Blog
How to get started with retail media as a brand advertiser
Learn more
Loading Insights...
ROAS tells you what your campaigns earned.
Incremental ROAS tells you what they actually caused.
For advertisers trying to prove that media spend is working, the gap between those two numbers is where the real answer sits. One number can make a campaign look like a strong performer, while the other shows whether it grew the business at all.
This post explains the difference in plain terms, shows how the two can diverge on the same campaign, and walks through how to start measuring the metric that holds up to scrutiny.
ROAS, or return on ad spend, is the revenue your advertising generates divided by what you spent to generate it. If you spend £10,000 and the campaign is credited with £100,000 in sales, that is a 10:1 ROAS. It is quick to calculate and easy to report, which is why most teams start here.
The problem is that ROAS counts every sale the campaign touched, including sales that would have happened anyway. A loyal customer who was always going to buy, a shopper who found you through organic search, someone who already had the product in their basket: standard ROAS often takes credit for all of them, and for word-of-mouth and other activity it had nothing to do with.
Incremental ROAS, or iROAS, strips that out. It measures only the additional revenue caused by the advertising, the uplift that would not have happened without it. You work it out by comparing a group of people who saw your ads against a comparable group who did not, then looking at the difference in their behaviour. That difference is the lift your media actually created. The shift is from measuring profitability to proving that advertising drives real growth.
Picture that same 10:1 campaign. You run a test that holds the ads back from a comparable group of customers, and that group still buys half as much as the people who saw the ads. That means half the revenue credited to the campaign would have arrived without it. Your real, advertising-driven return is 5:1, not 10:1. The campaign is still profitable, but it is doing half the work the headline number claimed. That gap is what iROAS is built to expose.
As third-party cookies fall away and consent rules tighten across EMEA, this distinction matters more, not less. The easy proxies that once propped up ROAS reporting are getting less reliable, so advertisers need measurement that rests on proper test design rather than tracking trails
When ROAS over-credits a campaign, the real cost is downstream: you start making budget decisions on the wrong number.
Channels that reach people who are already close to buying tend to post the highest ROAS, so they attract more budget. Judged on iROAS, some of that spend turns out to be paying to reach customers who would have converted anyway.
The opposite happens higher up the funnel. Channels such as display and online video, often look weaker on ROAS, because the sales they drive arrive later and through other channels. They can be creating genuinely new demand while a last-click view of ROAS hands the credit to whatever the customer clicked last.
iROAS is what lets you see that contribution and fund it with confidence, rather than cutting it because the surface number looked low.
Measuring incrementality means running a controlled experiment rather than reading a dashboard. The basic steps:
Geographic tests (advertising in some regions and holding others back) and audience holdouts are the two most common ways to build that control group. The method matters less than the principle: no control group, no incrementality.
A test is only as good as the control group behind it. A few things separate a result you can take to the board from one that only looks rigorous.
No. ROAS is still a fast, useful read on whether a campaign is broadly profitable, and it is fine for day-to-day reporting where you are not making a major budget call. The point is to know what it can and cannot tell you. Use ROAS to keep an eye on efficiency, and reach for iROAS when the question is bigger: is this spend growing the business, or just taking credit for sales we would have had anyway?
Most advertisers do not need to rebuild their entire measurement set-up to get going. Start with one campaign where the stakes are high enough to justify a proper test, and where you already suspect ROAS is flattering the result. Brand search and retargeting are common places to look, because they tend to reach people who were already close to buying.
This is the work Epsilon is built around. Our measurement runs on CORE Identity, our identity resolution layer, which ties advertising exposure to real purchases at transaction level, so the test and control groups reflect actual people rather than fragmented IDs. That is what lets us report incremental outcomes a brand can trust. When i-Run took this approach, over half its customers were recognised from day one and the programme reached a 3:1 incremental ROAS within four months.
ROAS and iROAS are one of three shifts that separate metrics which prove performance from ones that just flatter it.
Our Digital Metrics Guide walks through all three, covering reach versus unique reach and clicks versus cost per acquisition alongside this one, step-by-step.
