A buy-now-pay-later button is one of the easiest things to add to a checkout and one of the hardest things to evaluate honestly. The pitch is clean. Turn it on, customers split the cost into four payments, conversion goes up, average order value goes up, and the provider takes a cut you barely notice. Klarna, Afterpay, and Affirm all publish numbers that say exactly this.
The problem is that the pitch describes the best case and bills you for every case. A BNPL provider charges a merchant fee on every transaction it touches, including the ones from customers who were always going to buy and would have reached for a card the moment BNPL was not there. Those sales are not incremental, but the provider still gets paid. For some merchants that fee buys real, additional revenue. For others it is a discount handed to existing customers for nothing. This post is about telling the two situations apart before you sign.
Where buy-now-pay-later came from
Paying for goods in installments is old. US retailers ran layaway and store credit for most of the twentieth century, and furniture and appliance dealers built whole business models on financing. What changed in the 2010s was the checkout. Klarna started in Sweden in 2005, Affirm was founded in the United States in 2012, and Afterpay launched in Australia in 2014. The shared idea was to move installment lending out of a separate credit application and into the buy button itself. No long form, no hard credit pull for the basic product, just a choice at the moment of payment to pay in four.
The modern default is the pay-in-four structure: 25 percent down at checkout and three more equal payments every two weeks, with no interest if you pay on time. The provider earns most of its merchant revenue from the fee charged to the store, not from the shopper. Longer-term financing, the kind Affirm built for higher-ticket purchases, does charge the consumer interest and stretches over months. Both sit under the same BNPL label, which is part of why the category is confusing. They are different products with different economics.
Growth has been steep. Worldpay's Global Payments Report tracked BNPL online spending rising from 2.2 billion dollars in 2014 to 342 billion dollars in 2024, and a widely cited industry forecast put the global BNPL market near 560 billion dollars for 2025. The pandemic accelerated it, younger shoppers adopted it fastest, and by the mid-2020s it had moved from a novelty to a checkout option many customers expect to see.
How it works, and what the merchant actually pays
Mechanically the deal is simple. A customer picks BNPL at checkout. The provider pays the merchant the full order value minus its fee, usually within a day or two, then owns the customer relationship and collects the installments. If the customer never pays, that is the provider's loss, not the merchant's. The store has its money.
That fee is the whole question. BNPL is more expensive than a card. A typical credit card transaction costs a merchant roughly 2 to 3 percent. BNPL providers generally charge more, with published and reported rates landing roughly between 3 and 8 percent of the order, plus a fixed amount per transaction often around 30 cents. The exact rate depends on the provider, the product, the merchant's size, and how hard the merchant negotiated, and a small store on standard terms can easily pay 5 to 6 percent. So the first hard fact: every order routed through BNPL costs a few points of margin more than the same order on a card.
You do get real things for that fee beyond the financing. Because the provider is the lender, it absorbs payment defaults entirely, and many providers also offer fraud chargeback protection that can reduce a merchant's exposure compared with cards. That protection has limits. If a customer disputes a purchase because the item never arrived or was not as described, the merchant is generally still on the hook to prove delivery or refund the provider. BNPL shifts credit risk cleanly. It does not make a merchant's fulfillment problems disappear.
Where the conversion and AOV lift is real
The optimistic numbers are not fiction. They are just conditional, and the conditions matter.
The single most useful piece of evidence is a test Stripe ran across more than 150,000 global checkout sessions, showing some shoppers a BNPL option and some not. For sessions where BNPL was eligible, businesses saw up to a 14 percent increase in revenue from the combination of higher conversion and higher order values. The detail that matters most for the incrementality question: Stripe reported that more than two-thirds of BNPL volume came from net-new sales rather than customers switching off cards. The conversion lift was largest on the biggest transactions, between 500 and 1,500 dollars, though Stripe saw conversion rise across the full order-value spectrum. BNPL helped most where a financing option should help, on purchases large enough that splitting the cost changes whether someone can say yes.
