Retail media platforms love to report ROAS. The numbers are frequently impressive — 20x, 30x, 40x return on ad spend is regularly cited for sponsored product campaigns on major retail networks. But before you reallocate your entire marketing budget based on these figures, it’s worth understanding what these numbers actually mean — and what they don’t.
What the Standard ROAS Metric Measures
In most retail media reporting, ROAS is calculated as total attributed sales divided by total ad spend, using a defined attribution window (typically 7–14 days after a click). A 20x ROAS means the platform is attributing $20 of sales for every dollar spent on advertising. The problem: “attributed” is not the same as “caused.” If a consumer searches for a product they were already planning to buy, clicks on a sponsored product ad, and purchases — that sale shows up as an attributed conversion. But the sale would have happened anyway. The advertising didn’t drive incrementality; it captured intent that already existed.
Why ROAS Inflation Is Structurally Built In
Sponsored product campaigns are particularly susceptible to ROAS inflation because they often target branded search terms — consumers who are already searching for a specific product or brand. These consumers have already decided what they want. This doesn’t mean sponsored products are ineffective — they serve a real role in ensuring visibility at the moment of purchase. But 40x ROAS should not be read as “every dollar generates $40 of incremental sales.” That’s a different number, and it’s almost always significantly lower.
Incrementality Is the Metric That Matters
True incrementality testing — running holdout groups where a portion of your target audience is not served advertising, and comparing outcomes — gives a more accurate picture of actual campaign effectiveness. Industry data suggests average incremental ROAS for retail media campaigns (true lift, not attributed ROAS) is in the range of 2–5x for established brands. That’s still a strong return by advertising standards — but it’s a very different story from 30x. Don’t let the measurement complexity paralyse investment. Do build a measurement framework that distinguishes attributed from incremental performance, and run incrementality tests wherever platforms allow.