Polo Brain — how the AI ranks cards by real value, not headline rate

Polo Brain is the CardPolo feature you'll use most. Type a question in plain English — "Best card for my ₹400 Swiggy order?" or "Should I pay this $1,400 hotel on Sapphire Reserve or Amex Platinum?" — and you get back a specific card, the exact rupee/dollar value you'll earn, the reasoning, and a runner-up.

This post is what Polo Brain is doing under the hood, and why its recommendations often surprise people.

The problem Polo Brain solves

Every cardholder has stood at a checkout, three cards in their wallet, and asked themselves: which one of these earns the most here?

The honest answer is "I don't know" — because the math involves:

• Per-category reward rate for each card (not the headline rate, the specific category rate for THIS merchant)
• Monthly cap status — has the bonus category already maxed out this month?
• Forex markup if the merchant is international
• MCC exclusion check — does this merchant earn the bonus, or is it carved out?
• Point redemption value — at what rate will these points actually convert to rupees?
• Caveats — does the merchant qualify for the issuer's portal (Smartbuy, etc.)?

That's 6 variables × 3 cards = 18 things to check at the checkout. Nobody does it. So everyone defaults to whatever card is on top in their wallet — and loses 1-3% of every transaction to suboptimal swipes.

The Polo Brain workflow

Each query runs through five stages:

1. Parse the query

Polo extracts:

• Merchant ("Swiggy" → known as dining)
• Category (inferred or stated)
• Amount (in INR or USD)
• Forex flag (USD merchants are forex)

Where the merchant is ambiguous ("dinner out" instead of "Swiggy"), Polo defaults to a generic category rate without merchant-specific portal bonuses.

2. Load your cards

For signed-in users: pull from your saved wallet (the tracker feature). For anonymous users: use the demo wallet (HDFC Infinia + Axis Magnus + SBI Cashback).

Each card carries:

• Per-category reward rates (joined from card_category_rewards)
• Annual fee, welcome bonus details
• Forex markup percentage
• Reward redemption value (the "₹ per point" conversion)
• Monthly caps where applicable

3. Compute expected value per card

For each card in the user's wallet:

raw_earn = transaction_amount × per_category_rate / 100
local_value = raw_earn × redemption_value      (points → rupees)
forex_adjusted = local_value − (transaction × forex_markup/100)  if international
final_value = forex_adjusted − cap_haircut     (if cap is likely exhausted)

Where cap_haircut represents the realistic reduction if you've already used a chunk of this month's cap.

4. Pick the winner

Rank cards by final_value descending. Pick #1 as the recommendation, #2 as the runner-up.

5. Generate the reasoning

This is where Polo Brain's depth shows up. The reasoning explains:

• The exact rupee earned with the math
• Why the winner beats the runner-up (specifically — "10X SmartBuy doesn't apply because Swiggy is HDFC's portal only, not via the SmartBuy partnership")
• Any warnings: cap status, forex markup applied, MCC exclusions

The surprises that come up

Polo Brain frequently recommends cards that aren't the highest headline rate, because the math says so. Three common surprises:

Surprise #1 — SBI Cashback beats Infinia on a Swiggy order

A user asks: "Best card for my ₹2,500 Swiggy dinner?" Their wallet has HDFC Infinia (10X on Smartbuy = 10% headline), Axis Magnus (24X dining = 24% headline), SBI Cashback (5% online = 5% headline).

Polo Brain's answer: SBI Cashback, ₹125 earned.

Why? Infinia's 10X is only on Smartbuy partners — Swiggy is not one. So Infinia earns its base 3.3X = ₹82.5. Magnus's 24X is on Indian dining merchant category — Swiggy IS in this MCC, BUT the Magnus 24X is capped at ₹2L total monthly spend, and the per-rupee earn after the 0.20 redemption value is 4.8%. SBI's flat 5% on online (Swiggy qualifies as online order) lands ₹125 cleanly.

The "highest headline rate" loses because of the redemption math.

Surprise #2 — A no-fee card beats a premium card on small transactions

For a ₹400 Swiggy order:

• Magnus earns ₹19 (₹400 × 4.8% net)
• Infinia earns ₹13 (₹400 × 3.3%)
• SBI Cashback earns ₹20 (₹400 × 5%)

SBI wins by ₹1 on small orders. Polo Brain will tell you exactly that — "On orders under ₹500, SBI's flat rate edges out the premium cards by a few rupees. For ₹2,000+ orders, the gap widens."

Surprise #3 — The premium card wins only on large transactions

For a ₹50,000 international flight:

• Magnus 24X (uncapped on travel) = ₹2,400 net (5% net after redemption haircut)
• Infinia 10X Smartbuy = ₹3,500 if booked via Smartbuy travel portal, capped at ₹15,000 point bucket
• Chase Sapphire Reserve in the US wallet = $52 (3.25X UR × $0.015 redemption)

Polo Brain recommends Infinia for ₹50K Smartbuy flight bookings — but ONLY if booked via Smartbuy. If booked directly with the airline, Magnus wins.

The eligibility gate (Polo Match-specific)

When you take the Polo Match quiz, Polo applies a hard income-eligibility gate before ranking cards. The system prompt is explicit:

"The user's annual income (compute from income_bracket) MUST meet or exceed the card's income_eligibility_local. If income_eligibility > the bracket max, the card is INELIGIBLE. Note it in skipped_card_ids — never recommend it."

A user in the 'mid' income bracket (₹6L-15L) cannot be recommended an HDFC Infinia (₹40L+ income gate) even if Infinia would otherwise be the highest-net-value pick. Polo will surface the next best card the user actually qualifies for AND tell them why they didn't get Infinia.

This is the difference between "best card in the catalog" (vanity) and "best card you'll actually get approved for" (useful).

Where Polo Brain still gets it wrong

We are upfront about the limits:

1. Promotional offers. Polo doesn't track time-limited promotions (e.g. "10% off + 10X on Amazon Pay Days"). These are caught by user-side awareness, not the AI.
2. Hidden MCC mappings. When Swiggy bills as "SWIGGY" it earns dining; when it bills as "BUNDL TECHNOLOGIES" (their corporate entity), some cards categorize it differently. Polo errs on the user-favorable side; verify on a small purchase before assuming.
3. Bank-specific portal pricing. Smartbuy gift cards sometimes price higher than Amazon for the same merchant. The 10X earn is offset by the gift-card premium. Polo doesn't catch this.
4. Stale offer data. Reward rates and welcome bonuses change quarterly. Polo's catalog refreshes on a quarterly cadence; for the absolute latest, verify on the issuer's site before applying.

How to use Polo Brain well

1. Always pass the amount. "Best card for my Amazon order" is generic; "Best card for my ₹12,000 Amazon order" gets cap-aware reasoning.
2. Name the merchant. "Dinner" infers dining; "Swiggy" applies Swiggy-specific portal logic.
3. Mention currency. "My $400 United flight" tells Polo it is forex; if you say "₹35,000 United flight" Polo treats it as INR booking on an Indian portal.
4. Read the reasoning, not just the answer. The reasoning tells you WHY this beats the alternatives. That's where the learning compounds — over time you'll start to know which card wins for which category.

Try it

Polo Brain runs in demo mode without signin (HDFC Infinia, Axis Magnus, SBI Cashback). For YOUR wallet, sign up and add your cards via the tracker — Polo will then run against your actual portfolio.

10 free queries per day. Unlimited on Pro.

Related reading

• Net annual value vs headline rate
• Transfer partner sweet spots
• Card stacking 101