Incrementality

Advertising works – or at least it looks that way. Someone sees an ad for sneakers, clicks on it, and buys. From the advertiser’s perspective, that looks like a success. But maybe this person would have bought the sneakers anyway, completely without the ad. Incrementality answers exactly this question: How much did the advertising actually achieve – and what would have happened anyway without it? The term stands for the true effect of a measure, minus everything that would have occurred regardless. It originally comes from economics and today plays a central role in digital marketing and AI-driven ad measurement.

Why the real effect is hard to detect

The problem is called causality: just because two things occur together doesn’t mean one caused the other. Someone who catches the flu in winter while also drinking tea doesn’t get sick because of the tea. Likewise, not every ad that precedes a purchase is the cause of that purchase. Many people buy products they wanted anyway – and happen to click on an ad beforehand.

Classic metrics like click-through rates or conversions don’t capture this difference. They simply count all purchases that occurred after an ad click. That sounds helpful, but it’s misleading: a company that only targets people who were going to buy anyway sees great numbers – and still spends money on nothing.

How an incrementality test is structured

The cleanest method is a controlled experiment. A target audience is randomly split into two groups. The test group sees the ad. The control group – also called the holdout group – doesn’t see it. At the end, the purchase rates of both groups are compared. The difference is the incremental effect: the share of purchases that the ad actually triggered.

The principle comes from medical research, where it’s known as a randomized controlled trial. In the advertising industry, the control group is often called a ghost group or holdout, and the evaluation is now frequently handled by automated systems. AI models help cleanly separate the groups, identify confounding variables, and statistically validate the results – that is, check whether the measured effect isn’t simply due to chance.

An important benchmark here is the so-called ROAS – Return on Ad Spend, i.e., the ratio of advertising revenue to advertising cost. Incrementality tests often show that the true ROAS is significantly worse than the apparent one, because many of the counted purchases would have happened even without advertising.

Incrementality in practice: where the term shows up

In everyday life, incrementality is encountered mainly in reports about digital advertising platforms. Meta, Google, and other providers have developed their own tools that let advertisers test whether their campaigns deliver real added value. Interest in this has grown since classic tracking methods – such as tracking via browser cookies – have become more difficult due to privacy laws and Apple updates.

The term also appears regularly in news coverage about marketing technology. Companies that used to simply count clicks and conversions are now looking for more reliable methods. Incrementality is considered the more honest standard here – because it doesn’t ask what happened after the ad, but what happened because of it.

A common misconception, by the way, is confusing incrementality with reach. A campaign can reach millions of people and still have an incremental effect of nearly zero – if all those people would have bought anyway. Conversely, a small, precisely targeted campaign with few touchpoints can achieve high incrementality because it reaches exactly the right people.