ROAS is return on ad spend: attributed revenue divided by spend, reported by the advertising platform itself. Because every platform claims every conversion it can plausibly see, the same order is often counted by two or three of them, which makes platform ROAS unusable as a single source of truth.
That is not fraud, it is what attribution windows do. Each platform answers honestly from inside its own window, and the windows overlap, so the sum of everyone's claimed revenue is routinely more than the store took.
Worked through on the example store the calculators use: the platforms between them report €78.000 of purchases on €15.000 of spend, a ROAS of 5,20. The store's own books show €60.000 of net revenue for the same month, a MER of 4,00. Part of the gap is VAT and shipping inside the purchase value a platform counts, and the rest is the same orders claimed by more than one platform.
Keep it, but demote it: it is a relative signal for comparing two creatives inside one account on one day, and nothing more. The moment it is used to judge the whole budget, the overlap starts flattering whichever platform claims fastest, and that race is structural rather than a settings problem.
The one fair way to hold a platform's ROAS against your margins is break-even ROAS: the lowest ROAS that platform can report before its ads lose money, worked out from what an order leaves after goods, shipping and payment fees, in the same purchase value the platform counts. On the example store's €85 order that floor is 2,26, and Insights, Products gives every product its own.
nouz reads no platform's attributed revenue at all: the ad connectors fetch spend only, so no claim can leak into your statement. The blended alternatives, and when each one is the right tool, are set out in MER, POAS or ROAS.
