Payment processing fees: the quote is a percentage and the bill is a total

The version of this piece that has been up since 29 July was built on one merchant's statements. I read all of it, I did the arithmetic, and I published a figure of 64 basis points. You cannot check any of that. The statements are not mine to publish, the business would be identifiable from them in about four seconds, and so the whole argument rested on trust me, which is exactly the posture I complain about when a processor takes it. I had it wrong in method even where the arithmetic held.

This is the corrected version, rebuilt entirely from numbers anybody can pull in an afternoon: one published price list and one Federal Reserve data collection. It came out differently, and the part that moved is the part I was most confident about, which is a pattern I have now hit twice in a fortnight and have started treating as a signal rather than as bad luck: the sentence I would have defended hardest in an argument turns out to be the one carrying an assumption I never wrote down anywhere.

The headline rate is not a rate

Payment processing fees are quoted as a percentage plus a fixed amount. The fixed amount is where the quote stops describing what you pay, and because it is stated in cents while everything around it is stated in per cent, it slides past the part of the brain that compares offers, which I suspect is not an accident but cannot show. Stripe publishes 2.9% plus 30 cents for an online card, 2.6% plus 30 cents in person, an extra 1.5% when the card was issued outside the country, and 15 dollars when a dispute arrives.

Take the 30 cents on its own, and take a shop rather than a spreadsheet. On a 100 dollar order the fixed fee is 0.3 per cent, so the real cost is 3.2. On a 40 dollar order it is 0.75 per cent and the real cost is 3.65. On a 12 dollar order it is 2.5 per cent and the real cost is 5.4 per cent, which is a different business from the one described by the quote. The spread between those three shops is 2.2 percentage points. Every one of them was sold the same headline. Nobody lied to any of them.

Nothing in that paragraph is secret and none of it sits in small print. It is arithmetic the quote invites you not to do, and average order value is the single variable that decides the answer. Two shops on identical terms can sit a full percentage point apart with nobody treating either of them differently. You should compute this before you compare providers, because a provider who is cheaper at 90 dollars can be dearer at 20, and the crossover point between two published price lists is a single division that neither salesperson will do in front of you, partly because it takes ten seconds and mostly because whichever of them is on the wrong side of it would have to say so out loud.

Then the part you do not control at all

Underneath the processor's price sits interchange, which the processor collects and passes to the bank that issued the card. On debit this is regulated and the regulation is short.

The Federal Reserve's Regulation II says that a covered issuer "may not receive, for any electronic debit transaction, an interchange fee that exceeds $0.21 plus 0.05 percent multiplied by the value of the transaction, plus a $0.01 fraud-prevention adjustment, if eligible". That is the cap, and it has held since 2011. Small banks are exempt from it. The exemption is where the interesting number lives.

In 2023 the average interchange fee on a covered debit transaction was 22 cents on dual-message networks and 24 cents on single-message. On exempt transactions the average was 52 cents. Split by network, exempt dual-message came to 62 cents, down from 64 in 2021, and exempt single-message came to 27 cents. The Board notes that exempt averages "gradually increased from $0.43 since the regulation took effect to $0.52 in 2021, and remained steady through 2023".

Read those two numbers next to each other. Twenty two cents against sixty two. The same card, the same basket, the same shop, and the cost to you is nearly three times higher because of who issued the plastic. You cannot see it at checkout, you cannot decline it, and there is no version of your pricing page that changes it.

The whole debit and prepaid pool came to 34.12 billion dollars in 2023, growing at 3.9 per cent a year since 2021. I went looking for the equivalent series on credit, and after 3 hours across the Board's publications I am satisfied there is not one, which is the single biggest hole in this piece.

Working out your own effective rate, which takes about twenty minutes

This is the part worth doing. It needs no tools and no permission from anybody.

Pull 12 months of statements. Add every line the processor took, and there will be more of them than you remember: the percentage, the fixed fee on each transaction, monthly minimums, the gateway charge, chargeback fees, cross border, currency conversion, PCI charges, and the annual fee that arrives once and gets forgotten by March. Divide that total by what you actually sold across the same 12 months. The result is your effective rate, and it is the only rate that has ever cost you anything.

