Lending is a prediction

Seeing Someone Differently — Field Notes / K Baksh
kameel baksh

Seeing Someone
Differently.

What it makes possible and what it costs.

Scroll to explore ↓

What sits beneath the decision?

At its simplest, lending is a prediction about someone’s future.

A lender gives someone money today because it believes that person can and will repay it tomorrow. The price of being wrong is built into the business.

I began looking at Propel Holdings because of its third-quarter results. Revenue reached US$201.5 million in Q3 2026, up 32% from a year earlier, and the company announced its thirteenth consecutive quarterly dividend increase. But the numbers led me somewhere else: to the meaning of a lending decision.

Propel operates in the part of the market where traditional institutions often decline to lend. Its brands and partner programmes serve consumers in the United States and United Kingdom whom it describes as underserved.

But underserved is not one type of customer. Someone can be overlooked because a lender has too little information. Someone else may be unable to afford the loan. The first could represent a failure of the system to see; the second may be a necessary refusal.

A better decision isn't necessarily a yes. It is knowing when yes creates value for both sides.

That distinction changes how I think about the size of the opportunity. The total addressable market cannot simply be everyone who has been declined by a bank. It is the much narrower group who need suitable credit, can be reached economically, meet the rules, and can repay on sustainable terms.

It also changes what I think Propel might be building.

A lender needs to find customers, evaluate them, secure funding, issue or facilitate credit, collect repayments and manage losses. Each activity has its own economics. Propel increasingly has more than one way to put those capabilities to work: through consumer-facing lending programmes and through lending-as-a-service relationships.

In Q3, lending-as-a-service revenue reached US$13.5 million, up 170% year over year. UK revenue reached US$22.0 million, up 69%. Those figures describe different dimensions of the business and should not be added together as exclusive segments.

For a growth practitioner, that is intriguing. A system that works across products, partners and geographies might become more useful with every application and repayment outcome. But only if what it learns genuinely improves decisions—and if its distribution brings the right customers, not merely more of them.

The quarter offers clues, not proof. Reported net income was US$18.4 million, up 23%, slower than revenue. Provisions for loan losses and other liabilities represented approximately 53% of revenue. Propel also reported a lower cost of debt and improving acquisition efficiency.

All of those things can be true at once: the business is growing, some economics are improving, and a large part of its revenue must still absorb the cost of credit risk.

Then there is AI. Propel already uses AI-enabled decisioning. But what happens if better models become cheap and widely accessible? The ability to predict risk may improve across the industry, while the advantage of owning a model becomes less unusual. What could endure is harder to see: trustworthy repayment data, good judgment, compliant operations, funding relationships and the ability to serve people responsibly through a difficult cycle.

There is a human question here, too. Better approval rates tell us little by themselves about whether borrowers are better off after receiving the credit. That outcome belongs in the definition of a successful lending system.

I don't know yet whether Propel's strongest advantage will be its models, its distribution, its partnerships or the way these pieces work together. I do know that looking beneath a strong quarter has made the question more interesting.

Perhaps the most consequential innovation in lending isn't approving more people. Perhaps it is making better decisions about whom to approve—and why.

What I'm still trying to understand.

The numbers establish momentum. They don't settle its durability.

What happens after the loan is made?

How do repayment and customer outcomes differ by origination vintage, product, geography and first-time versus returning borrower?

What is really driving the yield?

How much comes from pricing, product mix and fees rather than improved selection of credit risk?

What is the true cost of finding the right customer?

Will acquisition efficiency persist when credit conditions tighten, and which channels attract borrowers with sustainable repayment outcomes?

Which part of the system compounds?

Can lending-as-a-service grow without comparable exposure, and what would remain distinctive if competitors gained similar AI tools?

Who benefits from the decision?

What evidence would show that borrowers are better off, not simply more likely to be approved?

Primary evidence: Propel Holdings Q3 2026 earnings release, November 3, 2026. USD unless stated. Company-defined and non-IFRS measures require the accompanying MD&A. Questions and future scenarios are the author's interpretations, not management guidance. Working editorial draft; not investment advice.

K Baksh
↑ Back to the opening