Visa wants you to pay for coffee with a card

Visa grows mainly by volume, but the remaining third of its growth does not come from raising prices: it comes from every purchase getting smaller, and the line that contributes most is charged per transaction, not per amount.

The prompt

Visa bills more every year. The CEO's question is simple: are we growing because the card gets used more, or because we charge more every time it is used? And if it is the second, are we the ones deciding it?

This is the classic shape of a profitability case: you are handed a result that moves and asked to say what moves it, with numbers instead of adjectives.

Where the data comes from

Visa is listed in the United States, so every year it files a 10-K: the mandatory annual report to the SEC. It is far more than a balance sheet, it is signed by the executives, and lying in it is a federal crime. It is free to search on EDGAR(opens in a new tab), the SEC's public archive, by ticker (V for Visa, MA for Mastercard, which files exactly the same form).

The parts that are useful in a case:

In a real interview you cannot look anything up: the interviewer holds the data. The 10-K is read beforehand, so you arrive knowing how the company makes money. Inside the interview you ask for one specific figure and say what for: "to separate volume from price I need payments volume and net revenue for the last three years, do we have them?". Asking for "all the data you have" costs you points.

Step 1: who keeps what

The question: a $100 coffee paid by card leaves the shop with less than $100, because the merchant pays a discount rate. Who keeps that difference, and how much of it is Visa's?

Four parties take part in a card purchase, which is why it is called the four-party model: the cardholder, the merchant, the issuing bank (the one that gave out the card) and the acquiring bank (the one that gives the merchant its card-acceptance service). Visa is none of the four: it is the network that connects them and sets the rules. The terminal, the gateway and the processor are plumbing; they charge, but they do not decide the model.

Two things the diagram deliberately keeps apart, because this is where the model gets confused most:

• The issuer authorises, not the acquirer. The acquirer and the network carry the request; the one that checks the limit, the balance and the fraud risk, and says yes, is the bank that will put up the money.
• The cardholder does not pay the merchant. The merchant is paid by its acquirer, a day or two after the purchase. The cardholder pays the issuer only on the statement. That is why the issuer is the one taking the risk: it fronts money it may never collect, and that is what makes its slice the largest of the three.

And interchange is always read off a table, never as a single number. Same $100 purchase, changing only the plastic:

The rule that explains the split: in payments, whoever takes the risk takes the big slice. The issuer pays the merchant now and bills the cardholder at the end of the month; if the cardholder does not pay, the issuer loses. The total the merchant pays is called the merchant discount rate.

That fraction each party keeps has a name: take rate. Visa's is tiny, 28.79 basis points, and Visa is still one of the most profitable companies in the world. That is where the case starts.

Step 2: building the tree

The question: for revenue to grow next year, what levers are there?

A lever is a variable you can move that changes the result. The craft here is writing the result as a formula and walking through it term by term:

revenue = volume × take rate
              │
              └── volume = transactions × average ticket
                                 │
                                 └── transactions = active cards × times each one is used

That is a driver tree, and it is what the interviewer expects to see by minute two. On the volume side there are two levers, and both are Visa's historical business: paying by card for what used to be paid in cash (cash displacement) and having more cards and more merchants that accept them. The latter is a two-sided network: each side pushes the other.

Step 3: volume or price

The data. Year-on-year growth, calculated on the figures published in the 10-K filings:

Revenue grows faster than volume every single year. If revenue = volume × take rate, that can only mean one thing: the take rate is rising. And it rises without exception, from 24.17 basis points in FY2021 to 28.79 in FY2025.

The arithmetic. If A = B × C, then (1 + growth of A) = (1 + growth of B) × (1 + growth of C). Two versions are worth having at hand:

• The quick one, the one you say out loud: for small rates, growth rates add up, so 11.34% − 6.98% ≈ 4.36%.
• The exact one: 1.1134 / 1.0698 = 1.0408, that is 4.08%.

Checking it by a second route, which is the most valuable habit of the craft: 28.79 / 27.66 = 1.0408. The same number from different data.

So FY2025 growth splits up: 6.98 points of volume and 4.08 of take rate, which give the 11.34 of revenue. Two thirds and one third.

Step 4: the take rate rises, but nobody raised a price

Here is the part that makes the case interesting. Nobody at Visa sets the take rate: it is a calculated number, revenue divided by volume. Its rising does not prove that Visa raised the price for anyone.

The corner-shop analogy: if it sells the same $100 coffees and also more $500 cakes, its average ticket rises without a single price being touched. The mix changed.

The data. Revenue by category, in millions of dollars, published in the 10-K FY2025:

Every line grows faster than volume (6.98%), and that is why the take rate rises: none of them is holding it back. What is missing is knowing which one rules.

Out of the $4,074 million that net revenue grew, each line contributes:

They add to 100% once the negative sign is counted. If it does not close at 100, there is an error somewhere.

Step 5: why the take rate rises

The line that rules, with 55.9% of the growth, is data processing. And that line is charged per transaction, not per amount. With that, two numbers from the 10-K are enough to close the case:

If purchases grow faster than money, every purchase is smaller: the average ticket falls 2.90% (1.0698 / 1.1018 = 0.9710).

There is the mechanism. A $100 coffee pays one processing fee. Five $20 purchases pay five. Same volume, five times the revenue on that line. The take rate is revenue over volume: the denominator did not move and the numerator multiplied. Nobody raised any price.

That is why cash displacement, which sounds like a marketing story, is literally the engine: every tip, every bus fare and every coffee that leaves cash and enters the card pushes the take rate up.

Two caveats so as not to oversell it, because the ticket explains part of it and not all:

• International (36.8% of the growth) lifts the take rate through mix: it yields more per dollar.
• Value-added services went from $8.8 billion to $10.9 billion (+23.9%) and lift the take rate by construction of the metric: they add to the numerator without adding volume to the denominator. They are not more profit per transaction.

The close, in 30 seconds

This is how a case ends: the interviewer asks for the answer as if you were in a lift with the CEO inside. Answer first, then the evidence, then the risk.

The concepts this case uses

To practise it

The same exercise with Mastercard's 10-K on EDGAR(opens in a new tab): it is the same business with a different mix, and the fiscal year is different, so the first job is noticing that one company's twelve months are not the other's.

Three questions to take away, which are the ones this case trains: what formula explains the result? how much does each term contribute in money rather than in percentage? and did somebody decide the number that moved, or did it move on its own?