I would set the year boundary at your own financial year rather than the calendar, because the annual charges land unevenly and a January to December cut can either double count them or miss them entirely.

Then do it a second time with refunds excluded from the sales figure. On most gateways the fixed portion of a refunded transaction is not returned, so a shop refunding one order in ten is paying the fixed fee on volume it never kept, and the two versions of the ratio will differ by more than people expect.

Do not negotiate on the headline percentage before you have both figures. It is the number a salesperson can move most easily and it is rarely the one costing you the most.

Five lines to ask for before you sign

I would ask for these in writing, and I would ask before the price conversation rather than after it.

The first is the fixed fee per transaction. It is 30 cents at published Stripe pricing and it decides your rate more than the percentage does.

The second is whether that fixed fee is returned on a refund. At the 4 providers whose terms I read it is not.

The third is the cross border surcharge. Stripe publishes 1.5 per cent on a card issued outside the country, and a shop with 20 per cent foreign customers is carrying 30 basis points of blended cost from that line alone.

The fourth is the dispute fee. Fifteen dollars is the published number, win or lose, and the phrase win or lose is doing more work in that sentence than it looks like, because a dispute you win still costs the fee, the evidence pack, and the hour of somebody's Tuesday that went into assembling it.

The fifth is the monthly minimum, which is the one line nobody quotes and everybody pays. It costs nothing in a good month. It is pure loss in a slow one, and slow months are exactly when it lands.

None of those five is a secret. All five are answerable in one email. It still surprises me how rarely they get asked, and it is a strange thing to be annoyed about on somebody else's behalf.

What the complaint record shows about the other end

Disputes turn a fee question into an operations question. The public file is unusually clear about how common they are, and the 12 month record is large enough to be worth reading rather than summarising.

Of 95,478 credit card complaints in the twelve months to July 2026, 27,503 concerned a purchase shown on a statement. That is 28.8 per cent, and every one of those has a merchant at the other end assembling evidence against a deadline. At 15 dollars a dispute on published pricing, before anybody's time is counted, a shop taking 30 of them a month pays 5,400 dollars a year for the privilege of defending itself. Add 40 minutes of somebody assembling evidence for each one and the real figure is closer to 240 hours and 5,400 dollars together, which is a person-month.

A short digression about the word rate, because it changed how I read every quote afterwards. A processor quotes a rate and bills a total. Those are two different objects, the industry has no incentive to name the difference, and it is the oddest piece of accepted vocabulary I have run into in this subject. Anyway, back to the arithmetic.

Where my earlier version was wrong

The old piece said the gap between a quoted rate and a real one was worth about 64 basis points and that most of it was recoverable by switching pricing models. The first half I can still support with a worked example. The second half I cannot.

Moving from a blended price to interchange plus exposes the interchange, which is 22 or 62 cents on a debit card depending on the issuer, and that portion does not move for anybody. What moves is the processor's own margin. So a switch reveals the split and reduces the part above interchange, and I presented it as though the whole gap were on the table. That was too strong and I want it corrected in the same place it was published.

What I could not establish

I could not find published averages for card-not-present credit interchange comparable to the Federal Reserve's debit series. The networks publish rate tables, which is not the same object as an average of what was actually paid, and the difference matters most when a shop's mix is unusual. Two shops on the same table can land 80 basis points apart.

I do not know what share of small businesses ever computes an effective rate, and the 12 dashboards I have looked at do not make it easy. My guess is that it is small, on the grounds that every processor dashboard shows volume and none of the ones I have seen shows a single figure for total cost divided by total sales, and that is a guess with nothing behind it except the design of those dashboards.

I also could not establish whether the fixed fee is returned on refunds at every provider. It is not returned at the ones whose terms I read, but I read 4 sets of terms, and 4 is not a market. I would like somebody with 20 of them to tell me I am wrong.

The thing I keep going back to

Twenty two cents against sixty two cents, on 34.12 billion dollars a year of debit interchange.

Two identical transactions priced nearly three times apart, and the difference decided by which bank happened to send the customer a card. I keep thinking about it more than is reasonable for a cost line I cannot change. The honest position is that a good chunk of what a small shop pays to accept money is set by decisions nobody at that shop will ever be part of, and the most useful thing you can do is measure the part that is yours.