Loyalty Program: How to Design One That Pays Back
Most loyalty programmes are a discount you have agreed to give forever, to customers who were going to buy anyway. Here is how to build one that isn't.
Start with the uncomfortable version. A points scheme hands money back to your best customers — who are already buying. The money lands where it was least needed.
That is not an argument against running one. It is the test every programme must pass at design time and every quarter after: is this changing behaviour, or paying for behaviour that was already happening?
Answering it needs the arithmetic of what a customer is worth over time, which our guide to calculating customer lifetime value sets out. Retention as a practice is wider than this page — service, onboarding, winback, pricing — and a programme is one lever in it. This page is the programme itself: whether to have one, which mechanic, what it costs, how to prove it.
What a loyalty programme is actually for
A programme chases exactly one of three goals well. Design for two and you get neither.
| Goal | What the design has to do | Where it fails |
|---|---|---|
| More frequent purchase | Put the reward on a clock, so waiting costs something | Rewarding cumulative spend instead of rhythm — heavy buyers coast, light buyers never arrive |
| Bigger basket | Set thresholds at the next realistic step up | Thresholds only existing big spenders can clear |
| Data and identity | Near-zero enrolment friction, plus one small reason to identify yourself | Over-paying for data you never use |
The third is what most Indian businesses need and most often mis-build. Sell offline or through marketplaces and you may not know the person at the counter today bought in March.
Stitching those purchases to one identity is worth doing even if it changes no buying decision — but judge it as a data project, on a data project's budget. It is also what makes real customer segments possible.
The mechanics, and what each one costs you
Six structures cover almost everything in market. Pick for the purchase rhythm and margin you actually have.
| Mechanic | Good at | What it costs | Main failure mode |
|---|---|---|---|
| Points | Frequency in monthly-or-more categories | A fixed share of every sale, forever, plus a liability | Earn rates so slow the first reward never arrives |
| Tiers | Protecting the top of your base with status | Little, if benefits are service rather than discount | Everyone lands in the bottom tier and stops looking |
| Paid membership | Commitment — the fee itself changes behaviour | Real value owed from day one, visibly beating the fee | Only heavy buyers join, so you discount your best |
| Cashback or wallet credit | Being understood instantly; redemption happens | The most expensive per rupee of perceived value | Customers wait for credit; full-price sales thin |
| Surprise-and-delight | Goodwill without a permanent entitlement | Low and controllable — you pick moment and volume | Cannot be forecast; easy to let lapse |
| Community or access | Considered, high-margin, identity-linked buys | Mostly time and people, not margin | Dies when whoever ran it moves on |
Points and cashback give away margin on every transaction in perpetuity; the other four cost effort. That decides reversibility. You can stop surprise rewards next week; you cannot quietly stop honouring points.
The economics to settle before you launch
Four numbers decide affordability. Settle them on paper before anything is built.
- Reward cost as a share of gross margin, not revenue. A 5% cashback on a 20% gross margin is a quarter of your margin, not a twentieth of your price. Model it at full redemption — the case you have to survive.
- Breakage — and never plan on it. Breakage is the share of earned rewards nobody claims. The cheapest way to raise it is to make redemption awkward, and a programme that is hard to use is one nobody enrols in next year. Upside, never a line in the model.
- The liability it puts on your books. Under IFRS 15 and Ind AS 115 — India's converged standard, effective for financial years beginning 1 April 2018 (ICAI WIRC Ind AS 115 material, 7 July 2018) — a point giving a discount beyond what you normally offer that class of customer is a "material right", and so a separate performance obligation. Part of today's sale price is allocated to the points and is not revenue until they are redeemed or expire (IFRS Foundation TRG Agenda Ref 48, 9 November 2015, on IFRS 15 B40, B41, B43).
- Incrementality. Of the rupees given away, how many bought a purchase that would not otherwise have happened? Most programmes never establish this, which is why most cannot be defended when a new CFO asks.
Report it inside your marketing KPI framework. A programme with its own private dashboard is one nobody switches off.
The design decisions that decide whether anyone uses it
Programmes rarely fail on strategy. They fail on four boring choices.
- Time to first reward. If it takes six purchases, the programme is invisible to everyone except people who were going to make six. Land the first reward on the second or third order, even if it is small.
- Legible value. "100 points" means nothing; "₹100 off your next order" means something. If a customer cannot say what their balance is worth in rupees, you have hidden what you pay for.
- Expiry. It limits liability and creates urgency, and it draws the commonest complaint in Indian schemes. Keep the window reachable, warn before it bites, never shorten it retroactively.
- Enrolment friction. Every field, app install and password taxes the goal. If identity is what you are buying, buy it cheaply: one identifier, one tap.
A quieter fifth: who owns the programme on a Tuesday. Without someone accountable for its monthly numbers it drifts into a discount nobody reviews. We scope programme design and measurement inside a wider marketing engagement.
When not to run a loyalty programme
Four situations where the honest answer is no. Decide before you build — a live programme is hard to withdraw.
| Situation | Why a programme fails here | Spend the money on this instead |
|---|---|---|
| Low purchase frequency | A mattress, a solar install, a wedding service — no rhythm to accelerate, and rewards expire before the next window opens | Referral and after-sales service |
| Thin margins | A 3% reward on a 10% gross margin is nearly a third of it, before platform and support costs | Basket and mix — attachments, bundles, cost to serve |
| Price is the only differentiator | It reads as a price cut, a competitor matches it in a week, and you have both lowered price for nothing | A non-price difference worth paying for |
| An unfixed product or service problem | Nobody leaves for want of points. A reward is a bribe to tolerate late delivery or unanswered support | Fix the defect, then service recovery |
The alternative is not doing nothing. It is the same money on a better experience — faster resolution, a human on the phone, a replacement sent before the customer asks twice. That budget reaches people who had a bad experience. A points scheme sends its to the already happy.
