The best loyalty metric is incrementality and almost nobody measures it

The argument

Incrementality is the only metric that isolates the causal effect of a loyalty programme. A member who redeems points, reaches a tier, or spends heavily may have done all three without the programme. Counting that behaviour overstates the programme's contribution, because the counterfactual is missing. Incrementality supplies the counterfactual by comparing members exposed to the programme with an equivalent group that is not, or by comparing periods with and without a change.

Across 217 profiled programmes, the most common tracked structural fact is tier status. Of 214 programmes with known status, 112 run tiers. That means more than half invest in a status mechanic, yet no incrementality measure appears for any of the 217. The tracked facts are sector, tier status, and whether tiers run. This is activity tracking, not effect tracking.

The sector pattern reinforces the point. Airlines and hotels run tiers in every profiled case: 61 of 61 airlines and 18 of 18 hotels. Grocery runs none of 18 and fuel none of 12. These sectors with high tier prevalence are also sectors where customers are easy to identify and where spend is frequent and partly habitual. A tier programme there mainly recognises existing high spenders; it does not prove they spend more because of the tier. The absence of incrementality measurement means the spend that would have happened anyway gets credited to the programme.

Vendors are not correcting this. Across 87 profiled vendors, 48 publish full or partial pricing. Their segments are enterprise, composable api, ecommerce smb, vertical specialist, and agency services. None of the recorded vendor facts is an incrementality measurement product or service. If the technology market for loyalty prioritised incrementality, the vendor facts would show it. They do not.

The strongest counter-argument

The strongest case against incrementality as the best metric is that it is expensive and often infeasible. A clean holdout requires a programme to deny benefits to a randomly selected group for long enough to measure a difference in behaviour. Most loyalty programmes cannot do this: members talk, customer service gets complaints, and small programmes lack enough members to reveal a true difference from random noise. A tier metric or redemption rate is cheap, immediate, and understood by management. For a programme with a modest member base, a holdout may be too small to detect a meaningful lift, and the operational risk of treating some customers worse is real. If the metric cannot be measured well, it cannot guide decisions. That is a genuine weakness, not a straw man.

What would change our mind

If at least 44 of the 217 profiled programmes, or 20 percent, published a current incrementality estimate, defined as member spend minus holdout spend over the same period, and repeated it annually, the claim would be false.

What follows if we are right

A programme manager should stop treating tier attainment, redemption rate, or revenue per member as proof the programme works. Instead, set aside a holdout or use before and after changes for every major benefit, and allocate reward spend only where the lift exceeds the cost. For the 112 programmes running tiers, this means tiers should be redesigned not to recognise high spenders but to change marginal behaviour: require a threshold that would not be met by the member's baseline, and test whether that threshold increases spend relative to a control group. Otherwise the programme pays for loyalty that already existed.