Deep Dive | The BNPL Pivot

Do you want to level up your FinTech market knowledge?

A little different from the Monday jobs drop. The weekly list helps you find new opportunities; pieces like this aim to help you nail the interview and perform once you're in the seat. Let me know (reply or comment) whether you find these macro breakdowns useful!

When Affirm reported its fiscal third quarter in May, interest-bearing loans were 71% of transaction volume, up 33% year-over-year. The product Affirm built its brand on, the 0% APR installment paid for by the merchant, is now a minority of what it does.

This is not a one-company story. The entire Buy Now, Pay Later (BNPL) category has run the same play: what began as a merchant-subsidized checkout button has become a consumer-funded lending product charging interest, late fees, and increasingly a monthly subscription. The economic engine no longer runs on merchants but on the consumer, and that single shift rewrites the unit economics, revenue mix, risk profile, and regulatory exposure of the whole category.

In phase one, BNPL was a B2B2C product wearing consumer clothes. Klarna, Afterpay, and Affirm solved a real merchant problem: shoppers abandoned carts at the final step, and "Pay in 4" closed them. Merchants paid 3% to 6% per transaction (a fee multiples higher than card interchange) and booked it as customer acquisition, not payment processing.

For the shopper, the math was frictionless: a $200 jacket became four $50 payments, no interest, no application, no compounding, and for a car repair or broken fridge, genuine cash-flow smoothing. The "shadow credit card" critique missed the economic shape: the merchant was paying for the credit in exchange for the conversion.

Fractionalizing $200 into $50 trains the shopper to register the $50, and when the purchase is a consumable rather than a necessity, that mental-accounting trap is the feature. What's new isn't the critique; it's that the merchant has stopped paying for it.

Two things broke the original model. Capital stopped being free, raising the cost of underwriting every loan. And the merchant-acquisition flywheel hit a ceiling because the largest checkouts were already onboarded. So the category took BNPL off the merchant rail and onto the open card network.

That's what the physical card is. The Affirm Card splits any purchase, anywhere Visa is accepted, into a 0% short-term plan or an interest-bearing monthly loan; Klarna's card does the same. This deliberately severs the merchant subsidy. Swipe a BNPL card at a coffee shop with no Affirm partnership and the merchant pays Visa interchange (roughly 1% to 1.5% for a small-issuer card like the Affirm Card (issued by Evolve Bank & Trust)), not 6% to Affirm, who keeps only a fraction after the issuing bank's cut.

That doesn't cover the cost of capital, underwriting, or default, so now the revenue comes from the consumer. Anyone who's built a take-rate model here knows it doesn't close without consumer fees. Affirm's filings make the trajectory plain:

Klarna's F-1 makes the reweighting especially legible. Merchant fees fell from 75% of revenue in 2020 to 57% in 2024, while interest income rose to 22% and late fees to 9%. Nearly a third of revenue now comes directly from the consumer, and the slope isn't flattening.

Sezzle is another example case. In 2025, total revenue equaled 11.4% of GMV (a take rate arithmetically impossible to extract from merchants alone). Its merchant processing fees actually fell in absolute dollars, from $17.8M in Q3 2023 to $14.8M in Q3 2024, even as total revenue grew 71%. By Q3 2025, consumer fees overtook merchant and partner income outright: $33.5M versus $26.0M. The merchant rail is shrinking while the consumer rail grows.

The card doesn't just change who pays, but it is changing what gets financed. BNPL began as a tool for the occasional big-ticket or emergency purchase; the anywhere-accepted card migrates it to daily spend, the $4 latte rolled into a $20 monthly minimum.

That changes the credit risk. A book of discrete, re-underwritten big-ticket loans is unlikely to behave like revolving daily-consumption balances, and the damage won't surface in loss curves until volume has compounded.

The most credible defense: BNPL, even at 36% APR, is a structurally better product than revolving card debt. The math holds. A 36% simple-interest loan on a $1,000 purchase over 12 months runs ~$200 in finance charges; the same balance at minimum payment on a card compounds for a decade and costs $1,500 or more. Fixed terms beat revolving, per-transaction re-underwriting beats static limits, and there are no compounding or over-limit fees. The CFPB has conceded most of this.

But the defense holds at the loan level and breaks at the portfolio level. A consumer with eight open BNPL loans across three providers has an undisclosed revolving line, and because most BNPL loans don't report to bureaus, the next lender can't see it. The per-loan virtue becomes systemic risk. Stacking is the part “BNPL as consumer credit” defenders don't have a clean answer for.

Sharper still: the cheaper-than-a-card framing assumes the user would otherwise have used a card. The more honest question is whether BNPL is displacing card debt or expanding total credit to people cards wouldn't approve. If it's mostly the latter, the comparison isn't BNPL versus credit cards, it's BNPL versus no debt at all.

The category spent five years acquiring tens of millions of users on a 0%-interest, merchant-paid product, building brand permission legacy lenders couldn't buy, then used it to issue cards and route those users into loans at up to 36% APR. And it is working at current spreads. The open question is what happens to credit losses on the card-issued, anywhere-accepted, consumer-paid book through the first real recession, and whether a regulatory framework that still treats BNPL as a checkout convenience catches up to what it has become.

If you're interviewing at a company with BNPL products. Know the revenue mix. Explain why the card changes the historical merchant subsidy model, why interchange (1–1.5%) can't cover cost of capital + underwriting + default, and why that forces consumer monetization.

If you work in growth or strategy. The lesson is sequencing: using a subsidized, zero-friction wedge to build brand permission and a user base, then monetizing those same users through another product.

If you work in risk or lending. Per-transaction re-underwriting looks conservative at the loan level and hides aggregate leverage at the portfolio level. Pressure-test whether your book is displacing card debt or extending credit to thin-file/declined consumers.

If you work elsewhere in fintech. The same pattern of launching a loss-leading offering and then increasing user monetization with other products is running across cards, neobanks, and embedded finance.

Sources & Notes