Why the published uplift numbers are not planning inputs
You will meet figures like "members spend 67% more" in every loyalty platform's marketing. Do not plan against them: we could not trace the circulating uplift numbers to a primary study with a named author and date, and the sourced ones are vendor research on self-selected samples.
The selection-bias reason is the genuinely useful point, and it is simple.
- People choose whether to enrol, and those who do expect to buy again — they self-select on the exact behaviour being measured.
- So members were heavier buyers before the programme existed; the comparison measures that difference.
- The gap appears even if the programme does nothing, so the number cannot tell a working one from a useless one.
- Published cases are survivors — programmes quietly shut down are not written up.
How to tell whether it worked
A programme with no holdout group cannot be evaluated. Worth saying bluntly, because almost none has one.
- Hold back a control group at launch. Before enrolling anyone, randomly withhold the programme from 5–10% of eligible customers. They buy as normal and never see the offer. The gap in revenue per customer over a full purchase cycle is the contribution.
- Already launched? Stage the next holdout. Roll the next change — a new tier, a bigger reward, a reactivation push — to a random subset. The original launch cannot be measured now.
- Read it as cohorts, not monthly totals. Group customers by enrolment month and follow each forward; a working programme shows later cohorts retaining better at the same age (our cohort analysis guide has the method).
- Treat redemption rate as a health check, not a result. High redemption means rewards are reachable and understood; it says nothing about whether anyone bought more. A programme can hit 90% redemption and still be margin leakage.
From our own work: the first thing we look for in an existing programme is a holdout group, and in the audits we have run we have almost never found one — so the client has paid for years with no way to say what it returned. We are not putting a figure on that; we have not counted it systematically.
Give any change a full purchase cycle before judging it — a quarter for a monthly category, nearer nine months for a quarterly one. Retention work takes three to six months to move; a programme is slower.
Loyalty programmes in India: what actually changes
Four things differ enough here to change the design, not just the copy.
- Enrol on a phone number, not an app install. A number is the one identifier almost every customer will give at a counter or at checkout, and it is the same identifier online and offline — which is what makes identity-stitching possible. An app install asks for a download and an account before any value arrives.
- The default expectation of a reward is money. UPI and cashback wallets set the reference point: rupees, instantly, no ceremony. Points converting at an unclear rate into a conditional voucher read as worse than cashback even when worth more. Choose a non-cash reward and the burden is on you to make its rupee value obvious.
- Offline redemption has to work at the counter. For kirana, pharmacy, salon and QSR businesses it happens in ten seconds, queue behind the customer, often on a shared staff device. If it needs an app open, a code read aloud or a manager's approval, staff stop offering it and the programme dies at the till.
- Enrolment data is regulated. That phone number is personal data. The DPDP Act 2023 is in force: the Digital Personal Data Protection Rules, 2025 were notified on 14 November 2025 with an eighteen-month phased compliance window, and every Data Fiduciary must issue a clear, plain-language consent notice stating the specific purpose for which data is collected (Press Information Bureau explainer, 17 November 2025). Section 5 covers the notice, Section 6 requires withdrawal to be as easy as consent, Section 8 erasure once the purpose is served (Act text via Laws as Code, NeGD — last updated 12 July 2026). So keep the consent record, give a working opt-out, and do not reuse a loyalty list as a general marketing list. Our DPDP Act guide for marketers has the detail.
Where to go from here
- Name the single goal — frequency, basket or identity.
- Price it against gross margin at full redemption.
- Check the four no-go conditions honestly.
- Set the holdout before launch day — impossible later.
- Pick the smallest mechanic that could work.
Frequently asked questions
Is a loyalty programme worth it for a small business?
Only if customers buy often enough for a reward to be reachable and your gross margin can absorb the giveaway at full redemption. Start with the simplest mechanic — a service-benefit tier, or occasional surprise rewards — which cost effort, not permanent margin.
Points, tiers or cashback — which should I choose?
Match the mechanic to purchase rhythm and margin. Points suit monthly-or-more categories with margin to spare. Tiers protect a valuable top segment with status rather than cash. Cashback is the most instantly understood in India, the most expensive, and it trains people to wait.
What is breakage, and should I build it into my business case?
Breakage is the share of earned rewards never claimed, and it makes a programme look cheaper than it is. Do not model it: the easiest way to raise breakage is to make redemption awkward, and a programme nobody can redeem is one nobody joins next year.
Do loyalty points create a liability on my accounts?
Usually yes. Under IFRS 15 and Ind AS 115, a point granting a discount beyond what you normally give that class of customer is a material right, and so a separate performance obligation — part of the sale price is allocated to it and is not revenue until redemption or expiry.
How do I prove my loyalty programme actually increased sales?
Withhold the programme from a random 5–10% of eligible customers at launch and compare revenue per customer against enrolled customers over a full purchase cycle. That difference is the real contribution. Without a holdout you are only measuring that heavier buyers enrolled.
Find out whether it would change anything
We model the reward cost against your real margin, set the holdout, and say honestly if the money belongs elsewhere.
Related guides
Make Digital Hangover a preferred source
One tap tells Google to show more of our SEO and marketing coverage in your Top Stories.