How we know what we know.

Where the data comes from

For every active card on this site, we first locate the issuer's own pages that carry the canonical numbers: the marketing application page, the Pricing & Terms disclosure, and where applicable the Guide-to-Benefits PDF and cobrand partner disclosures. Those URLs are stored alongside the card in our database, and every field on a card page traces back to one of them.

That rule is deliberate, because a paraphrase of a paraphrase is how stale and wrong numbers spread across this category. We do not lean on another site's summary or a competitor's read of the terms. If a number cannot be traced to an issuer page, it does not ship.

Three-tier sourcing

Not every fact carries the same weight of evidence behind it, so we tag each data point with the tier it came from. That way you can judge for yourself how firm a given citation is:

• Tier 1 (issuer-stated). APRs, fees, signup bonuses, category multipliers, and benefit caps, quoted verbatim from the issuer's own terms pages. This is the firmest tier, and most of what drives a decision lives here.
• Tier 2 (card art). Card material, color, network logo, and design details, taken from the issuer's product photography.
• Tier 3 (aggregated reports). Approval credit-score ranges and similar synthesized values that no issuer publishes directly. Because the issuer is silent on these, we cite the specific report URLs and quote the thresholds verbatim, so you can see exactly where the number came from. Tier 3 facts are labeled as such on the page.

When a fact does not fit cleanly into one of those tiers, we mark it "Not disclosed" rather than guess at it. An honest gap is more useful to you than a confident invention.

Caught the day it changes

Issuers revise card terms more often than they advertise, and a page written once will quietly drift out of date. To guard against that, a daily job re-reads every cited source page, hashes the content, and compares it against the previous day's hash. When anything moves, the affected card is flagged for a fresh scrape, which is how an elevated welcome offer surfaces here the day it posts rather than the week after.

Every card page carries a "Data current as of" timestamp linked to the issuer's official page. A stale stamp is a bug, and we want to know about it. Email support@ccdeck.com.

What language models do and don't do

Language models help us draft the editorial work: the card bios, the plain-English summaries, the prose that sits around the numbers. They never source a fact. Before any numeric or terms-based claim lands in the database, it is grep-verified against the raw bytes of the issuer's page. A second model pass would not count as a check here, since two models can be wrong in the same direction and the errors compound rather than cancel. The check is the bytes.

How we value welcome offers

A 100,000-point Membership Rewards bonus and a $200 cashback bonus do not compare on their face, so the welcome-offer filter on /cards needs a single normalized dollar figure to sort against. We compute that figure from a per-currency cents-per-point rate multiplied by the size of the bonus.

Those rates are set conservatively, at the value an experienced points user would actually budget for. We deliberately sit below the rosy transfer-partner sweet spots the points blogs chase, while staying above the floor of the issuer's own portal redemption. The aim is to rank cards honestly against each other across currencies, which a number you can only hit on a perfect redemption would undermine.

Free-night certificates are valued at their point cap times the issuer's currency rate. A Marriott welcome of "4 Free Night Awards, each up to 50,000 points" therefore computes as 4 × 50,000 × 0.7¢ = $1,400. That is the same accounting an experienced cardholder would run to decide whether the certificate clears the annual fee, which is the only comparison that matters at sign-up.

These rates are not fixed, since they move when the underlying programs change how they price awards. We re-audit them as the catalog grows, and when a valuation shifts materially we note it on the /updates feed so the change is visible rather than silent.

What we don't claim

We are not financial advisors, and nothing on this site is financial advice. We have no view of your credit score, your tax situation, or your goals. What we can give you is what a card charges and what it earns, accurately. The decision about whether that fits your life remains yours.

We also make no claim to list every card. The site indexes only the cards we have personally researched and verified, so the catalog grows deliberately rather than aiming for completeness. A card's absence means we have not audited it yet, and should not be read as a verdict that it is not worth carrying.

How we get paid

Some links on ccdeck are affiliate links. When you apply through one and are approved, the issuer or network pays us a referral fee. That fee has no bearing on where a card lands in any ranking, which is decided by the math alone. When a card without an affiliate program is the right answer, we say so and link straight to the issuer's own application page, fee or no fee.

Our terms carry the full disclosure language.

Mistakes

Anything on this site can be wrong, and the pipeline is built on that assumption rather than against it. If you find a number that does not match what the issuer is showing today, email support@ccdeck.com with a link to the discrepancy. Corrections are pushed within a day, and the change feed on /updates shows the diff so you can see what moved.