Academic work points the same direction. A 2024 NBER working paper by Tobias Berg and co-authors, studying BNPL from the merchant's side, estimated that offering it increased sales by around 20 percent, with the effect concentrated among lower-creditworthiness customers and products where the merchant held more pricing power. Their framing is sharp: BNPL works partly as price discrimination, a softer effective price for budget-constrained shoppers without cutting the sticker price for everyone. The lift comes from reaching buyers on the fence about affordability.
Put the evidence together and the merchant who benefits has a specific profile: higher average order values, since splitting the payment only matters when the payment is large; categories where the purchase is considered and sometimes deferred, such as furniture, mattresses, electronics, appliances, premium fashion, and travel; and a customer base skewed younger or thinner on credit. For those merchants, the BNPL fee is buying sales that would not have closed.
The vendor-reported numbers are louder than this and should be read as marketing. Affirm has cited internal figures of merchants seeing average order values 85 percent higher on Affirm transactions, and Shop Pay Installments has been promoted with claims of up to 50 percent higher average order value. Both are vendor-reported and self-selected. A customer who chooses to finance a purchase was probably always going to spend more than one paying in full, so a BNPL versus non-BNPL order-value comparison measures who chose the option as much as what the option did. Stripe's 14 percent revenue lift, from a real holdout test, is a more trustworthy ceiling than any side-by-side average.
Where it just adds a fee
Now the other half, the part the provider's deck does not feature.
If a customer was always going to complete the purchase, BNPL adds cost and nothing else. The clearest case is a low-ticket order. On a 35-dollar purchase, splitting into four payments of roughly nine dollars changes very little about whether someone buys. The friction was never the price. But if that shopper taps the BNPL button anyway, the merchant pays the BNPL fee on a sale a debit card would have closed far more cheaply. The CFPB's market report found the average pay-in-four BNPL loan was about 135 dollars in 2023, so a large share of BNPL volume sits at order sizes where the financing does no real conversion work.
The second case is your existing, loyal customers. A repeat buyer who already trusts you and already intended to order does not need a financing nudge. If your BNPL placement is prominent enough, those customers will use it, and every time they do you convert a card sale into a more expensive BNPL sale. The provider counts that transaction in the volume it reports back as a win. From your margin's point of view it is pure leakage.
This is the incrementality question, and it is the entire decision. Incremental revenue is the sales that happened because BNPL existed and would not have happened otherwise. Substituted revenue is the sales that would have happened anyway, rerouted through a pricier rail. The provider's fee falls on both, and its reported lift numbers tend to count both. A merchant who sees a big total BNPL number and concludes BNPL is working has not measured the thing that determines whether it is.
The way to find out is a holdout test, the method Stripe used. Show BNPL to a random portion of traffic and withhold it from a comparable portion, then compare total revenue and total conversion between the groups, not BNPL revenue in isolation. If the group that could see BNPL generated more total sales, the lift is real and you can size it. If overall conversion barely moved while BNPL volume was high, BNPL is mostly cannibalizing your cheaper payment methods and the fee is a cost with no return. Few merchants run this test. It is the single most valuable thing in this post.
One genuine reason to offer BNPL even when its direct lift is modest: customers abandon checkouts when they cannot pay the way they want. Baymard Institute's checkout research finds that too few payment methods is among the reasons shoppers abandon carts. If a real segment of your audience expects BNPL and bounces without it, the option earns its place as table stakes rather than as a growth lever. That is a legitimate argument, and a different one from the conversion-lift pitch. It should rest on your own abandonment data, not an assumption.
The regulatory and consumer-debt backdrop
BNPL did not grow up under the rules that govern credit cards, and that gap is now closing unevenly. It affects merchants more than it first appears.
