Your ROAS is lying. And systematically so.

Two accounts, same number, completely different reality

A ROAS of 4 can be excellent. Or worthless. The same number, in the same advertising account, can mean real growth or a self-congratulatory statistic. The difference is almost never in the campaign. It lies in attribution. And that is a setting, not a truth.

What attribution windows really measure

By default, Meta counts purchases within 7 days of a click and 1 day of a view. The second part is the exciting one: view-through attribution credits the ad with purchases from people who only saw it. No click, no measurable interaction, yet the revenue still counts. For strong brands with a large existing customer base, a significant portion of the reported revenue can consist of purchases that would have happened even without the ad.

In every audit, the first thing we check is the attribution composition. Accounts where the majority of revenue comes from view-through attribution usually don't tell a performance story, but a reach story. The crucial question is always the same: would this purchase have happened even without the ad? This is the question of incrementality, and it separates real growth from nice optics.

Why Meta and Google never show the same numbers

Each platform attributes successes to itself, according to its own rules and its own windows. GA4 with last-click logic systematically disadvantages social because social generates demand instead of just capturing it. Comparing Meta in 7-day click-plus-view with Google in last-click is like comparing two different sports and wondering about the different results.

The comparison only becomes clean with a uniform interpretation: same window, same data source, same period. A deliberately conservative 1-day click view on Meta often shows a completely different, more honest picture than the default view. It shows what the platform demonstrably triggers directly.

Without clean data, the algorithm decides blindly

Attribution is not just reporting. It is fuel for optimization. Browser signals are becoming increasingly fragmented due to tracking prevention and consent banners. Anyone who only measures with the pixel provides Meta with an incomplete picture, and the system optimizes for precisely this incomplete picture. The Conversions API closes this gap on the server side. The Event Match Quality shows you how well your signals are actually arriving. Both are part of the basic equipment, not the optional extras.

How to measure honestly

First: only compare identical measurements, same account, same window, same period. Second: use control metrics beyond the platform ROAS, such as MER or POAS on an overall level and the share of new customers. Third: check the attribution composition regularly, not just once. In the end, a ROAS is only as good as the question you ask of it.