METHODOLOGY
How we score every card
The full recommendation engine, published. Every weight, every filter, every assumption — so you can check our work instead of trusting our word.
Every recommendation on CardAdvisor comes from one algorithm, applied identically to every card we track. No editor reorders results. No issuer buys a slot. What follows is the actual engine — not a simplified story about it.
Step 1 — Hard filters
Before any scoring happens, cards are removed from consideration entirely when:
- —The annual fee exceeds the maximum you said you’re willing to pay.
- —The issuer’s stated minimum income is more than 20% above yours — recommending a card you’d be declined for helps nobody.
- —It’s your first credit card and the card isn’t an entry-tier product. Premium cards are a bad first application.
- —The card has been discontinued or closed to new applicants.
Step 2 — Five weighted factors
Surviving cards are scored 0–100 on five factors. The weights below are the real ones in production — not an approximation.
Category reward match
40%How much the card actually pays on the things you buy.
For every spend category in your profile we compute: monthly spend × the card’s reward rate in that category × the published point value × 12. Monthly earn caps are applied — a "5% cashback" card capped at ₹1,000/month is scored on the capped number, not the headline.
Fee ROI
25%Whether the card is worth its fee for you specifically.
Net value = annual reward value − effective annual fee (+ joining bonus in year 1). If your total spend crosses the issuer’s waiver threshold, the effective fee is ₹0. A ₹12,500-fee card can beat a free card — but only when the math clears the fee.
Perk relevance
20%Perks valued only if you would actually use them.
Lounge access scores zero unless you told us you travel. Fuel surcharge waivers count only if fuel is one of your top categories. Golf and concierge only register on premium tiers. We never inflate a card with perks you won’t touch.
Reward type alignment
10%Cashback people get cashback; miles people get miles.
Full marks when the card’s primary reward currency matches your stated preference, partial when you have no preference, zero on a mismatch — points you’ll never redeem are worth nothing to you.
Eligibility confidence
5%How likely the issuer is to actually approve you.
Cards where your income clearly meets the stated minimum score full weight. Borderline cases are discounted, and cards where you’re below the stated minimum are heavily discounted and shown with a warning — not hidden, not oversold.
Step 3 — Diversity rule
Your top five never contains more than two cards from the same bank. If one issuer's cards sweep the scores, the third one is skipped in favour of the next-best alternative — concentration risk with a single bank is a real cost the raw numbers don't capture.
The assumptions — stated, not hidden
Every reward calculation depends on what a point is worth, and that's where most card content quietly cheats. Our discipline:
- —Each card carries an explicit point value (in paise per point) drawn from the issuer’s own redemption rates — and every reward percentage shows that assumption alongside it.
- —Estimated values are shown in ₹ per year, never in raw points, so cards with inflated point counts can’t look better than they are.
- —Where transfer partners can stretch value further, we show the realistic base value first and the optimised ceiling separately — not the ceiling dressed up as the default.
- —All card data is sourced from issuer MITC documents and official fee schedules, and each card page shows when it was last verified. Stale data says so.
What can never change a ranking
Money. No issuer, bank, or network pays us for placement — and there is no mechanism in the engine to accept it. Rankings come out of the formula above.
Relationships. We hold no inventory of "partner cards". The same code path scores every card in the database.
Commissions. The site currently earns no affiliate income at all (see How we make money). If that ever changes, commission status will still have zero weight in scoring.
Check our work
Found a number that doesn't match the issuer's current terms? Tell us and we'll fix it — that feedback loop is part of the methodology too.