How to actually choose a payment provider in 2026

Every payments RFP in 2026 starts the same way. A spreadsheet lands in someone's inbox with twelve providers across the top, forty rows down the side, and a weighted scoring model at the bottom that confidently produces a winner to two decimal places. Three months later the chosen provider is live, the authorisation rate is two points below what the slide deck promised, and the merchant is quietly running a second processor on the side. The scoring model was not wrong. It was answering the wrong questions.

Pricing pages lie — not because anyone is being dishonest, but because the number on the page is the smallest variable in your total cost of acceptance. A 10 basis point difference in headline rate is worth a rounding error compared to a two-point swing in authorisation, a 48-hour delay in settlement, or a chargeback workflow that burns a full-time analyst. The providers that win on the pricing page often lose on everything else, and the merchants that optimise for the pricing page find out twelve months later.

Start with authorisation, because nothing else matters if the money does not move. Ask every shortlisted provider for their authorisation rate on your exact country mix, your exact card-brand mix, and your exact transaction profile — not their global average, not a reference customer in a different vertical, not a number from last year's annual report. A serious provider will produce this in a week from a sandbox replay of your traffic. A provider that cannot, or will not, is telling you something important about how they will behave once you are a customer.

Then ask about local acquiring. A transaction acquired domestically in the cardholder's country authorises at a materially higher rate than the same transaction acquired cross-border, settles at lower interchange, and avoids the cross-border fee that schemes apply on top. If your traffic is genuinely global, the right answer is rarely a single acquirer; it is an orchestration layer that routes each transaction to a local acquirer in the cardholder's market. Providers will tell you they have "local acquiring in 40 countries." Ask which 40, ask which BINs are actually routed locally today, and ask what happens to a Brazilian card on a Tuesday at 2am when the local acquirer's link is down.

Chargebacks are the second place pricing pages mislead. Every provider quotes a per-chargeback fee. Almost none of them quote the operational cost of the workflow around it. Ask to see the actual dispute console. Ask how evidence is assembled — is it a templated PDF the merchant uploads manually, or does the provider pull shipping data, 3DS results, device fingerprints and previous order history automatically? Ask what the provider's representment win rate is on disputes for merchants in your category, and ask whether they will share the underlying data or only the headline number. A provider with a real chargeback product will answer in detail. A provider without one will pivot to talking about fraud scoring.

The pricing page tells you what a transaction costs when everything works. The interesting question is what it costs when something breaks.

Settlement and treasury are the third blind spot. T+2 used to be the default; in 2026 it is a choice. Ask what the standard settlement cycle is for your currencies, what the cost is to compress it, and whether the provider can settle in multiple currencies into multiple accounts without forcing an FX conversion. If you sell in 14 currencies and the provider only settles in three, you are paying their FX desk every day for the privilege of being a customer. Ask about weekend and holiday settlement, ask about the cut-off times that determine which day a transaction lands, and ask whether the provider can settle into a virtual account structure that your finance team can actually reconcile.

Then ask the question that quietly decides everything: who owns the relationship when something breaks at three in the morning. Not the named account manager on the contract — the actual human who picks up the phone, has the authority to escalate to engineering, and can get a payment route re-enabled before the European day starts. Providers with a real merchant-success function will name that person, describe the escalation path, and commit to response times in writing. Providers without one will offer a ticketing portal and a four-hour SLA that begins when a junior agent acknowledges the ticket.

Integration quality is the fifth axis, and the one most easily tested before you sign. Spend an afternoon in the provider's sandbox with a real engineer, not a sales engineer. Build a working checkout, a working refund, a working partial capture, a working 3DS step-up, and a working webhook listener. Count the number of times you had to email support, the number of fields in the API that are documented as "optional" but actually required, and the number of edge cases that are only mentioned in a PDF buried three clicks deep. The sandbox experience is a leading indicator of the production experience, and it costs nothing to run.

Finally, look at the boring things. What is the provider's PCI scope and how does it interact with your own? What is their uptime over the last 24 months, broken down by region and by service, with the underlying status-page data rather than a marketing number? Who are their banking partners, and what happens to your money if one of those partners exits the relationship? What is the contractual exit path — can you take your tokenised card vault with you, in a portable format, without paying an exit fee that makes leaving economically irrational? The answers to these questions rarely change which provider you pick, but they often change the contract you sign.

The shortlist that survives this process is usually shorter than the one that started it, and the winner is rarely the cheapest. That is the point. A payment provider is not a commodity line item; it is a piece of infrastructure that touches every customer, every refund, every dispute, and every reconciliation your finance team will run for the next five years. Choose it the way you would choose a core banking partner, not the way you would choose a SaaS tool — because that is what it is.