Loyalty program: what it costs and what it must return
In short
A loyalty program is a standing offer of points, tiers or paid perks in exchange for buying again. It is a discount with a delay, so it is judged against margin: a 5% points scheme redeemed at half costs 2.5% of revenue and needs every customer to spend about 9% more to break even.
What is a loyalty program?
A loyalty program is a standing offer to give customers something back for buying again — points, a tier, a fee-paid perk — in the expectation that the promise changes what they do. It is a discount with a delay and a condition attached, and it should be judged the way you judge a discount: against margin.
That framing is missing from most of the category's writing, which describes mechanics and stops. The mechanics are the easy part. The hard part is that the offer costs a fixed share of revenue from the day it launches, while the behaviour change it is supposed to buy is uncertain, delayed, and difficult to attribute.
Types of loyalty programs
Four shapes cover almost everything in e-commerce. They differ less in mechanics than in who they ask to pay and when.
| Type | How it works | What it costs you | What it asks of the customer |
|---|---|---|---|
| Points | Earn per unit spent, redeem against a future order | A share of revenue, realised when points are spent | Patience, and the arithmetic to see the reward is small |
| Tiered | Spend thresholds unlock better terms | Concentrated on your best customers, who would return anyway | Sustained spend before anything happens |
| Paid | A fee buys ongoing benefits | The benefits, offset by the fee | Money up front, which is a real commitment |
| Value-based | Purchases fund a cause | A fixed donation per order | Nothing, which is why it rarely changes behaviour |
Points are the default, and they are the default because they are familiar rather than because they work best. Paid programs are the outlier worth attention: charging a fee selects for people who intend to buy again and makes the programme self-funding, which is the opposite of the points model, where the least engaged customers subsidise the most engaged.
What a loyalty program actually costs
Run the arithmetic before you choose mechanics, because the mechanics set the cost and the cost sets the bar.
Take the common setup: one point per unit of currency spent, one hundred points redeemable for a reward worth five. That is a 5% earn rate. Not all of it is spent, so the cost you carry is the earn rate multiplied by the share redeemed.
Then the break-even. If your contribution margin is 30% and the programme costs 2.5% of revenue, each remaining customer has to spend enough that the thinner margin still clears the old profit:
(1 + lift) × (0.30 − 0.025) = 0.30, so the lift required is 0.30 ÷ 0.275 − 1.
| Earn rate | Share redeemed | Cost of revenue | Revenue lift needed to break even |
|---|---|---|---|
| 3% | 50% | 1.5% | 5.3% |
| 5% | 50% | 2.5% | 9.1% |
| 5% | 100% | 5.0% | 20.0% |
| 10% | 50% | 5.0% | 20.0% |
A standard 5% points programme, redeemed at half, needs every customer to spend 9.1% more before it has paid for itself. That is the whole number, and it is not a number the category publishes. It also excludes the platform subscription, which sits on top and does not scale down when the programme underperforms.
Unredeemed points are not free. They are a liability you have promised to honour, and a burst of redemptions after a dormant year lands in a quarter that did not earn the revenue.
What is a loyalty platform, and what does it not do?
A loyalty platform issues points, tracks balances, renders the widget, gates tiers and fires the emails. That is genuine work and worth paying for once you have decided to run a programme.
What it does not do is tell you whether the programme is working. Platform dashboards report member revenue against non-member revenue, which is a comparison between people who opted in and people who did not — two groups that differed before the programme existed. No amount of dashboard resolves that. The only thing that does is withholding the programme from a random slice of customers and comparing, which almost no platform offers because the answer is sometimes no.
Why "members spend 3.1x more" is not evidence
The category's most repeated claim is that members who redeem rewards spend around 3.1x more annually than those who do not. It is published without a sample size, and one widely cited roundup of nineteen redemption benchmarks attaches a sample size to none of them.
Set the sourcing aside, because the deeper problem survives it. Redeeming a reward requires having already bought enough to earn one. The statistic therefore compares frequent buyers with infrequent buyers and reports, correctly, that frequent buyers buy more. It would look exactly the same if the programme changed nobody's behaviour at all.
The claim that would mean something is a comparison between customers offered the programme and comparable customers who were not. The same trap swallows loyalty campaigns, where a double-points weekend is judged by what members spent during it, and it is worth understanding how customer retention is measured before spending on a programme that promises to improve it.
When a loyalty scheme is the wrong answer
A programme is a poor fit when the required lift is implausible against your purchase cycle. If customers buy from you once every two years, a points balance expires or is forgotten before it can act, and you have bought a discount for the people who were coming back anyway.
It is also the wrong answer when the underlying problem is the product, the delivery time or the second-order experience. A programme does not fix a reason not to return; it pays people a small amount to overlook it, and the arithmetic above says how small.
The honest sequence is: measure what repeat behaviour you have now, decide what lift would justify the cost, and only then choose the mechanics. Doing it in the other order — platform first, measurement later — is how stores end up with a permanent 2.5% cost of revenue and no way to tell whether it bought anything. The wider question of what loyalty mechanics do to repeat rate is the one to settle first.
Common questions
Which type of loyalty program works best for e-commerce?
It depends on purchase frequency rather than on the mechanics. Frequent, low-value categories suit points; infrequent, high-value ones rarely justify any programme. Paid programmes suit stores with a genuine recurring benefit to sell, because the fee selects for customers who already intend to return.
How much should a loyalty program cost as a share of revenue?
Set the ceiling from your margin and the lift you believe is achievable, not from what competitors offer. The table above gives the break-even lift for common earn rates; if that lift looks implausible for your category, the earn rate is too generous whatever anyone else is doing.
Do unredeemed points save me money?
Only until they are redeemed. Outstanding points are a liability against future orders, and they tend to be spent in bursts after a promotion or an expiry warning. Budget for the full issued amount and treat low redemption as a timing effect rather than a discount.
Can a loyalty platform tell me whether my program is profitable?
It can tell you what it cost and what members spent. It cannot tell you what those members would have spent without it, which is the only comparison that answers the question. That requires withholding the programme from a random group, which is a decision about the business, not a platform feature.
Is a loyalty program worth it for a store with low repeat rates?
Usually not as a first move. A low repeat rate is more often a product, delivery or expectation problem, and a programme pays customers a small sum to overlook it. Find out why people are not returning before paying a permanent share of revenue to ask them to.
See it on your data
On a twenty-minute call we connect one ad account and one month of your orders, and work out the gap between what the platform claims and what your shop took. Profit and POAS stay blank until you enter your margin and costs per order.
Book a demo