When Can Sales Lift Be Called Incremental?

Incremental-sales measurement is the subject here: how to decide whether a change in product, brand, or category sales can be attributed to in-store media rather than merely observed during the campaign. The focus is the choice among before-and-after variance, test-versus-control, multivariate, and shopper-level approaches. The practical judgment is to match the strength and cost of the design to the claim, while making its comparison, timing, coverage, and assumptions visible.

This question comes after the outcome has been named. A featured-product purchase, brand halo, or category movement may be commercially relevant, but it is not automatically incremental. Incrementality asks a counterfactual question: what would the comparable shoppers, stores, or products have done if the media had not been present? No design observes that alternate world directly. The purpose of a measurement design is to construct a credible comparison or model for it.

A Sales Change Is Not Yet an Increment

The simplest approach is before-and-after comparison. A retailer measures sales before the campaign, during it, and perhaps after it, then describes the difference as a media-driven change. The IAB calls this sales variance and says it is a presumed media-driven delta. It recommends using the method only when incremental-sales studies are not possible.

The weakness is not that time-series movement is useless. It is that many forces move with time. A product may be entering a seasonal demand period. A price promotion may begin with the media. Store traffic may change because of weather, holidays, or local events. A competitor may be out of stock. The retailer may change shelf placement or assortment. Ordinary variation may be mistaken for a treatment effect. Before-and-after data can show that the period differed; by themselves, they do not show that the media caused the difference.

The first stronger alternative is test versus control, also called split testing. One group receives the treatment—such as media exposure—and another comparable group does not. The difference in behavior is called lift and can be examined through measures such as order value, purchase frequency, or survey-based perceptions. The central requirement is comparability: the groups should differ in the treatment being studied, not in all the other factors that influence behavior.

Random assignment is ideal because it reduces systematic differences between exposed and unexposed shoppers. When random selection is not possible, the IAB recommends intentionally designing groups with similar distributions of relevant characteristics. At the store level, it says test and control locations should be matched on pre-period sales, other in-store activity, prices and promotions, store size, and store format. The details are not administrative decoration. If the test stores ran a discount that control stores did not, a higher sales result has two possible explanations. If test stores were larger or had a different product range, the comparison begins with a structural difference.

Matching is not cosmetic. It is an attempt to make the untreated group stand in for the outcome the treated group would otherwise have produced. If the groups differ in store format, baseline demand, pricing, or concurrent activity, the measured difference contains both the media contrast and those structural differences. Random assignment can reduce this problem when it is feasible.