The Four-Party Model: Who Actually Touches a Card Payment

Strip away the plastic and a card payment is a trust problem. A merchant is asked to hand over goods to a stranger, right now, on the promise that a bank somewhere will make them whole. The stranger's bank has never heard of the merchant. The merchant's bank has never heard of the stranger. Neither bank will talk to the other directly for something this small.

Everything in card payments exists to make that promise cheap, fast and enforceable at planetary scale: billions of transactions between parties with no relationship, most of them completing in about a second.

Key idea: the card system does not move money when you tap. It moves a promise, instantly, and settles the money later in bulk. Keeping those two things separate is the single most useful idea in this course.

The machinery that carries the promise has a standard shape, called the four-party model, and once you can name the parties, most of the industry's jargon resolves into who is talking to whom.

The model names four roles. Two of them are banks, and each bank serves one side of the transaction.

The names are worth internalising because the industry uses them relentlessly. Issuing is everything on the cardholder's side. Acquiring is everything on the merchant's side. A single large bank often does both, but the roles stay distinct: they sit on opposite sides of every transaction, with opposite incentives on fees and disputes.

Between the two banks sits the card network: Visa, Mastercard, and their regional peers. The network does three jobs, and none of them is the one people assume.

• Routing. It carries authorization messages from the acquirer to the right issuer, worldwide, in well under a second. Card numbers begin with a bank identification number precisely so the network can route on the first digits.
• Rules. It publishes the scheme rules: a private legal system covering how disputes work, who eats fraud losses in which scenario, what a merchant may and may not do. Every participant contracts into it.
• Net settlement. Once a day, it computes what every member bank owes every other member bank, nets it down, and instructs settlement of the differences.

Gotcha: the network does not issue cards, does not lend, and does not hold your money. When people say "Visa charged me", every part of that sentence is wrong. The issuer charged them, over rails Visa operates.

The same shape underlies every card payment on earth: promises flow one way around the loop, money flows the other way, later.

At purchase time, the request travels merchant to acquirer to network to issuer, and the approval travels back the same path in reverse. Nothing but messages.

At settlement time, typically the next business day, the flow reverses and becomes money: the issuer pays the network's settlement process, the acquirer receives, the merchant gets paid, and the issuer collects from the cardholder at the end of the month for credit, or immediately for debit.

Hold onto the asymmetry: the message loop runs in about a second, per transaction. The money loop runs in batches, on a schedule, in bulk.

The issuer and the acquirer are both banks, but the economics could not be more different, and the difference explains most industry behaviour.

The issuer's business is the cardholder. It takes real credit risk on every credit card, funds the interest-free period, eats most fraud losses, and runs rewards programmes. Its revenue is interest, cardholder fees, and, crucially, interchange: a fee it receives on every transaction, paid by the merchant's side. Rewards points are largely interchange, recycled.

The acquirer's business is the merchant, and it is a volume business with thin margins. It earns a slice of each transaction for carrying the merchant relationship, the payout logistics, and an underappreciated risk: if a merchant takes payment for goods it never delivers and then disappears, the chargebacks land on the acquirer. That is why acquirers underwrite merchants like small credit risks, and why a new merchant selling event tickets a year in advance finds acquiring surprisingly hard to get.

Around the four parties has grown a service industry, and the labels confuse everyone because companies keep absorbing each other's roles.

Modern full-stack providers such as Stripe and Adyen collapse the whole column into one integration, and Adyen additionally holds its own banking licences, making it its own acquirer across much of the world. The four-party model does not change underneath; what changes is how many of the boxes one company operates.

Two variations compress the model, and both are worth recognising.

In a three-party scheme, one company is both the issuer and the acquirer. American Express is the classic case: it issues the card to the cardholder and signs up the merchant itself, so there is no interchange between separate banks, just one company's internal pricing. That is part of why its merchant fees have historically been higher and why its dispute handling feels different: the same firm sits on both sides.

An on-us transaction is the four-party model degenerating by coincidence: the cardholder's issuer and the merchant's acquirer happen to be the same bank. The transaction can be authorized and settled inside that bank's own ledger without crossing the network at all. In markets dominated by a few large banks, a meaningful share of domestic volume is on-us, which quietly matters for who earns fees and whose outages take payments down.

Here is the question that separates people who have absorbed the model from people who have memorised the boxes.

That gap between the promise and the money is not an implementation detail. It is where the next lesson lives: authorization, clearing and settlement are the three phases that turn the 09:00 promise into cash.

The four-party model runs on two layers of law, and practitioners need to know which layer answers which question.

The scheme rules are private law. Visa and Mastercard each publish rulebooks running to hundreds of pages that bind every issuer, acquirer and, through the acquirer's contract, every merchant. Chargeback time limits, liability for fraud, honour-all-cards obligations, data standards: these live in scheme rules and change on scheme schedules.

Public regulation sits on top and varies by jurisdiction. The EU capped interchange fees by regulation and mandated strong customer authentication through PSD2, both covered later in this course. The US capped large-bank debit interchange under the Durbin Amendment, a rule that has spent years being contested in courts and revisited by the Federal Reserve. Other markets run national schemes with their own rules entirely.

When a payments question starts with "is that allowed", the first move is always the same: work out whether the answer lives in scheme rules or in statute, because the two are amended by different people at different speeds.