Ad platform ROAS is not real ROI. Why last-click attribution and revenue-based math distort scaling decisions.
Most advertisers look at the ad platform report — "X euros spent, Y conversions, ROAS 4.2" — and treat that as real ROI. It isn't ROI: it's the platform's own ROAS, calculated on data the platform itself provided.
The last-click attribution problem
Last-click attribution inside the platform itself systematically overstates the platform's contribution: if a user saw an ad, then searched the brand on Google and bought a month later, both Meta and Google can simultaneously claim that sale for themselves.
ROAS counts revenue, not profit
A €100 sale with a €70 cost of goods and €20 in ad spend gives a ROAS of 5, but a real margin of only €10 — and that's before accounting for returns and logistics.
How real ROI is calculated
Real ROI is calculated from net profit, not revenue: (revenue − cost of goods − ad spend − operating costs), divided by ad spend. The gap between ROAS and real ROI on the same campaign can be twofold or more.
From last-click to a cross-channel model
Last-click attribution needs to be replaced with a view-through window and a cross-channel model — at minimum a simplified post-purchase "how did you hear about us" survey, or server-side analytics with a single source of truth outside the platforms' dashboards.
LTV, not a one-off sale
LTV, not a one-off sale, is what should go into the ROI formula for businesses with repeat purchases. A campaign with a ROAS of 2 in the first month can be far more profitable than one with a ROAS of 5, if the first brings in customers with triple the LTV.
In short
- The platform's ROAS is not ROI: it's calculated on data the platform itself provided.
- The gap between ROAS and real ROI on the same campaign can be twofold or more.
- Real ROI = (revenue − cost of goods − ad spend − operating costs) / ad spend.
- A campaign with a ROAS of 2 can be more profitable than one with a ROAS of 5 if it brings in customers with triple the LTV.
- Calculate marginal ROI by 30-60 day cohorts, reconciling it against your CRM and warehouse data, not just the Meta or Google dashboard.
There's no universal recipe here, but the mistakes listed above cost the most and repeat the most often.