The rewards engine
that shows its work.
PRISM reads 4,213 reward rules across 1,143 Indian cards — caps, exclusions, milestones and 79 transfer partners — and returns the exact card to swipe. Deterministic math. AI narration. Zero black box.
Points & Rewards Intelligence, Simulation & Maximisation
every rate, cap & exclusion as data
deterministic single pass
the full catalogue, every time
cashback floor + travel ceiling
Engine activity — last 24h
18,442 verdicts · peak 1,450/hr at 9 pm
Live inference
streaming- BigBasket · ₹1,640+₹49→ Amex MRCC · grocery multiplier14 ms
- HPCL Fuel · ₹2,000+₹65→ IDFC Power+ · fuel accelerator11 ms
- Amazon · ₹2,199+₹110→ SBI Cashback · 5% online9 ms
- IndiGo · DEL→BOM · ₹8,420+₹842→ Axis Atlas · 5x travel EDGE16 ms
- Swiggy · ₹431+₹21.6→ HDFC Infinia · 5x dining12 ms
Example verdicts on real card rules — run your own numbers in the calculator.
Five stages. One honest number.
White light goes in; your best card comes out. Every stage is inspectable — this is the exact path each rupee takes through the engine.
Your spend profile meets the card knowledge base — every card's earn rates, accelerated categories, merchant multipliers and MCC mappings, kept as structured data.
For each rupee, PRISM finds the winning rule: exclusions dominate, higher priority beats lower, and the most specific match (merchant > MCC > category) breaks ties.
Monthly and quarterly earning caps are annualised and enforced, so a '10x' headline rate never overstates what you'd actually receive after the cap bites.
Milestone bonuses trigger when your spend crosses thresholds, and quantified benefits — lounges, memberships, vouchers — are priced into the total.
Fees are subtracted (unless your spend earns the waiver), points are valued two ways — cashback floor and transfer-partner ceiling — and every card is ranked by net annual value.
The AI layer on top
PRISM's core is deterministic — the same profile always produces the same verdict, so you can trust and verify it. The AI copilot sits above: it retrieves facts from the card knowledge base, calls the engine for live computations, and answers in plain language with cited sources. Language model for the conversation, hard math for the numbers.
No black box
Every verdict is explainable: expand any card in the calculator and see exactly which rule matched each spend, where a cap clipped the earnings, and which milestones triggered. When a card would earn you nothing on rent or fuel, PRISM says “excluded” — not “trust us”.
One point, two truths
A reward point is worth ₹0.25 as statement credit and ₹1+ as a business-class seat. PRISM always shows both — the cashback floor and the transfer-partner ceiling — because the gap between them is where the game is won.
See PRISM run on your numbers
Twenty-nine spend fields in, a ranked verdict out — with every rule, cap and milestone on display.