Definition
Incremental Return on Ad Spend measures the revenue an advertising campaign caused — the sales that wouldn’t have happened without it — divided by the money spent to cause them. It’s the answer to a sharper question than standard Return on Ad Spend (ROAS) asks. Regular ROAS divides all attributed revenue by spend, and attribution has a habit of crediting sales the ad only witnessed. iROAS strips out that free-riding and counts only the lift.
The distinction lands hardest with people who were going to buy anyway. If someone already has your product in their cart, then sees a retargeting ad, then checks out, last-click attribution hands that sale to the ad. iROAS asks the harder question: would the purchase have happened without the impression? If yes, it contributes nothing to incremental return.
Disambiguation: iROAS is not a different flavor of the same measurement — it uses a different source of truth. Standard ROAS reads from an attribution model (last-click, multi-touch, or data-driven). iROAS reads from a controlled experiment or a causal model: a holdout test, a geo experiment, or Media Mix Modeling (MMM). The numbers can disagree dramatically, and when they do, the incremental figure is usually the more honest one for deciding where the next dollar goes.
Why it matters for marketing
The reason iROAS keeps gaining ground is that reported ROAS quietly overstates paid channels — especially the ones that intercept demand you’d have captured for free. Brand search and retargeting are the usual suspects: they post gaudy ROAS because they sit close to the purchase, catching conversions other channels (or plain organic intent) actually generated. Optimize a budget on reported ROAS alone and you pour money into channels that look efficient because they’re standing in the right doorway, not because they’re creating demand.
iROAS reframes the budget conversation around cause and effect. It pairs naturally with Incrementality testing and feeds cleaner decisions about scaling, cutting, or reallocating spend — and it connects to the deeper unit economics a business actually runs on, like Customer Acquisition Cost (CAC) and Customer Lifetime Value (CLV). The catch is that it’s harder to measure. You can’t read it off a dashboard; you have to run an experiment or build a model, and that friction is why plenty of teams still lean on the easier, rosier number.
See also: Return on Ad Spend (ROAS) · Incrementality · Holdout Campaign · Media Mix Modeling (MMM)
How to calculate
The formula looks like ROAS but swaps in incremental revenue:
iROAS = Incremental Revenue / Ad Spend
Where incremental revenue is the outcome for an exposed group minus the outcome for an unexposed control:
Incremental Revenue = Exposed Group Revenue − Control Group Revenue
The whole thing hinges on getting a trustworthy control — a comparable group that didn’t see the ad. There are a few standard ways to build one:
- Holdout / ghost-ad tests. A slice of the target audience is deliberately withheld from the campaign (or served a placebo public-service ad), and the difference in outcomes between exposed and held-out users is the lift.
- Geo experiments. Some regions run the campaign and matched regions don’t, and the gap between them estimates incremental effect at the market level.
- Media Mix Modeling. A statistical model attributes outcomes across channels and controls for baseline demand, seasonality, and other drivers to isolate each channel’s marginal contribution.
A worked example: a campaign spends $50,000. The exposed group generates $400,000; a matched control generates $300,000. Incremental revenue is $100,000, so iROAS is 2.0 — two dollars of caused revenue per dollar spent. Standard ROAS on the same campaign, counting all $400,000 as attributed, would report 8.0. Same campaign, wildly different story.
How to utilize iROAS
iROAS earns its keep in budget decisions where being wrong is expensive.
- Deciding whether to scale a channel. A channel with high reported ROAS but low iROAS is mostly harvesting existing demand. Scaling it won’t grow the pie the way the dashboard promises.
- Justifying upper-funnel spend. Brand and awareness channels tend to look weak on last-click ROAS and much stronger on incremental measurement, because their effect is diffuse and delayed. iROAS gives that spend a fair hearing.
- Settling attribution arguments. When teams disagree about which channel “deserves” credit, an incrementality test sidesteps the model war with an experiment.
- Cross-channel reallocation. Comparing iROAS across channels shows where the marginal dollar actually produces new revenue, which is the only comparison that matters for reallocation.
