The three flavours of ROAS
When marketers and finance teams argue about ROAS, they are usually arguing past each other because they mean different things by the word. There are three distinct versions, and the gap between them is where budget decisions go wrong.
Reported ROAS
What each ad platform shows in its own dashboard. Each platform uses its own pixel, its own attribution window, and counts any conversion it touched. Almost always the highest number and almost always the least reliable.
Attributed ROAS
Revenue distributed across channels using an independent multi-touch model. Reconciles double-counting and uses a consistent methodology for every channel. A much better truth than reported ROAS, and the daily operational metric most attribution platforms surface.
True incremental ROAS
The revenue that would not have happened without the ad. Strips out baseline or organic demand. The only version that answers "would we still have made this sale if we hadn't run the ad?" Typically derived from experiments or marketing mix modelling.
Why reported ROAS misleads
A customer discovers your brand on Instagram three weeks before they buy. They return via an email campaign. They search your brand on Google, click the ad, and convert.
Meta reports that Instagram delivered the sale. Your email platform reports that email delivered the sale. Google Ads reports that search delivered the sale. Three platforms. One sale. Triple counting.
This is structural, not fraudulent. Every platform has a commercial incentive to claim credit for anything it touched. A full walkthrough of this dynamic is in how to track ROAS by channel.
Incremental vs attributed: why both matter
Even well-attributed ROAS still counts revenue that would have happened anyway. The most common example is branded search. When a customer types your brand name into Google, they were going to buy. The ad you run on your own brand term simply intercepts that click. Attributed ROAS on branded search often looks excellent. Incremental ROAS on branded search is usually a fraction of that.
The practical split is this: attributed ROAS is your daily operational metric. It tells you, fairly, where channels stand relative to each other. Incremental ROAS is your strategic metric. It answers whether each channel deserves budget at all.
The same sale, three numbers
Consider a $300 sale influenced by Meta (first touch), Email (middle touch), and Branded Google Search (last click).
| Channel | Reported | Attributed (linear) | Incremental |
|---|---|---|---|
| Meta | $300 | $100 | $180 |
| $300 | $100 | $90 | |
| Branded Search | $300 | $100 | $30 |
| Total "credit" | $900 | $300 | $300 |
Reported totals $900 because each platform claims the full $300. Attributed totals the real $300, split evenly. Incremental also totals $300 but weights channels by their causal contribution: Meta introduced the customer, email nurtured, branded search merely caught the click that was already coming.
How to measure incremental ROAS
Three methods, each with different trade-offs. Most mature teams combine all three.
Holdout experiments
Randomly exclude a segment from seeing the ad and compare revenue against the exposed segment. The gold standard for causal measurement. Requires large audiences and disciplined experiment design.
Geo experiments
Pause spend in a geographic region and compare revenue trajectories. Useful for channels where user-level holdouts are impossible. Confounded by local seasonality and external events.
Marketing mix modelling
Statistically model revenue as a function of spend across channels, seasonality, and external factors. Produces incremental contribution estimates without pausing campaigns. Attriqs' MMM module runs three approaches (Bayesian, Ridge, Ensemble) entirely in a browser Web Worker.
Where MMM fits in
Multi-touch attribution answers "which channels moved this customer?" Marketing mix modelling answers "what would revenue have been without each channel?" The two are complementary:
- MTA gives you daily granularity at the campaign and keyword level.
- MMM gives you causal honesty at the channel and portfolio level.
Mature teams use MTA to tune the detail and MMM to size the big picture. The two rarely disagree about which channel is best, but they often disagree about how much budget that channel deserves.
What this means for budgets
Likely over-invested
- Branded search
- Retargeting to intent-qualified audiences
- Last-click-optimised bidding on existing customers
Likely under-invested
- Upper-funnel awareness channels
- Organic content and SEO
- Email nurture sequences
- Podcast and creator partnerships
The specifics vary by business. The pattern almost never does.
Common mistakes
Comparing reported ROAS across platforms. Each uses a different window and methodology. The comparison is apples to oranges.
Treating attributed ROAS as incremental. Attributed ROAS reconciles double-counting. It does not strip out baseline demand.
Pausing a channel without testing. Dropping branded search to "prove it wasn't incremental" can tank revenue if organic coverage is weaker than expected. Run a measured holdout or geo test first.
Ignoring time lag. Upper-funnel channels often show incremental impact weeks or months later. Short evaluation windows punish them unfairly.
Frequently asked questions
What is the difference between reported ROAS and true incremental ROAS?
Reported ROAS is the figure an ad platform shows in its dashboard based on conversions it can see. True incremental ROAS is the revenue that would not have happened without the ad. A customer who was already going to buy contributes to reported ROAS but not to incremental ROAS.
Why do ad platforms over-report ROAS?
Each platform reports every conversion it touched as its own, using its own attribution window and rules. When one customer interacts with Google Ads, Meta, and email before buying, all three platforms claim credit for the same sale. Summing platform-reported revenue typically exceeds actual revenue by 30 to 80 percent.
How much lower is incremental ROAS typically?
It varies by channel. Branded search is often dramatically over-credited, with incremental ROAS sometimes only 20 to 40 percent of reported ROAS. Upper-funnel channels are often undercredited. The gap is larger for brands with strong organic demand and for retargeting campaigns that reach already-interested customers.
Is attributed ROAS the same as incremental ROAS?
No. Attributed ROAS uses a multi-touch model to distribute credit fairly across channels but still counts revenue that would have happened anyway. Incremental ROAS adjusts for baseline or organic demand. Attributed ROAS is closer to reality than reported ROAS, and incremental ROAS is closer still.
How do you measure incremental ROAS?
Three main approaches: holdout experiments where a random audience segment sees no ad, geo experiments where a region is paused to see if revenue drops, and marketing mix modelling which statistically estimates incrementality from historical spend and revenue patterns. Attribution platforms typically combine attributed ROAS (daily) with MMM-driven incremental estimates (quarterly). Attriqs runs MMM with three statistical approaches in parallel (Bayesian, Ridge, Ensemble) so the incrementality answer is not dependent on any single model choice.
Does incremental ROAS matter for small businesses?
Yes, especially for small businesses. When budgets are tight, spending on non-incremental activity is the most expensive mistake you can make. A small business paying for branded search clicks that would have converted organically is throwing away budget that could fund new-customer acquisition.
Can incremental ROAS be negative?
Practically, incremental ROAS can be zero or near-zero when an ad serves only to already-intent customers. Technically, it is bounded at zero, because an ad cannot cause less revenue than no ad. But the profit impact can be negative if ad spend exceeds the incremental revenue it generates.