Dynamic Couponing: One Customer, One Offer

Walk into any cashback aggregator site at 11am on a Friday. You'll find every "FLAT20" and "BIGSALE" code from every major retailer in India, neatly catalogued, valid for anyone who clicks. The codes were meant for loyalty members, or for a specific cohort, or for a 48-hour campaign window. Within hours, they ended up in the open. The merchant pays the discount to everyone. The intended cohort never feels special. Attribution is impossible.

This is the bulk coupon problem. It is solved, completely, by dynamic couponing — unique, single-use, member-bound codes generated at scale by the loyalty platform. Solved by the architecture, not solved by trying harder.

The economics of leakage

Consider a mid-size retail brand running a 10-day promotion: code "MARCH20" gives 20% off, intended for 1.5 lakh top-tier loyalty members. Day 1, the code leaks to a couponing site. By day 3, it has 80,000 redemptions, most by non-members. Days 4-10, attribution is impossible — finance can't tell what the campaign drove versus what would have happened anyway.

Margins on the leaked redemptions are gone. The intended cohort's special treatment is no longer special — every walk-in had the same code. The 20% discount becomes a permanent floor for that period, and the next month the team has to discount deeper to drive any movement. The brand learns nothing about what would have driven incremental revenue.

Dynamic couponing — what it actually is

Dynamic couponing has three architectural properties:

1. Unique per member — a 1.5 lakh-member campaign generates 1.5 lakh distinct codes, each bound to a specific member ID. "FUNDLE-7HKQ2" works only for the member to whom it was issued.
2. Validated in real time — at POS or web checkout, the code is verified against the issuing member and the redemption rules. A leaked code fails instantly.
3. AI-sized — the discount level, offer type, and expiry are not uniform across the cohort. AI picks per member based on predicted lift, churn risk, basket affinity and margin tolerance.

The third property is what makes this more than a fraud-prevention play. Once you can issue 1.5 lakh distinct offers, you stop running one campaign and start running 1.5 lakh micro-campaigns. The economics change underneath the strategy.

A model to think about offer sizing

Most brands size discounts by gut — "let's do 20% off, that always works". The correct way to size is per-member, based on expected incremental margin. The model:

The same 20% blanket discount would over-spend on Champions (who would buy at 0%), under-spend on Hibernating (who need 30% to come back), and miss the cross-sell entirely on Loyals. Dynamic couponing makes the model executable; without it, you're back to "20% off, blast it"

The architecture that makes it work

Code generation

A modern dynamic-coupon engine should be able to mint 10 million codes in under 5 minutes. Codes are cryptographically random, member-bound at issue, and stored with metadata (member ID, offer type, value, expiry, channel of issue). Brand-prefix support keeps codes recognisable ("FNDL-7HKQ2") without sacrificing uniqueness.

Real-time validation at POS / checkout

The harder problem is real-time validation at the point of redemption. Fundle pre-integrates with 50+ Indian POS systems (Pine Labs, Ezetap, Mosambee, Innoviti, Ginesys, LS Retail, GOFRUGAL, Wondersoft) and ecommerce platforms (Shopify, WooCommerce, Magento). At checkout, the cashier or system enters the code; the platform validates member ID match, redemption rules, expiry, and stacking constraints in under 300ms. If validation fails, the cashier sees a clear reason.

Channel delivery

Codes are delivered through whatever channel the member prefers and the AI arbitrates: WhatsApp template (with the code embedded), RCS rich card with one-tap copy, SMS with branded short URL, app push, email, printed on the receipt as a QR. Each delivery is logged with attribution back to the issuing campaign.

Fraud and abuse — what to actually worry about

Dynamic couponing eliminates the worst category of coupon fraud (code resharing on aggregator sites) but introduces a new class: member-account abuse. If a single account can earn unique high-value coupons repeatedly, sophisticated abusers will create many accounts. Defences:

• Per-member velocity limits on coupon issue and redemption
• Cross-account device fingerprinting (same device, multiple new accounts → flag)
• IP / network heuristics (cluster of accounts behind one IP)
• Bill-image OCR uniqueness check (same bill not credited twice across accounts)
• Employee-misuse detection (POS cashier credentials flagging unusual patterns)
• Anomaly model trained on legitimate vs abusive patterns

Fundle's fraud module ships with these defences enabled by default. The right way to think about coupon fraud is the same way fintech thinks about transaction fraud: not whether it exists, but what fraction your platform catches automatically.

Attribution: the quiet superpower

Because every code is unique, every redemption maps perfectly back to the issuing member, channel, campaign, send-time, and offer variant. This is the foundation for proper incrementality measurement.

Combined with an automatic propensity-matched control group (Fundle holds out 5-10% of every targeted cohort by default), you can answer the only question the CFO cares about: did this campaign drive incremental revenue, or would those sales have happened anyway? Bulk codes can never answer this question; dynamic couponing answers it for free.

A 90-day playbook

1. Week 1-2: Audit your current promotional calendar. Identify the 3-5 campaigns that drive the most volume. These are your starting use cases.
2. Week 3-4: Deploy dynamic couponing on a single campaign. Generate unique codes, validate at POS, measure leakage (should drop to ~0) and redemption rate vs prior baseline.
3. Week 5-8: Add AI offer sizing. Configure per-segment offer rules; let the model pick the optimal offer per member from a slate of variants.
4. Week 9-12: Move all promotional campaigns to dynamic. Decommission bulk codes entirely. Report incremental revenue per campaign to the executive team for the first time.

Three deeper architectural choices that matter

Code format — short, branded, OCR-safe

Code structure is small detail with large operational consequences. Codes that are too long ("FNDL-7HKQ29832-VALID-MAR-2026") are mistyped at POS. Codes that are too short ("ABCD") collide and feel arbitrary. The sweet spot: 8-10 characters, brand prefix + 6 random alphanumerics ("FNDL-7HKQ2P"). Critically: exclude visually-confusable characters (0/O, 1/I/l). When members read a code off a paper receipt or a screenshot, the difference between 99% and 99.9% legibility is the difference between a clean redemption and an angry support call.

Stacking rules — explicit, not implicit

Dynamic coupons coexist with other promotions: store-level sales, category discounts, tier benefits, member-tier multipliers, festival-week promotions. Stacking rules — what can combine with what, in what order — are usually undocumented in the platform and informal in operations. The result: edge cases at POS where the cashier and the customer disagree on the final price. Best practice: explicit stacking rules per coupon type, configured in the platform, surfaced to the cashier at validation time ("This code stacks with tier discount but not with store sale"). The clarity prevents 80% of POS disputes.

Expiry and reminder economics

A coupon issued without an expiry is a liability you cannot model. Expiries shorter than 7 days create urgency but miss the natural shopping cadence of low-frequency categories. Expiries longer than 30 days lose intent — the recipient forgets. The right pattern: 14-21 day expiry by default, with a reminder push at day 10 ("3 days left to use your offer"). Reminders lift redemption 18-30% versus the same coupon without reminders. The compounding effect of dynamic delivery + reminder cadence is what separates 12% baseline redemption from 28% optimised redemption.

FAQs