The ROAS number on your dashboard is lying to you
Blended ROAS hides more than it reveals. A short case for pulling attribution apart before you cut a campaign that's actually working.
Every account I inherit has the same first conversation. The client opens the dashboard, points at a campaign showing 2.1x ROAS, and asks why we’re still running it. Nine times out of ten, that campaign is the reason the rest of the account converts — and the number on the screen has no way of telling them so.
Return on ad spend is a ratio, and ratios are seductive. One figure, one verdict, no reading required. But the numerator and denominator are measured by completely different systems, over different windows, with different ideas about who deserves credit. Treating the result as ground truth is how good campaigns get killed.
Where the number comes apart
Three things quietly distort almost every ROAS figure I look at:
- Attribution window. A 7-day-click window and a 30-day-click window can disagree by 40% on the same campaign. Nobody remembers which one is set.
- Model choice. Last-click flatters bottom-funnel brand terms and starves everything that creates demand. Data-driven spreads credit — but only if you have the conversion volume to feed it.
- The denominator is honest; the numerator isn’t. Spend is spend. Revenue is whatever your tracking managed to observe, minus what Consent Mode modelled, minus what Safari dropped.
A campaign doesn’t have a ROAS. A campaign has a ROAS under an attribution model, and the model is a choice you are making whether you know it or not.
What I do instead
I stopped trying to fix the number and started triangulating it. Three views, side by side:
| View | What it answers | When I trust it |
|---|---|---|
| Platform ROAS | ”Did the algorithm hit its target?” | Bidding decisions only |
| GA4 (data-driven) | “How does this fit the whole path?” | Cross-channel calls |
| WooCommerce / actual orders | ”Did money actually arrive?” | Anything the client sees |
When all three point the same way, act. When they disagree, that disagreement is the finding — usually it means a campaign is doing work that its own last-click number can’t see.
The rule I give clients
Don’t cut a campaign on a single ratio. Cut it when the incremental view — hold it out, watch what happens to total revenue — agrees with the dashboard. Until you’ve done that test, you’re not measuring the campaign. You’re measuring your tracking setup.