ROAS is not your return
Your ad platform grades its own homework. Incrementality, the marginal ROAS ceiling and blended MER are the three numbers that decide whether paid ads build profit.
Reported ROAS is the most trusted number in ecommerce and one of the least reliable. It is not that the platforms lie. It is that you are asking a question they cannot answer: reported ROAS tells you which sales the ad touched, and you need to know which sales the ad caused. Those two numbers are different, and the gap between them is where ad budgets quietly leak.
Three ideas close that gap. None of them requires a data team, and all three change what you do on Monday morning.
1. Incrementality: would this sale have happened anyway?
Your retargeting campaign shows an ad to someone who already has your product in their cart. They buy. The platform records a conversion and your ROAS looks superb. The honest question is whether they would have bought without seeing it. Often they would have. You paid for a sale you already had.
This is why retargeting and branded search almost always report the best ROAS in the account and almost always have the worst incrementality. They are positioned closest to a purchase that was already going to happen.
You do not need a formal study to get a usable read. The cheapest test in advertising is turning something off:
- Pick the campaign with a suspiciously high reported ROAS. Usually branded search or a retargeting set.
- Pause it completely for two weeks. Not a budget cut, a full pause, or you cannot read the result.
- Watch total store revenue, not that campaign's revenue. The campaign's number will go to zero by definition.
- If total revenue barely moves, the spend was not incremental and you have found money. If total revenue drops by roughly what the campaign was reporting, it was real, and you should scale it.
Run this on one campaign per quarter. A single honest answer about your biggest reported winner is worth more than a year of dashboard staring.
2. Marginal ROAS: every winner has a ceiling
The most common way to lose money on a working ad is to scale it. Reported ROAS is an average across all the spend in a campaign, and averages hide the thing you need to see: the return on the next dollar, which falls as you spend more.
The mechanism is simple. The platform spends your first dollars on the people most likely to buy. As you raise the budget, it has to reach further into a colder audience. Average ROAS drifts down slowly and reassuringly while the marginal return falls off a cliff.
At $1,000/day a campaign returns 4.0x, so $4,000 revenue.
At $1,500/day it returns 3.4x, so $5,100 revenue.
Average ROAS still looks healthy. But the extra $500 of spend bought $1,100 of revenue, a marginal return of 2.2x. If your break-even is 2.5x, that increase lost money while the dashboard reported a 3.4x winner.
Illustrative example. Not real campaign data.
So scale in steps and read the step, not the total. Raise budget by 20 to 30 percent, wait for enough conversions to mean something, then work out what the increase itself returned. When the marginal number crosses your break-even, you have found the ceiling. Sit at it.
3. Blended MER: the number that cannot be gamed
Per-campaign ROAS is self-reported by a party with an interest in the answer, and when several platforms all claim the same conversion, the numbers you are adding up overlap. Blended marketing efficiency ratio avoids the whole problem by refusing to attribute anything.
Total store revenue divided by total advertising spend, across every channel, for the same period.
No attribution, no windows, no platform's opinion. If MER is flat while a platform reports a new winner, the winner is moving sales around rather than adding them.
Use both. Per-campaign ROAS is fine for deciding which creative beats which creative inside one platform, because the bias applies to both sides equally. Blended MER is the number you steer the business by. When they disagree, MER is right.
What to actually do
- Set a MER target first, derived from your real contribution margin, and treat it as the goal. Per-campaign ROAS targets are a means, not the end.
- Judge creatives against each other, not against a threshold. Inside one platform the bias cancels out, so relative comparison is the honest use of reported ROAS.
- Scale in increments and read the increment. Never move a budget by a large multiple and judge it on the new average.
- Give a pause test enough time. Two weeks minimum. Anything shorter reads noise.
- Be most suspicious of your best-reported campaign. High reported ROAS and low incrementality look identical on a dashboard.
Where Wizzy comes in
The reason most stores do not run this loop is not that it is hard. It is that it is relentless: the marginal curve moves as audiences saturate, creatives fatigue on their own schedule, and a winner can become a loser inside a week. Wizzy watches spend against total revenue rather than platform-reported conversions, tracks the marginal return on each budget change so a winner stops getting scaled past its ceiling, and flags the campaigns whose reported ROAS is not showing up in your blended number.