Does loyalty spending buy behaviour or subsidise it?

Judge a loyalty programme by additional contribution after rewards and operating costs, while separating the behaviour it changes from purchases members would have made anyway.

Members spend more than non-members. The loyalty manager presents the comparison as proof that the programme works. Finance asks whether the company’s best customers simply had more reason to join.

Both observations can be true: members may be valuable, and the programme may create value. The comparison alone does not tell you how much additional behaviour the rewards caused.

Begin the economic review by defining the behaviour the programme is meant to change. Without that definition, points become a permanent expense attached to every purchase the business can identify.

Choose the commercial job

A programme might aim to increase visit frequency, retain suitable customers, encourage a profitable product mix or reduce service friction. Each job needs a different measure and potentially a different incentive.

Increasing purchase value is not automatically beneficial if the reward shifts demand toward lower-margin products. More visits can also be unhelpful if customers merely split the same spending into additional transactions.

Write the target as a behaviour and an economic consequence. For example, encourage a second purchase from customers whose first experience was successful, provided the additional contribution exceeds the cost of the offer.

That is a testable proposition. “Make customers more loyal” is an aspiration that can absorb almost any result.

Separate three kinds of spending

First, identify rewards attached to purchases that would probably occur anyway. They may still provide relationship value, but they are not automatically an acquisition or retention gain.

Second, identify additional behaviour plausibly encouraged by the programme. This is where a causal evaluation matters, because ordinary repeat purchasing can look like programme impact.

Third, identify timing shifts. An offer can bring next month’s purchase into this month without increasing the total over a longer period. The apparent uplift may reverse after the promotion ends.

Keep the observation window long enough to examine these effects where practical. A short redemption campaign can require a longer commercial follow-up.

Work through the contribution equation

Consider an illustrative retailer with 1,000 eligible customers. A planned comparison estimates 120 additional completed orders during the agreed period. Each produces €25 contribution before rewards and programme costs.

That creates €3,000 of estimated additional contribution. If rewards across the eligible programme cost €2,200 and incremental administration costs €500, the remaining contribution is €300 under these assumptions.

The example is hypothetical and ignores no costs intentionally hidden elsewhere; a real model must include all relevant incremental costs and uncertainty. If the order estimate is imprecise, show a range rather than treating €300 as a guaranteed result.

Also test the effect of returns and substitution. A reward may encourage a purchase that replaces a more profitable purchase the customer would otherwise make.

Understand redemption instead of maximising breakage

Unused rewards can reduce immediate cost, but a programme that customers cannot understand or use may provide little reason to remain engaged. Low redemption is not automatically a success.

Research summarised by the American Marketing Association shows that point exchange-rate design can affect redemption behaviour. The practical question is whether your rules create understandable value for the intended customer.

Ask why rewards remain unused: lack of awareness, an unattractive benefit, difficult thresholds, inconvenient timing or simple lack of need. Those explanations imply different changes.

Finance should also handle the accounting treatment of earned rewards under the applicable standards. Marketing’s operating model should not assume that an unredeemed promise has no economic consequence.

Build a credible comparison

Where feasible, test a specific offer among eligible customers using planned assignment and a fixed follow-up period. Compare contribution after costs, not only redemption or purchase frequency.

GeoLift’s methodology illustrates the need to assess detectability before geographic experiments. The same planning discipline applies to choosing any suitable test: available data and sample size must support the intended decision.

If a full programme cannot be withheld, narrower tests of benefits, timing or communication may still be useful. Be clear that the result concerns the tested change, not the entire historical value of loyalty.

When relying on observational comparisons, account for obvious customer differences and state the remaining uncertainty. A sophisticated adjustment does not automatically remove every selection effect.

Keep the operating costs visible

Track support contacts, disputed rewards, fulfilment problems and technology costs. These can rise as a programme becomes more complex, even while headline engagement improves. Include staff time spent explaining exceptions at the point of service.

Look separately at customer groups. A programme may be useful overall but inefficient for customers who already purchase frequently or harmful for customers who misunderstand the conditions.

The CMO should leave the review with a specific decision: retain the design, adjust a benefit, simplify rules, test another behaviour or stop an uneconomic element. The purpose is not to prove that loyalty is always worthwhile. It is to identify which promises create a relationship the customer values and a business result the company can afford.