Platform Retain Loyalty Platform

Reward the behaviour that grows tomorrow.

The best loyalty programme pays for behaviour change, not for purchases that would have happened anyway. Earn, burn, tiers and referrals on a ledger finance can reconcile, with AI choosing who gets what, and when.

  • Finance-grade points ledger
  • Rewards tested against holdouts
  • Fraud and abuse protection built in
A woman holding a gift box and her phone, smiling
Second-order reward sentDay 21 · worth ₹150 · margin-safe

Reward decision

  1. Double points0.74
  2. Free delivery0.58
  3. No incentive0.33
Repeat rate First-time buyers who order again
Retained margin Contribution after reward cost
Active members Members earning or redeeming
Reward waste Incentives given to people who would buy anyway

Why change

Most loyalty programmes are a discount with a logo.

Points pile up, finance worries about the liability, and nobody can say whether the programme changed a single customer's behaviour.

Everyone earns the same

Rewards are not matched to the behaviour you actually want more of.

The economics are opaque

Finance cannot easily reconcile cost, liability and breakage.

Points are hard to use

Customers see a balance but rarely experience any value.

Abuse goes unnoticed

Points farming, refund loops and fake referrals leak margin quietly.

What it does

Predict, decide, deliver, validate.

A rules engine for the programme you want to run, a ledger for every point ever issued, and intelligence to spend your reward budget where it works.

01 Programme design

Earn, burn, tiers and campaigns, without a release cycle.

Configure how members earn and redeem, how tiers are reached and kept, and what counts beyond transactions: reviews, referrals, milestones. Launch a double-points weekend or a tier-fast-track in an afternoon.

  • Checkout integration, so points can be earned and burned where the order happens.
  • Experiences and perks, not only discounts: early access, free delivery, priority service.
  • Gamified missions, streaks and challenges that reward habits.
Member · Gold tierActive
4,280points available
₹428redeemable value
620expire in 30 days
Mission: 3 orders this month2 of 3 done · 500 bonus points2/3
Platinum in ₹3,200 more spendUnlocks free express delivery74%
Referral: Sana placed her first order300 points pending 14-day return windowPending
Review written for last order50 pointsPosted
Wallet shown in app, web and at checkoutIllustrative

02 Reward intelligence

Spend the reward budget where it changes something.

Uplift models estimate, per customer, how much a reward would change what they do next. The engine then chooses between no incentive, points, a perk or a discount, within the cost and margin limits finance has set.

  • Uplift, not propensity: target the persuadable, skip the sure things.
  • Churn and repeat-purchase signals timed to the moments that matter, like the second order.
  • Always-on holdouts, so reward ROI is measured, not asserted.
Second-purchase accelerator · day 2120,000 first-time buyers
GroupLikely without rewardUpliftDecision
Sure thingsHighLowNo incentive
PersuadablesMediumHighDouble points
Price-sensitiveLowMediumFree delivery
Lost causesVery lowLowNo incentive
Holdout——Control
Reward cost capped at ₹6L · margin floor respectedIllustrative

Illustrative segmentation of a fictional cohort.

03 Finance-grade ledger

Every point accounted for, from accrual to expiry.

Each earn, burn, refund, reversal and expiry is a ledger entry with a source and a timestamp. Outstanding liability, breakage and reward cost are reportable at any moment, and reconcile to the books.

  • Liability and breakage reporting, by programme, tier and merchant.
  • Refunds and cancellations handled, including partial returns and clawbacks.
  • Fraud review queue for suspicious earning and redemption patterns.
Points ledger · todayReconciled 99.98%
EntryPointsValueState
Accrued on orders+1,84,200₹18,420Posted
Redeemed at checkout−62,900₹6,290Posted
Reversed on refunds−8,150₹815Posted
Expired−12,400₹1,240Posted
Flagged: rapid earn-refund loop−3,000₹300Held
₹41.6Loutstanding liability
11.8%expected breakage
3accounts in review
Exports to finance dailyIllustrative

Agents at work

AI recommends. The ledger authorises.

Agents predict, propose and personalise. They never move value on their own say-so: deterministic policy, entitlement checks and the ledger authorise every financial action.

