What the 67-Score measures

The 67-Score is a single number from 0 to 5 that summarizes how well a credit card serves a young adult with average spending habits. Higher scores indicate better all-around value for our audience.

The five criteria

Criterion	Weight	What we measure
Rewards value	30%	Estimated annual rewards based on $2,500/month benchmark spend
Fees and costs	25%	Annual fee, foreign transaction fees, penalty APR
Welcome bonus	15%	Dollar value relative to spending requirement
Benefits and perks	15%	Travel insurance, purchase protection, cell phone coverage, lounge access
Accessibility	15%	Credit score requirement, approval likelihood for young adults

How it works

Each criterion is scored 0 to 5 independently, then multiplied by its weight. The weighted scores are summed to produce the final 67-Score.

Example: Apple Card

• Rewards: 3% on Apple, 2% Apple Pay, 1% else = ~$450/yr on benchmark spend = 3.0/5
• Fees: $0 annual fee = 5.0/5
• Welcome bonus: None = 0.0/5
• Benefits: Daily Cash, no foreign transaction fees = 3.0/5
• Accessibility: Good credit needed = 3.0/5
• 67-Score: 0.30(3.0) + 0.25(5.0) + 0.15(0.0) + 0.15(3.0) + 0.15(3.0) = 0.9 + 1.25 + 0 + 0.45 + 0.45 = 3.1/5

Editor overrides

In rare cases, our editor may override the algorithmic score when the formula does not capture qualitative factors (e.g., exceptional customer service, ecosystem value). Overrides are documented per card.

Updates

Scores are recomputed quarterly and whenever card terms change materially.

---

The computed figures

Every dollar figure on a calculator, worth-it, bonus-history or comparison page comes from the same small set of inputs and formulas, documented here. If a number on the site disagrees with this page, this page is the bug.

Spending profiles

Earnings are modelled on three named profiles rather than one, because "how much does this card earn" has no single answer. Category shares are derived from Consumer Expenditure Survey patterns for under-35 households and rounded to readable numbers; the same shares scale across all three levels.

Profile	Monthly spend	Share of spend by category
Light spender	$1,200	Groceries 18% · Dining 12% · Travel 6% · Gas & transit 8% · Online 10% · Streaming 2% · Everything else 44%
Typical spender	$2,400	same shares
Heavy spender	$4,500	same shares

Headline figures ("earns about $X a year") always use the Typical profile. Category sliders on calculator pages let you replace the profile with your own numbers; the page's static tables never move.

Earn-rate arithmetic

annual dollars = monthly spend × 12 × earn rate × point value

A 2% cash-back card is stored as rate: 2, pointValue: 0.01 — identical maths to a 2×-points card redeeming at one cent. There is no separate cash-back formula.

Point valuations

We value points and miles at a flat $0.01 unless a card's own terms state otherwise. This is deliberately conservative:

• It is what a cardholder gets from the simplest redemption (statement credit or portal booking) with no research and no transfer partners.
• Transfer-partner valuations — the "this point is really worth 1.8¢" arguments — depend on redemptions most people never make, and they are the standard way premium cards are made to look better than they are.
• A conservative baseline means our figures understate rather than overstate. If you extract more value than we model, good; the reverse would be a problem.

Where a card's structure genuinely changes the maths (fixed-value redemptions, guaranteed multipliers), that is recorded per card and reflected in its rates.

Welcome bonus valuation

The current offer is stored with the date we recorded it and a dollar value at the same $0.01 baseline. Bonuses are shown with that date attached, because they change frequently and an undated bonus figure is worthless. Spending requirements are not deducted — they are a condition, not a cost — but they appear in the card's own review.

"Is the annual fee worth it" — the breakeven model

A fee card is not compared against earning nothing. It is compared against the best no-annual-fee card in the same category, chosen automatically as the highest typical-profile earner among $0-fee, rate-complete, currently-open cards (ties broken by 67-Score, then alphabetically, so the choice is stable).

margin = (fee card earnings − baseline earnings) − annual fee
breakeven = the monthly spend where margin reaches $0

If the margin never reaches zero, we say so plainly: the fee has to be justified by credits and perks, not by earn rates. Statement credits are not counted automatically — a $120 dining credit is only worth $120 if you would have spent it anyway, so credits are described in the card's caveats rather than silently added to its value.

Comparisons

Two-card comparisons use the same engine, net of annual fees, and report the spending level where the winner changes (the crossover). The verdict shown on a comparison page is computed at page render from live card data, so it cannot go stale; any editorial commentary appears separately, marked as such.

Bonus history statistics

Where we have tracked two or more offers for a card, we report the current offer's percentile against that card's own history, plus its minimum, median and maximum. These are statements about our tracking window — which begins when we started recording that card, not when the card launched — and they are descriptive, never advice about whether to wait.

Where the data comes from, and its limits

• Rates, fees and APRs are recorded per card in typed fields, with the date a human last verified them against the issuer's terms.
• Cards whose rewards structure cannot be captured honestly in flat category rates — rotating quarterly categories, choose-your-own categories, brand-locked multipliers, first-year-only promotional rates — are excluded from computed pages entirely rather than approximated. Their reviews stay up; their calculator and worth-it pages do not exist.
• Caps and qualifiers ("4× on the first $25,000") are recorded as caveats and shown next to the numbers they qualify. Our earnings figures do not apply caps, so a heavy spender in a capped category will earn less than we model — the caveat line is where that is disclosed.
• Machine-readable versions of everything above are published at /data under CC BY 4.0.

Corrections

Issuers change terms without notice and our data can lag. If a number here is wrong, the issuer's own application page controls — and we would like to know: the correction path is on our editorial guidelines page.