Why your ROAS doesn't match your bank account

Is the platform counting sales it cannot see?

It is not, and that is the first gap. Any sale whose report never arrived is missing from the platform's numerator, so its ROAS is understated. This makes the reported figure pessimistic, and unevenly so: campaigns whose buyers block more are penalised more.

Is the same sale being counted twice?

It can be, and this one runs the other way. If a browser report and a server report of one sale carry different event identifiers, the platform sees two sales. Reported revenue then exceeds real revenue, and the campaign looks twice as good as it was.

Are you comparing different windows?

Almost certainly. The platform credits a sale to an ad clicked days earlier, so its month contains sales your bank recorded in a different month. Meta also removed the seven and twenty-eight day view windows in January 2026, which moved the totals again without any campaign changing.

Is ROAS the wrong measure for the question?

Often. ROAS compares revenue to ad spend and ignores cost of goods, shipping, payment fees and returns. A campaign can hold a healthy ROAS and lose money on every order, which is a real problem that no amount of better tracking will show you.

How to make the two comparable

1. Compare one day at a time, in one time zone, before comparing a month. Most reported gaps are boundary effects.
2. Fix double counting first, because it is the only cause that overstates. Check that the browser and server reports of one sale carry the same identifier.
3. Then close the reporting loss, so the platform is judging campaigns on all the sales rather than the ones that got through.
4. Keep ROAS for steering ad spend and a margin-based measure for deciding whether the business works. They answer different questions and neither replaces the other.