Klarna's IPO and the BNPL Unit Economics Problem

Buy Now Pay Later has two different economic tests. The lender needs revenue to cover funding, credit losses and operations. The merchant needs incremental contribution from completed sales to exceed the extra acceptance and servicing costs. Passing one test does not establish that the other works.

Klarna's September 2025 IPO prospectus makes a useful historical case study. The financial figures below describe that filing's periods, not Klarna's current earnings or share valuation. The regulatory section separately reflects developments checked in October 2026.

The IPO Numbers: Keep the Denominators Separate

The prospectus reports $112 billion in GMV for the twelve months ended 30 June 2025 and credit-loss provisions equal to 0.52% of that GMV. It also reports a $21 million net profit for calendar 2024 but a $152 million net loss for the first six months of 2025. Calling the business simply profitable at IPO would obscure that distinction.

Klarna defines transaction margin dollars as revenue less processing/servicing costs, credit-loss provisions and funding costs. This is a non-IFRS measure, not net profit. Its provision measure includes future-loss provisions and realized losses; it is not a count of customers who defaulted.

A merchant negotiating a BNPL contract should use its own quote. A company-wide revenue ratio is not a price list, and interchange is not the whole card MDR. Compare the same country, product, transaction profile, refund treatment and settlement terms. The PSP pricing comparison explains the same normalization problem for card acceptance.

A Reproducible Credit-Loss Sensitivity

Illustration, not a forecast: hold the disclosed $112 billion GMV constant and increase the provision-to-GMV ratio by 0.5 percentage points. The mechanical additional provision is:

$112 billion × 0.005 = $560 million.

This is a pre-tax, all-else-equal expense sensitivity. It does not predict cash losses or establish an eight- or ten-percentage-point operating-margin movement. To calculate a margin change, use the same-period revenue denominator and specify which margin is measured. Credit policy, product duration, funding, growth and recoveries would not necessarily remain fixed in an actual stress event.

For a merchant, this calculation suggests due-diligence questions: what happens to acceptance, reserves, service levels or commercial terms under stress? Request the contractual answer rather than infer it from an equity-market narrative.

Merchant Break-Even: A Hypothetical Example

Suppose a merchant has these inputs. They are invented round numbers for the calculation, not Klarna prices or measured conversion results:

• 1,000 completed orders at an average $100 basket.
• Contribution before payment fees of $30 per order, after product and fulfilment costs.
• Card acceptance at 2%; the proposed BNPL method at 4%.
• No change in basket size, returns, servicing cost or customer mix.

Contribution after the payment fee is $28 per card order and $26 per BNPL order. If all 1,000 existing orders moved to BNPL, the merchant would give up $2,000. Around 77 genuinely additional orders would replace that contribution: $2,000 ÷ $26 = 76.92, or about 7.7% more completed orders than the starting 1,000.

If only 200 existing orders switch, the fee difference is $400. Around 16 incremental BNPL orders would cover that difference under the same assumptions. Migration and incremental sales therefore need separate measurement.

A proper pilot should also capture partial refunds, non-refundable provider fees, cancellations, fraud allocation, support contacts, basket mix and the time to receive usable funds. Higher checkout completion can still leave less contribution if the new orders have lower margins or more returns.

Use a holdout or carefully matched cohort where practical. Record eligible checkout sessions and provider declines as well as completed orders, so underwriting does not disappear from the conversion denominator. Shoppers who already intended to buy are not automatically incremental demand because they selected BNPL.

Test Merchant Circumstances, Not a Universal Vertical Ranking

These are evaluation questions, not reported industry conversion benchmarks. For product differences, use the Klarna, Afterpay and Affirm comparison alongside a current quote and agreement.

Regulation: Three Different Developments

United Kingdom. The FCA began regulating third-party Deferred Payment Credit on 15 July 2026. Its guidance distinguishes credit supplied by a separate lender from credit supplied by the seller itself, and excludes earlier agreements from the new regime. Relevant lenders need authorization or temporary permission. Affordability, information, support and complaints requirements matter to the customer journey; this is no longer an unspecified 2025–2026 forecast.

Australia. ASIC's licensing guidance sets 10 June 2025 as the relevant start for regulated BNPL credit activities and explains permissions and transitional arrangements. The precise permission depends on the activity. A merchant should confirm its provider's status and contractual responsibilities rather than assume every business offering a BNPL button is itself the lender.

United States. The CFPB records that it withdrew its 2024 BNPL interpretive rule on 12 May 2025. That withdrawn interpretation is not a current blanket statement of BNPL card-issuer obligations. Its withdrawal also does not establish that every BNPL arrangement is exempt from other applicable law.

Ask which product and contracting entity serves each market, how regulation affects eligibility and checkout, and who handles complaints and refunds. There is no single global BNPL compliance-cost percentage to insert into a model.

The AI Narrative: Attribution Is Not Causation

In its IPO prospectus, Klarna said its assistant handled 69% of customer-service chats in the twelve months ended 30 June 2025 and attributed approximately $39 million in 2024 savings to it. These are company-reported operational claims, not an independent estimate of AI's effect on total profitability.

Headcount reductions alone cannot isolate automation from hiring policy, restructuring, outsourcing or changing transaction mix. A support metric also does not prove better credit underwriting. Ask for resolved cases, escalations, repeat contacts and customer outcomes alongside cost per case.

For merchants, the actionable test is contribution after acceptance, fulfilment, returns and support. For readers studying the provider, distinguish historical reported profit, transaction contribution and hypothetical credit-loss stress. Those are related questions, but they are not interchangeable.