What loyalty programs are, the types that work, the ROI benchmarks, and how to decide build vs buy.
Short answer
A customer loyalty program rewards repeat purchases to lift retention and revenue. The types that work are points, tiers, paid (fee-based), referral, cashback, and coalition programs. Bain research shows a 5% lift in retention can raise profit by 25% to 95%, but the return depends on adoption and offer design far more than the technology. Most programs underperform because they reward existing behaviour instead of changing it.
A loyalty program rewards people for coming back. Points, tiers, or perks give a repeat customer a reason to choose you again instead of a rival.
The tech is the easy part. Design is the hard part. A program works when it changes behaviour, not when it hands a discount to people who were already buying. The goal is to pull an occasional customer into a regular one, lift the average basket, or turn a buyer into a referrer.
The options
Points programs
Earn and burn on spend. Simplest to launch, easiest to ignore. Best for high-frequency retail.
Tiered programs
Status unlocks perks. Works when aspiration drives the category, like beauty or travel.
Paid (fee-based) programs
Members pay for perks. Amazon Prime economics. Strong when repeat demand is already high.
Referral programs
Members bring members. The highest-intent acquisition channel you own.
Cashback programs
Money back on spend. Simple to grasp, strong for everyday purchases.
Coalition programs
A shared currency across brands. High reach, but complex to run.
Loyalty pays off through retention, not novelty. The figures below come from Bain and McKinsey. Treat them as ceilings you earn with good design, not defaults you get for launching.
| What the research measures | The finding | Source |
|---|---|---|
| Retention to profit | A 5% lift in retention can raise profit by 25% to 95% | Bain & Company |
| Active vs inactive members | Active members spend about 10% more; redeemers spend 25% more | McKinsey |
| Paid vs free programs | Paid members are 60% more likely to increase spend, vs 30% for free | McKinsey |
| The category ceiling | Amazon Prime members spend roughly 4x non-members over their lifetime | McKinsey |
One caveat runs under all of it: adoption. We've seen it first-hand on a receipt-scanning platform for Aldi Ireland and a digital engagement platform for the utility Energia: first-year adoption usually lands between 15% and 30% of active customers. Plan for that band, not the headline case studies, and treat 50%+ adoption as the exception, not the model.
Estimated incremental annual revenue: $48,000.
01 Your numbers
Five inputs, no signup. We use them to model repeat revenue against build and SaaS cost.
Share of orders from customers who have bought before.
Share of customers who will sign up and use the program.
Most mature programs see 10–25% lift on member spend.
We benchmark a custom-built app from RaftLabs against a standard SaaS loyalty platform of the same scope.
Build cost and yearly maintenance scale with the modules you pick. SaaS cost scales with add-ons and feature-tier add-ons. Numbers update live.
02 Revenue impact
Modelled against your inputs, before platform cost.
Modelled from your inputs and program benchmarks. Real-world results vary by category and offer design.
03 Build vs buy, 3-year view
| Option | Year 1 profit | Year 2 profit | Year 3 profit | Break-even (mo) | 3-yr total profit | 3-yr ROI |
|---|---|---|---|---|---|---|
| Build once with RaftLabs | $36,800 | $46,800 | $46,800 | ~2.6 | $130,400 | 1304% |
| Rent a SaaS platform | $36,000 | $36,000 | $36,000 | n/a | $108,000 | ~300% |
Build-cost model: $10,000 core plus selected modules, plus 12% annual maintenance. SaaS model: $12,000/year base plus add-ons. Ember highlight marks the better option on each metric.
Cost tracks scope. A focused MVP with one or two mechanics is a different budget from a tiered program with wallet integration and multi-store sync. We break the numbers down in two guides: the loyalty program development cost breakdown and the loyalty program ROI deep-dive.
Where it fits
Retail
Points and cashback on frequent baskets. The classic fit.
Restaurants
Visit frequency and reorder are the levers to move.
Travel and hospitality
Tiers and status drive direct bookings over OTAs.
Healthcare
Recurring treatment cycles reward gentle re-engagement.
E-commerce
Repeat purchase and referral compound online.
SaaS
Retention and expansion matter more than points.
A loyalty program is only as good as the systems feeding it. Earn events have to reach the program from wherever a purchase actually happens. That might be a POS terminal at checkout, an order webhook from Shopify or WooCommerce, an ERP or invoicing feed for B2B and trade purchasing, or a photographed receipt when there's no direct checkout to integrate against at all.
Points are a financial liability the moment they're issued, not just a UX detail. Every unearned point is money promised but not yet paid out. That's why the operational side of a program includes breakage modelling (the share of points that will never be redeemed), an expiry policy, and a liability report finance can actually read. Skip this and a program can be popular and unprofitable at the same time.
The data worth tracking goes beyond redemption counts: member LTV by cohort, program attribution against a holdout group (so you know what the program actually changed, not just what happened anyway), and churn signals from engagement drop-off. Redemption rate alone tells you almost nothing about whether the program is working.
Most of the failure modes are predictable, and predictable means preventable.
A flat earn rate that pays the same person for the shop they were always going to do subsidises loyalty the program didn't create, rather than changing behaviour. A points balance nobody checks might as well not exist, so programs that hold attention pair earn with a reason to come back between purchases, not just a static number. Earn that outruns redemption and expiry compounds quietly until finance asks where the money went. And the ROI benchmarks above are ceilings, not defaults: treating adoption as guaranteed instead of the 15-30% most programs actually see in year one is how the math stops working before launch.
For a longer look at why engagement drops off after launch, see our guide to the loyalty program engagement gap.
Rent SaaS when scope is small, integrations are standard, and you need to launch in weeks. Build custom when you need deep POS or CRM integration, unusual mechanics, or when SaaS fees outgrow a one-time build inside three years. When SaaS stops paying off, a custom loyalty platform is the move, see how that played out for Musgrave's SuperValu and Centra.
Proof
We built an AI loyalty platform for Musgrave's SuperValu and Centra that automates receipt validation with 99.9% accuracy, acquiring 1,062 users in 4 weeks.
We rebuilt Energia's loyalty rewards platform from WordPress to Next.js in 12 weeks, migrating 300K+ user records and delivering a platform that recorded 1,100 logins and 87 new registrations in its first 24 hours at 99.9% uptime.
We built a gamified mobile loyalty platform for a medical spa, letting patients earn points with built-in urgency mechanics and redeem rewards, supporting a model where lifetime value depends on returning every 3-6 months.
Sanbra Fyffe's Instantor Rewards app reached 5,000 plumber signups in 3 months, increased average order value 25%, and logged 100+ receipt uploads in the first month with a tiered loyalty system built for trade customers.
A US marketing agency replaced five fragmented marketing tools with GrowViral: 2.5x higher conversion rates, 12+ integrations, and referral campaigns their clients can embed on any website without developer support.
LoyaltyPass runs loyalty programs inside Apple Wallet and Google Wallet. Customers earn and redeem without a separate app download, and businesses get real engagement data without the friction of a native app.
Work with us
Tell us what's not working in your business. We'll find the real problem and tell you exactly what it would take to fix it.