Facebook Ads ROAS Calculator

How the number is derived

The calculation is identical to any ROAS, revenue ÷ spend, but the attribution is materially different and that difference is the whole story.

Meta's default attribution has narrowed over time and now typically credits conversions within a seven-day click and one-day view window, with modelled conversions filling gaps left by tracking restrictions.

That makes reported ROAS a partly modelled figure rather than a direct observation.

An example

$20,000 of spend reporting $54,000 of revenue is a 2.7 ROAS in the platform.

The same period's actual store revenue attributable to paid social, measured by holdout or by blended analysis, might be $41,000, a 2.05 real ROAS.

At a 55% margin, 2.7 implies $9,700 of profit and 2.05 implies $2,550. Both are profitable; only one justifies scaling aggressively.

The gap is not fraud; it is view-through credit, modelling, and conversions that would have happened anyway, but a business planning on the reported figure is planning on a number 30% too high.

The catch

View-through attribution credits an impression nobody clicked, which is defensible for brand advertising and generous for direct response.

Modelled conversions also mean the reported figure includes estimates. That is a reasonable response to signal loss and it means the number carries uncertainty the interface does not display.

What this changes

Track blended ROAS, total revenue ÷ total ad spend, alongside the platform figure. For a business where paid social is the main channel, blended is much closer to the truth.

Then run a holdout test periodically: suppress advertising in a matched region or audience for two weeks and measure the revenue difference. It is the only method that measures incrementality directly.

Why incrementality testing is worth the disruption

A holdout costs real revenue for the duration and produces the one number no attribution model can give: how much of the reported return would have happened anyway.

The results routinely surprise. Retargeting campaigns reporting spectacular ROAS frequently show low incrementality, because they reach people already intending to buy; broad prospecting reporting mediocre ROAS often shows high incrementality.

Running one a quarter and reallocating budget on the results is more valuable than any amount of in-platform optimisation, because it corrects the direction of spending rather than its efficiency. Accounts that have never run one are optimising against a measurement they have not validated.

Compare the platform's reported revenue against the store's own order data for the same period, matched by first-time customer status. The overlap is usually smaller than expected.

That comparison takes an afternoon and gives a defensible adjustment factor, which is considerably more useful than either accepting the platform figure or dismissing it.

Repeating it quarterly keeps the factor current, since attribution behaviour changes as platform and browser restrictions evolve.

The comparison is most revealing during a period when spend changed sharply, since that is when the relationship between reported and real performance is easiest to observe.

Running the comparison at the cohort level: new customers acquired in a month, against spend in that month. Is cleaner still, since it removes the timing mismatch between when advertising is paid for and when the revenue it generated is recorded.