ROAS has become the dominant metric of performance marketing. If it's above target, you scale. If it's below, you cut. It's simple, universally understood, and consistently misleading in ways that cost brands real money.
That's not an argument against using ROAS. It's an argument for understanding what it's actually measuring, and what it's not.
Blended ROAS inflates results by taking credit for organic conversions
Problem 1: Attribution Overcounting
Your reported ROAS includes conversions that would have happened without the ads.
When someone who was already planning to buy visits your site, abandons their cart, sees your retargeting ad three times, and then completes the purchase on day four, your retargeting campaign gets credit for that sale. In last-click attribution, it gets 100% of the credit. In data-driven attribution, it still gets substantial credit because it was a recent, observable touchpoint.
But the attribution system cannot tell you that this person was going to buy anyway. It can only observe that an ad appeared before a conversion occurred. The counterfactual, would they have bought without the ad, is invisible to every attribution model.
The degree of overcounting depends heavily on the channel and audience. Retargeting campaigns targeting recent site visitors often show reported ROAS that's 3–5x the actual incremental ROAS, because organic purchase intent in that audience is already high. Prospecting campaigns targeting cold audiences have less overcounting because organic intent is genuinely low.
Without running incrementality tests to measure the gap, you don't know how much of your reported ROAS reflects advertising that actually caused conversions versus advertising that showed up to people already in the purchase funnel.
Problem 2: The Platform Boundary Problem
Each platform calculates ROAS from within its own data silo.
Facebook's ROAS is calculated from conversions Facebook can attribute to Facebook, within Facebook's attribution window, using Facebook's observable data. Google's ROAS is the same, for Google. TikTok the same, for TikTok.
When a customer sees a Facebook ad on Monday, reads a review article on Tuesday, watches a YouTube video on Wednesday, clicks a Google Search ad on Thursday, and converts, Facebook might claim that conversion (7-day click window). Google will definitely claim it (last click). The customer's actual revenue is $80. Your combined platform reporting shows $160 in attributed revenue from a single order.
This is not an edge case. It's the structural reality of running multi-platform advertising. When you add up all your platform-reported ROAS figures, the total implied revenue is virtually always higher than your actual revenue, sometimes by 50%, sometimes by 200%.
Any comparison of ROAS across platforms, "Facebook ROAS is 4x vs. Google ROAS is 6x, so we should shift budget to Google", is comparing numbers that are measured by different methods, in different data bubbles, with different attribution windows. This is why platform attribution is structurally self-serving. It's not an apples-to-apples comparison.
Problem 3: ROAS Optimizes for Efficiency, Not Growth
A 6x ROAS is not always better than a 3x ROAS.
A 6x ROAS at $10,000 spend is $60,000 in attributed revenue. A 3x ROAS at $100,000 spend is $300,000 in attributed revenue. Efficiency and scale are different objectives, and optimizing for efficiency alone can systematically limit growth.
This matters in practice because the tactics that produce the best ROAS, retargeting warm audiences, bidding on branded search, running bottom-of-funnel campaigns, are typically the tactics with the lowest incremental value. You're maximizing efficiency by spending on people who were already going to convert.
The tactics that drive actual business growth, prospecting new audiences, building brand awareness, reaching people who don't know you yet, typically produce lower ROAS because organic purchase intent is low. The attributed numbers look bad. But the incremental impact can be substantially higher.
A marketing organization that is evaluated purely on blended ROAS targets will systematically underinvest in growth channels and overinvest in harvest channels. The ROAS looks great until the prospecting funnel depletes and the warm audiences are exhausted.
Problem 4: ROAS Ignores Profit Margins
A 4x ROAS on a product with 20% gross margin is a 0.8x profit ROAS, meaning the advertising is generating less in gross profit than it costs. The campaign is unprofitable, even with a 4x ROAS that sounds strong.
A 2x ROAS on a product with 65% gross margin is a 1.3x profit ROAS, generating positive contribution after ad spend and cost of goods.
Most ROAS targets in performance marketing are set without reference to gross margin, and the resulting targets are either too high (for high-margin products that could profitably scale at lower ROAS) or too low (for low-margin products where even a strong ROAS barely covers costs).
The metric that actually matters is contribution margin per acquisition, gross profit minus ad cost, but this requires knowing product-level margins and syncing that data with your advertising measurement. Few brands do this well.
What to Track Instead (or Alongside)
iROAS (incremental ROAS): Measures the return on ad spend from conversions that advertising actually caused. Requires incrementality testing (holdout tests or geo experiments) to measure. More accurate than ROAS for evaluating whether a channel is genuinely worth the spend. Not available without running experiments.
MER (Marketing Efficiency Ratio): Total revenue divided by total ad spend, using your actual revenue from your backend system rather than platform-attributed revenue. Blended, directional, doesn't tell you which channel is working, but it tells you whether your overall marketing investment is generating growth. Harder to game than ROAS because it uses real revenue.
Contribution margin per channel: (Revenue × gross margin) minus ad spend, by channel. Requires margin data at the product or order level. More complex to build but the most honest picture of whether advertising is generating profit.
The Organizational Problem
The deeper issue with ROAS isn't measurement, it's incentives.
When performance marketing teams are evaluated on ROAS targets, they rationally optimize for ROAS. They double down on retargeting and branded search (high ROAS, low incrementality). They pull back from prospecting and brand channels (low ROAS, high incrementality). Over time, the warm audience depletes, new customer acquisition slows, and the ROAS metrics that looked healthy turn out to have been measuring the exploitation of existing demand rather than the creation of new demand.
The fix requires leadership that understands the difference between attributed ROAS and incremental growth, and that evaluates marketing on the latter. Marketing organizations that measure success by MER trends, new customer acquisition, and incremental ROAS from periodic tests tend to make better long-run decisions than organizations that live and die by platform-reported ROAS.
That's not a measurement problem. That's a management problem. But it starts with understanding why the ROAS number in your dashboard is telling you a story that's only partially true.