The consumer picture explains the scrutiny. The CFPB found that most pay-in-four BNPL loans went to borrowers with subprime or deep subprime credit scores, and that around 63 percent of BNPL users held more than one such loan at the same time. Default rates have run lower than on credit cards, roughly 2 percent of loans against about 10 percent for cards in the CFPB's window, but the simultaneous-loans pattern is the worry. Because most BNPL lending has not been reported to credit bureaus, one provider often cannot see what a borrower already owes another. Critics call the result phantom debt: real obligations invisible to the system meant to track them.
Regulators have moved at different speeds. In the United States the CFPB issued an interpretive rule in 2024 treating pay-in-four BNPL reached through a digital account as a form of credit card lending, which would have carried dispute and refund rights, then withdrew that rule in May 2025 under a new administration. The legal question is not fully settled, but federal pressure has eased for now. The United Kingdom is going the other way: the FCA has confirmed BNPL will come under formal regulation from July 2026, bringing affordability checks and complaint rights. Australia brought most BNPL under its consumer credit law in 2025. A merchant selling across borders should expect BNPL to look more like regulated lending over time, not less.
The credit-reporting picture is shifting fastest, and it changes the consumer's calculation. Affirm began reporting its pay-in-four loans to the major credit bureaus through 2025, and FICO announced score versions that incorporate BNPL data. As BNPL starts to affect credit scores, the format loses some of its consequence-free feel, which over time may cool casual use. None of this is a reason to avoid BNPL, only a reason to treat it as a financial product with a moving regulatory floor.
The consumer-debt context also carries reputational exposure. BNPL skews toward financially stretched, lower-credit, often younger shoppers, and a meaningful share juggle several loans at once. A merchant offering it is, in a small way, part of that system. That is not a verdict against BNPL, since plenty of customers use it sensibly. But it is a reason to present it as one payment choice among several rather than the default, and to weigh brand fit first.
A decision frame
Strip away the marketing and the choice comes down to a few honest questions.
Start with order value and category. If your average order is well into the hundreds of dollars, and your products are considered purchases that customers genuinely defer, BNPL has real room to convert buyers that price was blocking. If your average order is small and your category is routine, the financing has little conversion work to do and the fee is likely to outrun the benefit.
Then look at your customers. A younger or thinner-credit audience is more likely to need and want BNPL, which tilts the lift toward incremental. Established repeat buyers with their own credit lines are more likely to use BNPL as a card substitute, which tilts it toward leakage.
Then do the arithmetic on the fee. If you pay 5 percent to a BNPL provider against 2.5 percent on a card, the extra 2.5 points is the price of admission. For it to pay off, the incremental sales BNPL brings must cover that surcharge across all your BNPL volume, including the substituted sales you are now overpaying on. On a thin-margin product the bar is high.
Then, if you can, run the holdout test. It is the only way to replace the provider's number with your own. A merchant who has run it and seen real incremental lift should offer BNPL with confidence and negotiate the rate hard. A merchant who has not is guessing, and the provider is happy to let them guess, because the fee arrives either way.
Buy-now-pay-later is neither a growth hack nor a trap. It is a financing product with a real merchant fee, a real conversion effect in the right conditions, and a real tendency to charge you for sales you already had. The merchants who win with it know which of their sales are incremental. The merchants who lose never asked.
Council summary
This post argues that BNPL has a real but conditional conversion effect, and that its merchant fee falls on incremental and substituted sales alike, so the only honest test is a holdout experiment. Review confirmed the load-bearing figures against primary sources: Stripe's 150,000-session test and 14 percent revenue lift, the CFPB's 135 dollar average pay-in-four loan and its subprime and simultaneous-loan findings, the NBER merchant-side estimate of roughly 20 percent higher sales, and the US, UK, and Australian regulatory timeline. The Stripe passage was corrected to note conversion rose across the full order-value range rather than only on large baskets, and the Affirm and Shop Pay Installments AOV claims are kept but clearly framed as vendor-reported. The takeaway: BNPL pays off when order values are high and the audience skews younger or thinner on credit, so run the holdout test before trusting any provider's number.
Comments