ROAS and Its Limitations

Return on ad spend is revenue divided by ad spend. It is quoted constantly, it is the basis of most advertising decisions, and it is more fragile than its simple definition suggests.

Both numbers are estimates

The denominator is close to known. The numerator is not: it depends entirely on an attribution model that decides which sales to credit to which advert. Different models produce materially different revenue from the same campaign, and none of them is measuring the alternative where you had spent nothing. The ratio is real arithmetic on top of two estimates.

View-through credit is the largest distortion

Modern attribution includes a term for sales that happened after someone saw an advert but never clicked it. On a platform where a great deal of selling happens inside the app and the click never happens, this term can be very large. Whether those sales would have occurred anyway is unknowable from inside the data, and the standard assumption is that a portion of them would. This is why a high ROAS on one platform and a much lower one on another is not necessarily a contradiction to be explained away; it is frequently two attribution models describing the same sales differently.

What it is genuinely good for

Comparing similar things. ROAS is useful as a relative measure between campaigns with the same attribution model, the same creative category and a comparable time window. It is genuinely useful there, and it is the correct instinct to want a single number.

Where it misleads

Across time. A quarter with a promotion and a quarter without are not comparable, and neither is a quarter before and after a tracking change.

Across margin. The same return means something very different on a low-margin sale and a high-margin one, and ROAS has no way to see the difference.

Across platforms. Comparing a self-attributing network's number with a modelled one measures the models.

As a decision rule on its own. A campaign with a high ROAS and a small spend may be far less valuable than a lower one carrying most of the budget. Scaling decisions are made on absolute contribution, not on a ratio.

The habit worth building

Use ROAS to compare like with like, and use absolute numbers and margin to decide. If the ratio and the absolute contribution disagree, the absolute figure is closer to the truth, because it reflects what actually arrived.