Personalisation has a ceiling

The argument

Across 217 profiled programmes, 112 of 214 with known tier status run tiers. That is a ceiling for recognition itself, not just for personalisation. If a programme cannot sustain a public tier, it lacks the repeat purchase frequency or margin to pay for individual modelling, which costs more than a tier rule.

The sector counts make the ceiling visible. Airlines run tiers in 61 of 61 profiled programmes with known status, and hotels run tiers in 18 of 18. Retail runs tiers in 25 of 46, restaurants in 2 of 22, grocery in 0 of 18, coalition in 2 of 15, banking in 1 of 14, fuel in 0 of 12, and telecom in 2 of 7. Individual-level customisation requires finer distinctions than tiers can provide, so the sectors with zero or near-zero tier adoption are not about to leap into it. They have already hit the cheaper ceiling.

Mechanism explains the split. Airlines and hotels collect many interactions per member per year and control scarce inventory, so a tier seat or room upgrade pays for itself. Grocery and fuel have frequent transactions but thin margins, so individual offers must repay their cost on low-value baskets. Restaurants and banking have infrequent or regulated contact, which limits the behavioural data needed to tailor without seeming intrusive. Personalisation is not a software feature that removes these economics; it is an operating cost that must be recovered per member.

Even the vendor market shows the limit from the supply side. Of 87 profiled vendors, 48 publish full or partial pricing, and the rest do not. If personalisation had no ceiling, vendors would have no reason to hide pricing, because demand would be unbounded. Instead the market segments split into ecommerce-smb with 26 vendors, enterprise with 19, composable-api with 18, vertical-specialist with 17, and agency-services with 7. These segments serve different ceilings, not one infinite demand for personalisation tools.

The strongest counter-argument

Airlines and hotels prove that personalisation can be total within a sector. With 61 of 61 airlines and 18 of 18 hotels running tiers, the ceiling looks like a matter of sector maturity, not a hard limit. Grocery, fuel and banking may simply be late. If machine learning makes individual offers as cheap as tier rules, a grocery programme could personalise without sacrificing margin, and the current zero or one running tier would reflect a lag, not a wall. The counter-argument deserves weight because the cost of computation has fallen while data collection has risen, and no physical law prevents a fuel programme from tailoring offers per fill-up.

What would change our mind

We would abandon the claim if any zero-tier or one-tier sector crossed to a majority of its profiled programmes running tiers and those tiers used individual behaviour rather than demographic groups. Concretely, if grocery moved from 0 of 18 running tiers to a majority of those 18, and the tiers differentiated by individual purchase behaviour, the ceiling would be falsified. The same test applies to fuel and banking without any change in their underlying margin structure. We would also change our mind if more than one of those sectors reached majority tier adoption while keeping their current thin margins, because that would show the ceiling was organisational rather than economic.

What follows if we are right

A programme operator should spend on personalisation only where the member's repeat interaction and the programme's margin per member exceed the cost of individual modelling. For an airline or hotel, that condition often holds, so tailor the offer around scarce inventory. For a grocery or fuel programme, the condition does not hold for most members, so invest in broad price benefits and simple recognition instead of chasing one-to-one customisation. The practical consequence is to set a personalisation budget per member, not per programme, and to stop promising hyper-personalisation to every member in every sector.