A 16x ROAS includes sales the brand would have made anyway. Incrementality measures only those that wouldn't have happened without the ad.
The difference seems subtle but it changes everything. When a brand sees a 16x ROAS, what it's seeing is the total sales attributed to the ad: the sales that came through the ad plus the sales that would have happened anyway, with or without advertising. It's a number that looks good on a dashboard, but it doesn't answer the question that actually matters to the person approving the budget: how much did I grow thanks to this investment?
A 16x ROAS includes sales the brand would have made anyway. Incrementality measures only those that wouldn't have happened without the ad. It's a lower number, but it's defensible.
- What it measures: attribution counts total sales with the ad; incrementality, only those that wouldn't have happened without it.
- Typical number: 16x ROAS in attribution vs 3x real incremental.
- Methodology: last-click for attribution vs control group for incrementality.
- Defensible: attribution no; incrementality yes.
How incrementality is measured
Incrementality answers exactly that question. To measure it you need a control group: a set of users comparable to the one that saw the ad, but that didn't see it. The difference in sales between the two groups is the incremental revenue — the revenue that wouldn't have existed without the campaign. It's experimentation methodology, not attribution.
- It requires designing the experiment and reserving inventory for the control group.
- You have to wait for statistical significance before reporting.
- The number is almost always lower than the traditional dashboard number.
- For a marketing team, a lower number feels like a risk.
The high number is what puts the contract at risk: when the advertiser discovers that half of those sales would have happened anyway, they lose confidence in the report and, eventually, in the platform. The lower number, on the other hand, is defensible.
How we do it at AndesML
At AndesML we report incrementality by format, brand and campaign. It's not an add-on offered to the premium client: it's the central metric of the report. Each brand can see how much impressions, clicks and sales grew thanks to the advertising campaigns, not just the total. And the retailer can use that same information to defend its rates: if the inventory generates real incremental sales, the price of the space justifies itself.
Incrementality also serves to optimize. When you know which campaigns generate real sales and which only capture sales that would have happened anyway, you can move budget toward what works and stop funding what doesn't.
It's the difference between a report that looks good and a report that renews the contract. And in a commerce media business, what renews the contract is what sustains the revenue line month after month.
The mistake of measuring with the wrong metric
The problem with reporting only attribution isn't that the number is a lie: it's that it hides the truth. A 16x ROAS tells the advertiser that every dollar invested generated 16 — but a significant portion of those sales would have happened anyway. The customer who was going to buy that product regardless bought it, and the ad simply appeared along the way. Attributing that sale to the campaign is technically correct under last-click, but strategically misleading.
When the advertiser discovers, months later, that half of the attributed sales would have happened anyway, trust doesn't recover. And without trust, there's no renewal. The high number destroys what the low number protects: the long-term relationship.
How an incrementality experiment is designed
Designing an incrementality experiment isn't trivial, but it's not dark science either. The total audience is split into two comparable groups: one that sees the campaign (treatment group) and one that doesn't (control group). The control group reserves a portion of inventory that isn't monetized during the experiment. At the end, sales from both groups are compared: the difference is the incremental revenue, the revenue that wouldn't have existed without the advertising.
- Define the control group size to reach statistical significance.
- Ensure both groups are comparable in behavior and composition.
- Keep the control isolated from the campaign throughout the experiment.
- Wait for sufficient volume before reporting a number.
Incrementality isn't a lower number that makes the business look bad: it's the number that makes the business last. It's what allows the retailer to defend its rates and the advertiser to trust they're paying for real growth, not for sales that would have happened anyway.
