How we rate crypto cards
One formula for every card, recomputed from the data on every build. Here is exactly what goes in, how much each part weighs, and what happens when a provider does not publish a number.
The short version
Each card gets a score from 0 to 10. Fees and rewards decide 56% of it, because they determine what the card actually costs or pays you. Ratings from App Store and Google Play users count for 10%, so real-world experience with the app weighs in. The remaining 34% covers limits, where you can get the card, KYC, who holds your money, supported coins, the provider's track record, app store ratings and how much the provider publishes. Version 1.2.
Commissions are not an input. The formula reads only the card data shown on each card page. Whether we have an affiliate deal with a provider, and how much it pays, cannot change a score. Cards without any affiliate link are rated the same way.
Criteria and weights
| Fees (28) |
|
|---|---|
| Rewards (28) |
|
| Spending limits (4) |
|
| Availability (8) |
|
| KYC and privacy (4) |
|
| Custody (5) |
|
| Coins and networks (4) |
|
| Track record (5) |
|
| App store ratings (10) |
|
| Data completeness (4) |
|
How a number becomes a score
Measurable values (fees, cashback, limits, country counts) are compared across all cards in the catalog, so a score always means “how this card does compared with the others today”:
Higher is better (cashback, limits): score = 10 × (value − P10) ÷ (best − P10)
Scores are capped between 0 and 10.
P90 and P10 are the 90th and 10th percentiles across cards. Using them instead of the single worst card stops one extreme outlier from squashing everyone else. Limits, country counts and the number of coins and networks are compared on a logarithmic scale, so going from 10 to 50 countries counts for more than going from 150 to 190.
Some parts are scored on a fixed scale instead:
- Custody: self-custody 10, partly custodial 6, custodial 3.
- KYC: no KYC 10, light 7, full 4; one extra point when the wallet works without KYC and only the card needs it.
- Apple Pay and Google Pay: 5 points each when supported, 1 when “coming soon”.
- Track record: card age up to 36 months (half), a named issuer (30%), no security incident in the last 12 months (20%).
- Perks: each distinct perk, yield programme or points scheme counts, up to six.
The criterion score is the weighted average of its parts, and the total is:
Every card page shows the criterion scores, the individual part values and any deduction, so you can recompute the total by hand.
No-KYC cards use the same formula. A card issued without an ID check at signup gets the top KYC score, 10 out of 10, which puts it at most 2.4 points out of 100 ahead of a card with full KYC. What usually comes with skipping KYC shows up in the other criteria: these cards are custodial (3 out of 10 for custody), charge 2.8% to 5.49% to load, are new, and rarely publish limits, issuers or country lists, so more of their parts are estimated and their data-completeness score is lower. The no-KYC cards page explains the catch.
Early access deduction
Some cards launch in stages: the app runs on one phone platform only, sign-ups open from a waitlist, and the other platform and large markets are announced as “coming soon”. Their terms can look excellent on paper while most people cannot use the card yet, and launch terms often change once it opens up. While a provider says its card works on only one of iPhone and Android, we take 5 points out of 100 off the total. The deduction is temporary: it goes away once the card ships on both platforms and opens the markets it has announced. It currently applies to Ethena Pay.
Fees, in detail
- Cost of the card in year one: the cheapest card to issue (virtual or physical), plus the annual fee and 12 monthly fees on the free plan.
- Purchase fee: the percentage charged on every purchase, with fixed per-purchase fees converted at an average purchase of $50.
- Foreign-currency fee: the extra charge on a non-USD purchase, converted the same way.
- Top-up fee: the fee on depositing crypto.
- ATM fee: the cost of withdrawing $200. A card with no ATM access scores 2 out of 10 on this part.
Rewards, in detail
The main rewards figure is the effective cashback: what the free plan pays on $1,000 of spending in a month, after monthly bands and payout caps, multiplied by how much the reward coin is worth to a typical holder. Cashback that can only be spent on the card counts at 90%.
| Dollar or euro stablecoins and cash | × 1 |
|---|---|
| Your choice of asset, including a stablecoin | × 1 |
| Bitcoin | × 0.9 |
| Large-cap tokens such as AVAX | × 0.85 |
| The provider's own or ecosystem token (ETHFI, XPL, WHYPE) | × 0.75 |
| Points that cannot be redeemed yet | × 0 |
The best plan figure looks at every plan on $3,000 of monthly spending and subtracts the plan's yearly fee (divided by 12). Plans unlocked by staking tokens or by trading volume count at 75%, because the lock-up has a cost we cannot price.
Missing data
When a provider does not publish a figure, we do not guess it. That part gets the 25th-percentile score of the cards that do publish it: worse than a typical card, but not zero, so a card is not wrecked by one undocumented fee and not rewarded for hiding it. Estimated parts are labelled on the card page. The separate data completeness criterion (5 points) counts how many of 15 key facts are published.
App store ratings
We read the rating and the number of ratings from the App Store and Google Play (US storefronts) and record the date. To stop five 5-star ratings from beating thousands of 4.5-star ones, we blend them with a prior: blended = (100 × 3.5 + Σ rating × count) ÷ (100 + Σ count). Some listings cover a whole wallet app rather than the card; the card page says so.
Sources and conflicts
Every figure carries its source. When sources disagree we trust, in order: the provider's official site and documentation, then DefiLlama, then SpendNode, PaymentScan and other third-party guides, then Telegram reports. If a conflict cannot be settled, we use the more cautious value (the higher fee or the lower reward) for the score and show both figures on the card page.
What the rating is not
- It is not a hands-on review. We do not invent user reviews or test results.
- It does not weigh how safe or solvent a provider is beyond what is public (issuer named, incidents reported).
- It is not financial advice. Which card suits you depends on your country and how you spend.