No sales team would ever pay commission on a deal that was already signed. That’s how so many tiered loyalty programmes work though: the biggest rewards go to the customer who was going to buy anyway.

Look at who’s in your top tier right now. They’re definitely valuable customers. But a tier that’s built on spend just hands out the biggest rewards to the people who needed the least persuasion. They were coming back anyway. The programme was just paying them for it.

It looks like a win on all the reports. The top tier members spend more so the tier must be working. They spent more before they were top tier too. The tier didn’t create that behaviour. It just found it, put a badge on it and started to discount it.

Upside’s 2026 ‘Escaping the Loyalty Plateau‘ research (summarized by eMarketer in July 2026) found that across grocery/fuel/convenience and restaurants, increasing overall spend was the last of the ways in which loyalty programs were found to impact members. It also identified a ‘brand first’ segment (20% of restaurant consumers and 10% of grocery shoppers) who would redeem rewards regularly but would probably be shopping with the brand anyway.

That’s the real margin risk in tiered loyalty: the rewards are going to be on moments that would have happened anyway.

Why did spend become the default way to build loyalty tiers?

The vast majority of marketing leaders were operating under the same playbook: points per dollar, tiers defined on trailing 12 month spend and an annual requalification cliff. That was the model that made sense at the time. Spend was the one number that every system and every finance function could agree on and you could draw your tier lines at spend percentiles in the space of an afternoon.

Tiers still work in the same way. What’s changed is everything else around them.

Membership has saturated. The same Upside research shows the average consumer is in 17.4 loyalty programmes but actively using just 8.4 of them so joining doesn’t in any way indicate commitment. Measurement has come a long way towards marketing. Holdout groups, propensity scores and real time triggers which used to require a full blown analytics project can now be run by the majority of enterprise lifecycle teams. ‘What would this member have done without the reward?’ is finally a question that can be answered.

None of these changes in themselves is dramatic. Within an enterprise programme they compound into something much more difficult to see.

Why does the margin leak grow at enterprise scale?

A small program a generous top tier is a rounding error. At enterprise scale the three forces all come into play.

Sorting. A spend threshold is a filter for customers who already spend the most. Their behavior is also the most predictable in the file, and the least likely to change because of a reward. Every benefit added at the top routes more margin to the members least sensitive to it.

Flattering measurement. Most program reporting compares members with non-members, or top tiers with lower ones. Those groups were different before the program touched them, so self-selection gets reported as lift and the program “proves” itself every year.

Split ownership. In our view, this is what lets the first two persist. Marketing owns the benefits, finance owns the liability, and merchandising owns the margin. Each can do its job well while nobody is accountable for incremental margin, the one number that would expose the problem.

Designing loyalty tiers around incremental behavior

Here’s the reframe: a tier is a contract to pay for a specific change in behavior, priced against the margin that change creates. Spend can still count toward qualification. It just can’t be the whole test.

The math is simple. If half the customers you reward would have bought anyway, each reward really costs double, because you’re paying two people for every one you persuaded. If three-quarters would have bought anyway, it costs four times as much.

Model each member’s baseline first. Work out what each active member is likely to spend and do in the next period with no intervention at all, using frequency, recency, category breadth and tenure. A decent propensity model, calibrated against a holdout is all that’s required to get started.

Give each tier one behaviour to buy. Spend is an outcome; the behaviours are what drive it. Think of a second purchase in the 60 day window, a first purchase in a new category, app adoption or a move to full price. Let your own cohort tell you which of those behaviours are the best predictors of lifetime value and then pick one per tier.

Qualify on lift not level. Progression has to be tied to movement against the member’s own baseline with a spend floor underneath. A member predicted at $400 who spends $650 has lifted. A member predicted at $3,000 who spends $3,000 has not lifted at all, despite clearing every threshold.

Price rewards on contribution margin not on revenue. A reward of 6% of revenue on a product line which has a 30% contribution margin is effectively handing back a fifth of the profit on every order before you even get to the purchases that would have been made anyway. Hold the hard benefits for the behaviours you’re trying to change. For the high, stable spenders, use recognition (early access, service priority, visible status etc). They need appreciation, not persuasion.

Hold out permanently. Take a small randomised group out of each tier benefit and major journey and read the results over a full purchase cycle. Retirement of any benefit whose lift doesn’t cover its cost is the way to go (just as you would with an underperforming paid channel).

A tier that buys a behaviour still needs the right messenger and every channel prices that job differently.

Which channels should be used to carry the message of each tier?

Once every tier has a target behaviour then channel choice stops being a habit and becomes an economic decision. The metric that matters is cost per incremental outcome i.e. which channel is going to change the target behavior in the most cost effective way without wearing the member out.

Channel	Cost per message	Best used for	Watch out for
Email	Very low	Tier progress, category-expansion nudges, top-tier recognition	Each send looks free, so volume creeps up
Push	Very low; opted-in devices only	Second-purchase and expiry nudges	One tap to opt out, rarely reversed
SMS	Highest of the four	Requalification alerts, limited top-tier access	The most expensive place to subsidize baseline spend; gate it tightly
In-app	Near zero; active app users only	Benefit discovery, app adoption as a tier behavior	Reaches the already engaged, so it’s easy to over-credit

Making the program prove its margin

The principle at work here is simple. A loyalty program is an offer to every member, paid for in margin. An offer which changes behavior is an investment. An offer which changes nothing is just a discount with a logo on it.

Three moves follow from that.

• Make incremental margin the program’s headline number. Not member count, redemption rate, or member-versus-non-member spend. We expect finance teams to start asking loyalty for the same proof of incrementality they already demand from paid media, so bring them into tier design now, not just the liability review.
• Fund the holdout before the next benefit. A permanent control group gives up some attributed revenue, and someone will call that a cost. It’s far cheaper than another year of paying for demand you already had.
• Change the logic before the logo. New qualification rules, behavior-based paths, and rebalanced top-tier benefits can all sit underneath your existing tier structure. Members see more relevant offers; finance sees margin move the right way.

This is the work that Mavlers’ lifecycle and CRM teams do in conjunction with the loyalty and marketing platforms that the enterprise already runs: they link up the tier events, the holdouts and the channel journeys so that incremental margin is actually reported rather than just assumed.

Tiered programs crept into the subsidization of habits simply because spend was easy to measure and every report would say they worked. The solution starts with one question for every tier, every benefit and every message: what did it actually change?

A tier should be the price you pay for a change you can see. Everything else is a discount your best customers never asked for.

Summing it up

Fixing margin leakage doesn’t require a new loyalty program. It means rebuilding each tier on four things: a predicted baseline for every member, one behavior per tier worth paying for, rewards priced in contribution margin, and a permanent holdout that shows whether any of it worked. Spend still counts, but as a floor for qualification rather than the reason for the reward.

Tiered programs drifted into paying for customers’ existing habits because spend was easy to measure and every report said the program worked. The fix starts with asking one question of every tier, benefit and message: what did it actually change?

A tier should be the price you pay for a change you can see. Anything else is a discount your best customers never asked for.