Methodology - How We Verify
Where the facts come from
Every card in the database is documented from three source classes:
- Operator primary sources - fee schedules, cardholder agreements, supported-country lists, issuer disclosures. Fetched and dated.
- Independent editorial sources - established reviews and media. Checked for affiliate incentives before being treated as independent.
- Community reports - Reddit threads, forum posts, social media with concrete details (dates, amounts, ticket numbers). Linked individually.
Every fact on a card page traces to a source you can open. Facts that cannot be verified are labeled "not documented" - never guessed.
How scores work
Each card gets seven sub-scores (0-10). They are editorial ratings of the documented evidence per dimension - a card with excellent terms but no published fee schedule scores low on fees, because "great but undocumented" is not verifiable. The total is a weighted mean:
| Dimension | Weight | What it rates |
|---|---|---|
| Privacy | 20% | KYC depth, custody model, Tor support, data-sharing pattern. Privacy leads deliberately: without KYC-free spending there is no financial self-determination, and a full-KYC card cannot substitute for a no-KYC one. |
| Bitcoin flow | 23% | How directly bitcoin gets on and off the card: on-chain top-up > in-app conversion > collateral-only > none. Also whether the sale timing is documented - the tax moment. |
| Fees | 18% | The full stack the provider publishes: monthly, top-up, FX, ATM, inactivity. Undocumented fees cap this score. |
| Reliability | 14% | Operator history, incidents, how programs handled freezes and outages, issuer transparency. |
| Rewards | 9% | Cashback after caps and tier gates - the honest net, not the headline number. BTC-denominated beats token-denominated at equal rates. |
| Availability | 9% | Documented regions, delivery, excluded countries. |
| Limits | 7% | Spend/ATM/load ceilings as published. |
Worked example: privacy 4, btc_flow 8, fees 6, reliability 7, rewards 5, availability 7, limits 6 → 0.20·4 + 0.23·8 + 0.18·6 + 0.14·7 + 0.09·5 + 0.09·7 + 0.07·6 = 6.2.
Rules the weights sit on top of:
- Red flags gate the ranking. A card with an unresolved red flag on record cannot appear above any card without one - it is listed in its own group with the flags documented. This is deliberate: documented warnings beat a good score.
- Missing sub-scores cannot raise the total - the mean runs over the dimensions actually scored.
- "Verified" is earned. A card is marked data-quality "verified" only where individual facts carry a dated source check; otherwise it is "reported" or "unverified". Unverified cards are not recommendable.
How freshness works
Facts carry a check date. Operator pages are re-verified on a schedule (90-day fact re-verification; material changes such as fees, availability or status trigger a review update). Reviews generated with AI assistance are drafts until a human approves them - approved versions show their generation date. The change feed records what changed and when.
Conflicts of interest
Rankings are never sold. Some outbound links carry referral programs; every primary button with a referral link is visibly tagged, and the ranking logic never sees affiliate data. Providers cannot pay for placement, pay to remove cons, or pay to stay out of the graveyard.
Errors
When we get something wrong, the correction is dated and visible. Report errors via the contact page.