Incremental return on ad spend (iROAS)
Incremental return on ad spend (iROAS) divides the incremental revenue advertising caused — the lift over a control group — by ad spend. Unlike attributed ROAS, which credits every conversion an ad touched, iROAS isolates causation using experiments such as geo holdouts or ghost-ad tests. It answers a different question: not how much revenue was attributed to ads, but how much would not have happened without them.
What this means
iROAS = incremental revenue ÷ ad spend, where incremental revenue is the difference between an exposed group and a comparable control that did not see the ads. It measures the causal lift attributable to advertising, not the total revenue that touched an ad.
Why it needs an experiment
Attribution models assign credit to ads after the fact, but cannot tell whether the buyer would have converted regardless. Incrementality requires a counterfactual — a holdout audience, a geo split, or a ghost-ad placebo — so the lift over control reveals the revenue that advertising truly added. Without a control, you cannot compute true iROAS.
- iROAS = incremental revenue ÷ ad spend
- Incremental revenue = exposed minus control group
- Requires a holdout, geo split, or ghost-ad design
Why it misleads
Experiments have noise, and a poorly designed control (contaminated audience, too small a sample) produces an unreliable lift. iROAS also varies by audience and time, so a single test is a snapshot. Treat it as evidence about causation that complements, not replaces, ongoing attributed reporting.
How it appears in analytics and logs
A low iROAS despite a high attributed ROAS means ads were taking credit for conversions that would have happened anyway — the spend is not as incremental as attribution implied.
Diagnostic use case
Use incremental ROAS, measured with a holdout or geo experiment, to test whether ad spend caused revenue rather than merely co-occurring with conversions it would have won anyway.
What WebmasterID can help detect
WebmasterID measures first-party conversions by source, supporting the holdout and exposed-group revenue comparisons that an incrementality test needs.
Common mistakes
- Calling attributed ROAS 'incremental' without a control group.
- Drawing strong conclusions from an underpowered experiment.
- Assuming one test's lift holds across audiences and seasons.
Privacy and accuracy notes
iROAS is estimated from aggregate group comparisons in an experiment, not per-person tracking. This page is educational, not legal advice.
Related pages
- Return on ad spend (ROAS)
Return on ad spend (ROAS) is the revenue attributed to advertising divided by the cost of that advertising, usually expressed as a ratio or percentage. It answers 'how much revenue did each unit of ad spend bring back'. ROAS is not ROI — it ignores product margins and other costs — and its numerator depends entirely on the attribution model, so the same campaign can show very different ROAS under different rules.
- Marketing efficiency ratio (MER)
Marketing efficiency ratio (MER) is total business revenue divided by total marketing spend over a period, across every channel at once. Unlike per-channel return on ad spend, it claims no attribution: it asks how much revenue the whole marketing budget produced, including organic and brand effects. As an industry convention it is read as a trend over time, and pairs with channel-level ROAS rather than replacing it.
- Data-driven attribution: promise and caveats
Data-driven attribution (DDA) assigns credit using a model trained on a site's own conversion paths rather than a fixed rule like last-click. Done well it credits assist touches more fairly. Its caveats are real: it needs enough conversion volume, it is a model not a measurement, and it cannot see touches that were never tracked.
- Attribution analytics
First-party outcomes for lift testing.
Sources and verification notes
- Google Ads Help — About conversion lift / experimentsLift measurement via holdouts; iROAS formula is an industry convention.
Last reviewed 2026-06-24. Facts are checked against primary/official sources where available; uncertain specifics are marked “Data not yet verified” rather than guessed.