Marketing Atlas · Reference · Attribution
Updated May 2026 · Reference route · written diagnostic
Return on Ad Spend (ad-attributed revenue / ad spend) as the headline number on agency reports. The metric that hides waste behind a clean ratio.
The numbers underneath
The shift this concept produces
Before and after the operator applies the discipline named here. Source: SC install benchmarks across categories, 2024-2025.
Section 01 · Quick definition
In one read
ROAS reporting is the convention of using Return on Ad Spend (revenue attributed to ads divided by spend on those ads) as the lead metric for evaluating a paid-media program. The number is clean, easy to compute, and easy to compare across campaigns.
The structural read
It is also routinely platform-reported, double-counted across channels, gross of margin, and silent on incrementality. A 4x ROAS report can be produced by a program that is bidding on the operator's own brand terms, paying for conversions that would have happened anyway, and losing money on every truly new customer.
Section 02 · Why it matters
Origin.
ROAS is the most-cited metric in operator-side paid media and the one most likely to be wrong about the question the operator actually has. The question the operator cares about is “is this program making me money on net new customers, after margin and after fulfillment.” The number ROAS answers is “what is the ratio of platform-attributed revenue to the spend the platform recorded.” The two questions can produce wildly different answers under the same data.
Mechanic.
The structural risk is that ROAS rises when the program does more harvesting (branded search, retargeting, last-touch on email) and falls when the program does more prospecting (cold audiences, top-of-funnel display, broad targeting). An operator who hires an agency to maximize ROAS is paying the agency to harvest more aggressively from the existing demand baseline, which is exactly what the agency will do.
The practical stake is that ROAS can rise while net new customer count falls, and the report will read as a strong month every time. The fix is to read ROAS alongside new customer count, contribution margin per customer, and blended cost per acquired customer.
Section 03 · How it runs
ROAS at its simplest is ad-attributed revenue divided by ad spend, expressed as a multiple. The calculation depends entirely on which revenue counts as ad-attributed and which spend counts as ad spend. Each platform answers both questions in its own favor. Google Ads attributes revenue to a click within a configurable lookback window of up to 90 days. Meta attributes within its own click and view windows. The two platforms can both claim the same conversion, producing channel-level ROAS that double-counts at the operator's level.
Each platform reports its own ROAS using its own attribution model, its own conversion window, and its own definition of ad-attributed revenue. The numbers are accurate within the platform's view of the world. Across channels, the same conversion is often counted twice. Adding channel-level ROAS does not produce blended ROAS.
Blended ROAS is total platform revenue divided by total ad spend across all channels. It is the operator-side view and is closer to what the bank statement implies. It does not solve incrementality (some of the revenue would have come without ads) but it does solve double-counting. Most agency reports do not lead with blended ROAS because blended ROAS is usually lower.
Margin-aware ROAS uses contribution margin instead of revenue: (revenue minus COGS minus fulfillment) divided by ad spend. The number is the one the CFO will accept. A 4x revenue ROAS at 30% contribution margin is a 1.2x margin ROAS, which is breakeven on direct cost and loss-making on customer acquisition cost.
Incremental ROAS measures the lift the ad program produced above what would have happened without it. Measured cleanly with geo-holdout tests or scaled-spend pauses. Almost never reported by agencies because it almost always shows lower numbers and because the test design takes work to set up. Incremental ROAS is the number that survives.
The shift this concept names
Before applying this concept
“ROAS went up, so the campaign got better.”
After applying this concept
Incremental ROAS measures the lift the ad program produced above what would have happened without it. Measured cleanly with geo-holdout tests or scaled-spend pauses. Almost never reported by agencies because it almost always shows lower numbers and because the test design take...
Section 04 · Common misunderstandings
Misunderstanding 01
“ROAS went up, so the campaign got better.”
ROAS goes up when the program targets cheaper conversions, which is usually existing demand the operator already had: branded search, returning customers, retargeting. Acquiring a new customer from a cold audience almost always lowers ROAS in the short term. A rising ROAS often means the program retreated into the demand baseline, not that it acquired more efficiently.
Misunderstanding 02
“Channel ROAS adds up to total ROAS.”
Channel ROAS does not add up. Each platform claims credit for conversions on its own terms, often double-counting cross-platform paths. Blended ROAS computed against total revenue and total spend is the only number that maps to the bank statement.
Misunderstanding 03
“A higher ROAS target is more disciplined.”
A higher ROAS target restricts the program to harvesting closer to the buy intent. In paid search, that means bidding more aggressively on branded terms. In paid social, it means retargeting site visitors. Both produce strong ROAS reports and both fail to acquire new customers. Discipline is hitting a margin-aware blended ROAS target, not maximizing platform ROAS.
Misunderstanding 04
“ROAS is the right metric because the CFO understands it.”
CFOs understand contribution margin, customer acquisition cost, and payback period. ROAS is a marketing translation of those concepts that drops the margin and the customer count. The translation is always lossy. Reporting marketing performance to the CFO works better with the CFO's native metrics than with the marketing translation.
Misunderstanding 05
“Incrementality testing is too expensive to run.”
Geo-holdout tests cost lost revenue in the holdout cells and a few weeks of analyst time. The cost is real and is far smaller than the cost of running an entire program against a non-incremental ROAS target for two years. The tests answer a question that ROAS cannot answer.
Section 05 · Diagnostic questions
What is the difference between channel-level ROAS and blended ROAS for the last twelve months, and what does the gap imply about double-counting?
What is the difference between channel-level ROAS and blended ROAS for the last twelve months, and what does the gap imply about double-counting?
What is the contribution margin per acquired customer, and what is the margin-aware ROAS at that margin?
What share of attributed revenue comes from branded search and retargeting versus cold audiences and prospecting?
Has the program ever been geo-holdout tested or scaled-spend paused to measure incrementality?
If branded search were paused for thirty days, what fraction of that revenue would still occur through organic and direct?
Is new customer count rising, falling, or flat while ROAS has been reported as rising?
What ROAS target was set, and does it correspond to a contribution-margin target the CFO has signed off on?
Stan's take . four chunks
A ROAS that goes up while net new customer count goes down is the cleanest signal that an agency is harvesting branded baseline rather than building demand. The ratio rises because the program retreated into the demand the operator already had.
The customer count falls because no incremental customer is being acquired. The report reads as a strong month.
The bank account confirms the strong month because the operator's own brand is producing revenue. The agency takes credit for the brand the operator built.
Reading ROAS alongside new customer count exposes the pattern in one chart. Almost no agency report shows the two side by side. The reason it does not show them is the same reason it should.
Section 06 · Adjacent concepts
Section 07 · Sources
Official documentation on how Google Ads computes and reports ROAS, including which conversion events are included and how the lookback window applies.
Reference for how Meta calculates purchase ROAS, the click and view attribution windows used, and the difference between platform-reported and operator-side measurement.
Practitioner reference on the gap between channel-reported ROAS and operator-side blended ROAS and the implications for paid-media governance.
Industry overview of geo-holdout and scaled-spend test designs for measuring the lift a paid-media program produces above baseline.
Independent reference on the structural limitations of ROAS as a decision metric, with worked examples of margin-aware and incremental alternatives.