What the card said, and what the statement said

5 published numbers, checked against our own money on 2 cards. 2 paid what they promised, 1 did not, and 1 turned out to be a cost that appears on no fee schedule at all.

Every other number on this site was copied out of a fee schedule. We read what the issuer published, wrote it down, and linked the source. That is honest work and it is also the ceiling of the method: a fee schedule can only tell you what the issuer chose to put in one.

So we started buying things. Below are the 5 figures we have produced ourselves so far, on 2 of the 4 cards an editor here actually holds. Small sample, and we would rather publish it small than wait until it looks impressive.

The log

Each row gives the issuer's claim, what we got, and enough method that you can go and disagree with us.

The finding that does not fit in a fee schedule

One euro cost us a dollar sixteen. Tria itemises that charge better than almost any card we have used, and the itemisation is still incomplete: the two fees it names, Visa's 1% and its own 0.50%, come to 1.50%, while the real distance from the interbank rate was 2.64%. The missing 1.12% sits inside the exchange rate itself. It is not a line item anywhere, in any document, so no amount of reading terms would have found it.

This is worth being precise about, because it is not an accusation of overcharging. Tria advertises up to 3% and we paid 2.64%, which is inside the claim. The point is narrower and, we think, more useful: a card can be honest about its ceiling and still route most of the cost through a channel that has no name on your statement. Our own guide to hidden FX fees was built by reading 170 fee schedules, and this is the one thing that method structurally cannot see.

The comparison across our two cards makes it concrete. Ether.fi charges 0% FX on euro purchases and settles them out of a dollar balance. Tria charges up to 3% and buries about 1.1 points of that inside the rate. Two Visa cards, both self-custodial, neither of them lying. A reader comparing them on published fees alone would still miss the difference that shows up on the statement.

What we could not settle

The Tria cashback row is the one we most wanted to nail down and could not. Six cents came back on a dollar-sixteen purchase, and cashback pays to the cent, so that figure sits comfortably with the advertised 6% and just as comfortably with 5%. The honest answer is that the test was too small, and the fix is to spend twenty dollars rather than one. We are publishing it in that state deliberately. A log where every measurement resolves into a clean story is a log that quietly dropped the boring ones.

Why the sample is this small

Measuring a card costs money and takes months. The Tria yield figure needed a thousand dollars parked from January to May. The physical card that produced the euro purchase was paid for on 24 January. There is no way to do this at the pace of a comparison site, which is exactly why comparison sites, ours included, run almost entirely on transcription.

So the plan is slow. We hold 4 cards, we have measured 2 of them, and each new figure gets added here with its method attached. If a number below turns out to be wrong, we would rather hear it from a reader than keep it tidy. The card pages carry the underlying screenshots, redacted where an on-chain hash would expose a wallet.

Method

Foreign-exchange cost is measured against the European Central Bank reference rate published for the day of the purchase, not against the issuer's own quoted rate, because the quoted rate is the thing being tested. Yield is measured as a round trip: money in, money out, annualised over the days held, with swap costs included, since a depositor cannot avoid paying them. Cashback is read off the settled transaction rather than the running total, which on both cards lags by days. The full process, including what we do not do, is on how we test.

Figures on this page were produced by an editor spending their own money on cards they hold. Issuer claims are quoted with the date they were read. The card catalogue and its sources are on the dataset page.

About the author. Mihail B. runs Sweepbase, an independent crypto-card comparison and research project tracking 170 cards across regions, networks, and BIN sponsors. New research drops are announced on his personal LinkedIn and the Sweepbase company page.