COCA Card

×

How the editorial score is calculated

Every number in this section comes from a deterministic formula applied to this card's verified data fields. Nothing is hand-tuned and nothing is voted on: the same record under the same scoring version always produces the same numbers.

1 · Eleven criteria, fixed weights

Each criterion is scored 0–100 from the raw fields. Money and percentages pass through fixed step ladders (an FX fee of 0% scores 100, 0.5% → 90, 1% → 75, 2% → 40, above 3% → 0), boolean features earn listed points, and compatibility is shrunk towards the mean so a card tested on 2 of 2 services does not outrank one tested on 16 of 18. The weights are identical for every card and add up to 100.

Base score = sum of contributions ÷ sum of the weights actually known.

Transparency measures whether the provider itself discloses what a customer needs to know: the legal issuing entity and its licence, the fee schedule, limits, the custody model, reward terms, the country list, plus an official site and a public channel. Each item is graded by where our editors verified it — on the provider’s own site 100, a reliable third-party report 60, a text citation 40; items we could not source are left out rather than scored 0. A provider whose issuer page names no legal entity scores 0 on that item.

Sign-up access asks whether a new reader can actually get the card today: open sign-up 100, invite-only 70, waitlist 10, closed to new users 0. Editors set availability.onboarding.mode after checking the sign-up flow; without an explicit entry the mode is assumed from the card status (active → open, beta → invite, waitlist → waitlist). A great card you cannot obtain should not top the list — and because the published terms of a waitlisted or closed card have not been tested by real users yet (you only learn the truth once you can spend with it), its rank score is additionally multiplied by 0.95 (invite), 0.80 (waitlist) or 0.70 (closed).

2 · Unknown is not zero

A field we have not verified is stored as null, which is deliberately different from a verified “no” and from a verified zero. A criterion with too little known data is dropped from the average instead of scoring 0, so a card is never punished in product quality for a gap in our research. Criteria covered reports how much of the weight survived — it tells you how much of the Score rests on real data.

3 · Confidence — how well documented the record is

Quality and evidence are measured separately. Twelve data sections each contribute weight × completeness × source factor × freshness factor. Source factor: issuer’s own page 1.00, other URL 0.90, plain-text citation 0.85, no source 0.75. Freshness factor: verified within 180 days 1.00, within a year 0.90, older 0.80, never verified 0.85.

4 · Rank score — what the list is sorted by

Product quality alone would let a thin record with three flattering criteria beat a fully documented card. So for ranking, every criterion still unknown is imputed a weak 40 at its own weight, and the result is then damped by confidence. Consequence: filling in a criterion that scores above 40 always improves the rank, and a fully documented card can never be overtaken by an equally good card with gaps. Ties break on coverage, then on confidence.

5 · Penalties

Editor-recorded deductions, plus one automatic deduction when the issuing entity cannot be identified, are subtracted in raw points after the weighted average.

• -3 Issuer / regulated entity not identified (automatic)

Reproducibility

Scores are recomputed whenever the card record changes and cached in the database; the raw record stays the source of truth. Only numeric, boolean and enum fields are read — translated text never moves a score.

v7 · 2026-09-05