It is the most common observation in day-to-day work and the least often said out loud: you add up what the advertising platforms report for a month, compare it with the shop and land above it. How far above is in your own numbers. Anyone seeing this for the first time looks for a tracking error. Often there is none.
Every platform counts for itself
A conversion rarely has just one touchpoint. Someone sees a display ad, clicks a search ad three days later, comes back through the newsletter a week after that and buys. Three platforms, three claims, and each reports the order as its own.
That is not fraud. Each platform sees only its own contacts and counts by its own rule. From its point of view the report is correct. It is just that the sum of three correct reports about the same order is three orders.
Four reasons that add up
- Different attribution windows: Meta looks back differently from Google Ads by default. The same click still falls inside one platform’s window and outside the other’s.
- Different models: where one assigns everything to the last click, the other distributes. Both report the same conversion, each by its own rule.
- Post-view: an ad seen but not clicked counts for some platforms. The others do not see it and attribute the same purchase to their click.
- Modelled conversions: where consent is missing, platforms fill the gap by calculation. Those conversions may have happened, or may not. It cannot be verified.
What the difference does to ROAS
An example with round numbers: €40,000 media budget, €95 average order value. The platforms together report 1,200 conversions; the shop shows 850 orders.
| By the platforms | By your orders | |
|---|---|---|
| Conversions | 1,200 | 850 |
| Revenue | €114,000 | €80,750 |
| ROAS | 2.85 | 2.02 |
Both values use the same order value; the difference comes solely from the number of conversions. The second is total revenue over ad spend, not the effect of individual campaigns. Which of the 350 extra reports are credited twice and which have other causes, the totals cannot say. Only the individual orders show it. Anyone scaling on the first figure should settle that question first.
What helps and what does not
What does not help: picking one platform and ignoring the rest. Nor does scaling the figures down proportionally. That produces a number nobody can justify and only moves the problem.
What helps: your own measurement that counts every conversion once and distributes it by a rule you can inspect and change. Then the question is no longer “which platform is lying” but “which share do I assign to which touchpoint, and why”. The second question can be answered.
Idea clave
Platform figures are not an independent truth. They are the self-report of a party with an interest in the outcome. That is fine, as long as you do not add them up.