  • Programme health watched daily, from liability to redemption failures.
  • Abuse patterns caught early, with evidence attached for the reviewer.
  • Campaign ideas with a forecast, cost, expected uplift and the holdout design included.
Agent activity  Live
Day-21 cohort scored for uplift20,000 first-time buyers · 10% holdoutAutomated
6,140 persuadables issued double pointsWithin ₹6L campaign capAutomated
Proposed: tier fast-track for lapsing Gold membersForecast cost ₹2.1L, uplift range attachedNeeds approval
3 accounts with earn-refund loopsPoints frozen, evidence sent to reviewerHuman review
Manual credit of 50,000 points requestedAbove role limit, no approval on fileBlocked
Every step logged with inputs, reason and outcomeIllustrative

You decide what the agents may do here.

AutomatedRunning the programme

Accrual, redemption, expiry, tier moves and rewards inside an approved campaign budget.

Human-approvedChanging the economics

New earn rates, new tiers, campaign budgets, manual adjustments and releasing frozen accounts.

RestrictedHard stops

Issuing value beyond a budget, redemptions without entitlement and edits to posted ledger entries.

What's inside

A programme marketers enjoy and finance signs off.

Run it as your loyalty engine, or put the intelligence on top of the programme you already have.

Earn and burn rules

Configurable rates, expiry, campaigns and exclusions.

Tiers and privileges

Qualification, retention and downgrade logic.

Wallet and points ledger

Transaction-grade balance with full history.

Reward catalogue

Vouchers, products, perks and experiences with fulfilment.

Checkout integration

Earn and redeem in supported storefronts and at point of sale.

Referrals and advocacy

Rewards for quality referrals, reviews and milestones.

Missions and streaks

Challenges that build habits, not one-off spikes.

Uplift modelling

Customer-level estimate of what a reward would change.

Holdouts and reward ROI

Incremental profitability, with selection-bias warnings.

Liability and breakage

Reporting that reconciles to finance.

Fraud and abuse controls

Limits, pattern detection and a review workbench.

Multi-brand and partner rules

Merchant-specific earning and settlement.

How it runs

From a customer moment to measured retention.

Each step is a named, logged stage. Nothing autonomous is switched on until its data, permissions, monitoring and rollback have been checked.

  1. Identify customer and eligibility
  2. Read balance and margin
  3. Estimate retention opportunity
  4. Select reward under policy
  5. Issue or redeem through the ledger
  6. Reconcile with merchant and finance
  7. Measure incremental retention

How you'll measure it

Loyalty measured in margin, not in members.

We agree the numerator, denominator, cohort and success gate with you before anything goes live, and compare against a baseline or a holdout, never against a hunch.

  • 90-day repeat purchase rateCustomers reordering within 90 days, divided by eligible first-time buyers, against a holdout.
  • Incremental retained marginContribution profit in treatment minus control, after reward cost.
  • Active member rateMembers earning or redeeming in the period, divided by enrolled members.
  • Redemption completionSuccessful redemptions divided by redemptions started.
  • Ledger reconciliationPoint and currency obligations matched to book entries.
  • Fraud loss rateConfirmed fraudulent reward cost divided by reward value issued.

Questions

Things people ask before they start.

We already run a loyalty programme. Is this a replacement?

It can be, but it does not have to be. Many teams start by adding uplift targeting and holdout measurement to the programme they have, and move the ledger later if at all.

How do you know a reward caused the purchase?

A randomly selected holdout group receives no reward. The difference in repeat rate and margin between the two groups, after reward cost, is the incremental effect.

Can finance trust the numbers?

Every accrual, redemption, reversal and expiry is an immutable ledger entry. Liability and breakage reports are built from those entries and reconciled to your books daily.

What stops people gaming it?

Earning limits, return-window holds on points, pattern detection for farming and refund loops, and a review queue where a person decides on flagged accounts.

Composable by design

Useful alone. Sharper when connected.

It works as a standalone loyalty engine. Connected, rewards become one more thing the platform can decide, deliver and prove.

Keeps your existing stack

Integrates with your commerce platform, point of sale, CRM and finance systems. Existing member balances can be migrated or left in place.

Runs where your data lives

Small language models trained on marketing data are deployed inside your cloud or data centre, so customer data and the models reasoning over it stay under your control.

Getting started

Start with one reward, one cohort and a holdout.

Agree the economics with finance, integrate one store, and test a single intervention such as the second-purchase reward.

  1. Define economics and earning policy with finance
  2. Integrate one store, test refund and expiry paths
  3. Pilot one intervention with a holdout
  4. Extend to tiers, referrals and partners