Ads Manager reports a beautiful ROAS, but your revenue does not follow. Here is how attribution really works, and how to triangulate the truth with your own data.
Meta reports a strong ROAS, but the revenue in your bank account does not follow. That is not a bug, it is how attribution works. Every platform claims every purchase it can connect to a touchpoint, within the window you configure. The fix is not a better dashboard but triangulation: platform data for direction, your own order data for truth, and blended metrics as the referee between the two.
Why does Meta differ from your real revenue?
Attribution answers a simple question in a self-serving way: which purchases may this platform take credit for? Meta checks whether someone clicked or saw an ad within the attribution window, and counts the purchase if so. Google does the same for its own touchpoints. Add up the reported revenue from all your channels and you will almost always land above your actual revenue. Everyone is claiming the same purchase.
Since iOS 14 there is an extra layer on top: part of the conversions can no longer be measured directly and gets estimated through modeling. The result is a dashboard that gives direction but is not bookkeeping. Read it as bookkeeping and you will draw the wrong conclusions, structurally.
What does an attribution window actually do?
The attribution window determines how long after a click or impression a purchase still gets credited to the ad. A seven-day click window means: if someone buys six days after clicking, Meta claims that purchase in full. Even if that person read an email in between, googled your brand and came back through a review site. Wider windows make your numbers look better without changing anything about reality.
View-through attribution goes a step further: a purchase counts if someone merely had the ad on screen. That can be useful signal for some analyses, but the difference between seeing an ad and being convinced by an ad is exactly the difference between attribution and reality.
In practice this means two things. Pick one window and hold on to it, so your numbers stay comparable over time. And never compare campaigns or tools running different windows as if they measure the same thing: a difference in settings can be bigger than the difference in performance you think you are seeing.
What is over-claiming and why does it affect you?
Over-claiming is taking credit for revenue that would have existed without the ad. The clearest example is retargeting: you show ads to people already in your funnel, with products sitting in their cart. Part of them would have bought anyway, but the last click happened to run through the ad, so the platform claims the entire purchase. That is why retargeting almost always looks fantastic in Ads Manager, while its true effect can be far smaller.
This is why we weigh new-customer share so heavily. Revenue from people who did not know your brand yet is much harder to over-claim than revenue from your existing customer base. Growth comes from strangers becoming customers, and that is exactly the movement you want to measure cleanly.
How do you triangulate the truth?
You do not need a perfect attribution model, you need multiple independent sources that keep each other honest. This is how we approach it with the brands we work with:
- Platform data for direction: use Meta numbers to compare ads and concepts against each other, not as absolute truth.
- Your backend for truth: revenue, orders and new-customer share from Shopify are the numbers you steer on.
- MER as the referee: total revenue divided by total ad spend. If spend rises while MER collapses, you are buying revenue you already had.
- Trend breaks as evidence: raise or lower spend deliberately and watch what total revenue does in the weeks that follow.
Turn this into a fixed weekly ritual. Every week, put spend, reported revenue, actual revenue and new-customer share side by side in one overview. After a few weeks you will know how big the gap between platform and reality normally is for your brand, and deviations from that pattern become the real signal to act on.
Your bank account is the only dashboard without an attribution model.
What role does platform data still play?
A big one, as long as you read it relatively. If concept A reports twice as many purchases as concept B, at comparable spend and the same window, that is a reliable signal, even when both absolute numbers are inflated. For creative decisions, budget allocation within an account and finding winning hooks, platform data remains the fastest instrument you have. We make creative decisions on Meta data every day, across 65+ brands. We just never trust the absolute number without holding it against the backend.
Conclusion
Meta is not lying, it is answering a different question than the one you are asking. Attribution measures what the platform may claim; you want to know what your spend causes. Read platform numbers relatively, steer on your own backend and let MER be the referee. Want to know how big the gap between dashboard and reality is for your brand? Book a call and we will gladly look at your numbers together.