Comparison: iROAS vs. related measures
| Metric | What it measures | Source of truth | Main weakness |
|---|---|---|---|
| Incremental ROAS (iROAS) | Revenue the ad caused per dollar spent | Experiment or causal model | Harder and slower to measure |
| Return on Ad Spend (ROAS) | Attributed revenue per dollar spent | Attribution model | Overcredits demand-harvesting channels |
| Marketing Efficiency Ratio (MER) | Total revenue per total marketing dollar | Blended (all revenue / all spend) | No channel-level or causal detail |
| Media Mix Modeling (MMM) | Each channel’s modeled contribution | Statistical model | Needs lots of data; not real-time |
iROAS and MMM both aim at causation; ROAS and MER describe association. The right tool depends on how consequential the decision is and how much measurement rigor you can afford.
Best practices
- Match your control carefully. The estimate is only as good as the comparability of exposed and unexposed groups. Sloppy geo matching or a self-selected holdout produces confident nonsense.
- Size tests for real signal. Incrementality lives in the difference between two groups, so small or noisy tests can’t detect modest lifts. Power the experiment before you run it.
- Run incrementality where the stakes justify it. Testing has real cost, including the revenue you forgo by holding customers out. Reserve it for big-budget channels and decisions that will move money.
- Expect the honest number to be lower. iROAS almost always comes in below reported ROAS. That’s the point, not a problem — it’s the inflation being removed.
- Re-test periodically. Incrementality isn’t a constant. It drifts as audiences saturate, creative fatigues, and competitors move. A result from last year isn’t a fact this year.
Future trends
Two forces are pushing iROAS from a specialist technique toward a default. Privacy changes and signal loss have made deterministic, click-based attribution less reliable, and experiment-based and modeled measurement degrade more gracefully in a world with fewer identifiers. At the same time, platforms and independent tools have made holdout tests, geo experiments, and lift studies far easier to run than they were even a couple of years ago.
The likely direction is a blended stack: MMM for the strategic, top-down view; incrementality experiments to calibrate and validate it; and attribution kept for fast, tactical optimization with the understanding that it’s directional, not gospel. iROAS is the metric that ties those together, because it’s the one that answers the question a CFO actually asks — what did the spend cause?
FAQs
What’s the difference between ROAS and iROAS? Standard ROAS divides all attributed revenue by spend. iROAS divides only incremental revenue — the sales the ad caused — by spend. iROAS is almost always the lower, more honest number.
How do you measure incremental revenue? By comparing an exposed group to a matched control that didn’t see the ad. Common methods are holdout/ghost-ad tests, geo experiments, and Media Mix Modeling. Incremental revenue is exposed-group revenue minus control-group revenue.
Why is my reported ROAS so much higher than my iROAS? Because attribution credits your ad with sales it merely witnessed, especially on demand-harvesting channels like brand search and retargeting. iROAS removes those, so the gap is the inflation.
Which channels usually have the biggest ROAS-to-iROAS gap? Brand search and retargeting, because they sit close to the purchase and intercept demand generated elsewhere. Upper-funnel channels often show the reverse: modest ROAS, stronger incrementality.
Is a higher iROAS always better? For measuring efficiency of caused revenue, yes — but a channel can have high iROAS at low volume. Read it alongside scale and total incremental revenue, not in isolation.
Do I need to hold out customers to measure it? For a true experiment, yes — a holdout or geo control is what makes the lift causal. Media Mix Modeling estimates incrementality without a live holdout, but it needs substantial historical data.
How often should I re-measure iROAS? Periodically, because incrementality drifts with audience saturation, creative fatigue, and competition. Treat a single test as a snapshot, not a permanent constant.
Can I optimize campaigns to iROAS in real time? Not really — it’s slower and experiment-based. Most teams optimize tactically on attribution and use iROAS to set strategy, validate channels, and correct the attribution picture.
Related Terms
- Return on Ad Spend (ROAS)
- Incrementality
- Holdout Campaign
- Media Mix Modeling (MMM)
- Multi-Touch Attribution (MTA)
- Marketing Efficiency Ratio (MER)
- Return on Investment (ROI)
- Customer Acquisition Cost (CAC)
- Conversion Value Optimization (CVO)
- View-Through Conversion (VTC)
Sources
- Google — Measure incrementality with experiments (Think with Google): https://www.thinkwithgoogle.com/marketing-strategies/data-and-measurement/incrementality-testing/
- Meta — About conversion lift / incrementality: https://www.facebook.com/business/help/399156503623312
- Interactive Advertising Bureau (IAB) — measurement resources: https://www.iab.com/guidelines/