Incremental ROAS (iROAS) measures the revenue your advertising actually caused, divided by what you spent to cause it. It removes baseline sales that would have happened without any advertising at all.
Standard ROAS, which most platforms report, divides total attributed revenue by total spend. The problem is that attributed revenue includes purchases people would have made regardless of seeing your ad. A customer who was already going to buy from you looks just the same in the data as one your ad converted.
The formula
Incremental ROAS = Incremental revenue / Advertising spend.
Incremental revenue is calculated by comparing a test group that saw your ads against a holdout group that did not. The difference in revenue between the two groups, scaled to represent the full market, is the incremental revenue your campaign caused.
Why standard ROAS misleads you
Imagine you run a retargeting campaign aimed at people who already visited your website. Many of these people intended to buy anyway. Your standard ROAS might look excellent at 8x because the platform attributes all their purchases to the ad they clicked. But your iROAS might be 2x because most would have converted without the ad.
You spent money to reach buyers who were already on their way. The campaign looked efficient but it was not creating demand. It was just taking credit for it.
How to calculate iROAS in practice
- Run a holdout test, keeping 10 to 20 percent of your audience or a set of matching regions out of the campaign
- At the end of the test, record total revenue from the test group and the holdout group
- Normalise revenue by population or baseline share to make the comparison fair
- Calculate the revenue difference as a percentage, then apply it to total revenue to estimate incremental revenue
- Divide incremental revenue by total spend during the test period
If your iROAS is significantly lower than your standard ROAS, you are likely over-investing in channels or audiences that convert regardless of your advertising. This is one of the most common and costly mistakes in digital marketing.
What is a good iROAS?
This depends on your margins and growth objectives. For most brands, an iROAS above 1.0 means the campaign is generating positive return. Whether that is "good enough" depends on your target ROAS, your contribution margin, and whether you are optimising for short-term profit or long-term customer acquisition.
The more useful question is how your iROAS compares across channels and campaigns. That comparison shows you where your budget is actually working.
iROAS vs ROAS: which should you use?
For day-to-day bidding and optimisation, standard ROAS signals are useful. For budget allocation decisions, channel investment reviews, and annual planning, you should use iROAS. These are different jobs and the right metric depends on which decision you are making.
A brand that only uses standard ROAS for budget allocation will systematically over-invest in channels that harvest existing demand and under-invest in channels that actually grow it.
Is incremental ROAS the same as marginal ROAS?
They are related but not identical. Marginal ROAS looks at the return on the next pound of spend, usually measured by incrementally increasing or decreasing budget. Incremental ROAS measures the return caused by advertising versus not advertising at all. Both matter, but they answer different questions about your spending efficiency.
Can I calculate iROAS without running a holdout test?
Marketing mix models (MMM) can estimate incremental contributions without a controlled experiment, but they are less precise. Where possible, calibrate your MMM with at least one holdout test to validate the model's channel coefficients. Without any experimental validation, you are relying entirely on statistical assumptions.
My iROAS is negative. What does that mean?
A negative iROAS means the holdout group outperformed the test group. This can happen when your ads are actually suppressing conversions, which is rare, or more commonly when the test groups were not properly matched at the start and the holdout was already a higher-value group. Review your experimental design before drawing conclusions.