gets more chances to earn. That’s what I was designing the early-payment workflows to make possible.