How we test crypto cards

What a hands-on review requires

All five have to happen before a card is flagged Tested hands-on. Four out of five is not a hands-on review, it is a well-informed reading, and it gets the Verified against issuer docs label instead.

The card is obtained the way a reader would obtain it, through the public application flow, and funded with our own money. That is what makes the last step possible. You cannot check a settlement rate against a statement nobody generated.

How each number is produced

Three rules, because three kinds of claim account for nearly every disagreement between a landing page and a statement.

• Foreign exchange. Measured against the European Central Bank reference rate for the day of the purchase, never against the rate the issuer quoted. The quoted rate is the thing under test, so using it as the yardstick would guarantee a clean result.
• Yield. Measured as a round trip. Money in, money out, annualised across the days it sat there, with the cost of swapping into and out of the position included. A depositor cannot avoid paying that cost, so excluding it would publish a rate nobody receives.
• Cashback. Read off an individual settled transaction rather than the running lifetime total, which on both cards we have measured lags by days or weeks.

Where the arithmetic cannot separate two candidate answers, we say so and keep the row. The full log of results, including the inconclusive one, is on advertised vs measured: 5 figures across 2 cards so far.

What gets published, and what gets blacked out

A card page for a card we hold carries the screenshots behind the claims, captioned with what each one proves rather than what it depicts. Before any of them goes up, four things come out: on-chain transaction hashes, issuer transaction IDs, account handles, and card numbers.

The hashes matter most. A single spend transaction on a self-custody card resolves to the wallet that funded it, and from there to its full balance and history. Publishing one to prove a fifteen-cent cashback payment would be a poor trade.

What we do not claim

• One measurement is one measurement. A rate confirmed on a single purchase, in one country, on one day, is reported as exactly that. Issuers vary rates by region and tier, and our sample cannot see it.
• We do not test what we did not buy. No sandbox accounts, no demo environments, no issuer-arranged trials.
• A hands-on flag does not mean every field was measured. Most fields on a tested card still come from the issuer's published terms. Only the figures in the measurement log were produced by us.
• An account of one person's experience is not a rate. When a card declines a merchant category or suspends an account, we describe what happened to us and say how many times we reproduced it. We do not turn it into a percentage.

Why only 4 cards

Each one costs an application, a KYC check, real money on the card, and then months of waiting before there is anything worth reporting. That rules out testing at the pace a comparison site publishes. A site claiming hands-on coverage of a hundred and seventy cards would be describing something other than what this page describes. The individual costs behind our own figures are in the measurement log.

The rest of the catalogue is honest about its own depth: 153 cards carry Verified against issuer docs, 7 are announced products with nothing to test yet, and 6 carry no verification badge at all. The tier counts and the rating maths are on the methodology page; sourcing, independence and corrections are on the editorial policy page.

If you can show we got it wrong

Every measurement above ships with its method so that it can be argued with. If you run the same test and get a different number, we want it. Use the Report a data error button on any card page or email contact@sweepbase.net. Applied corrections show up in the public data changelog.