Turn off a campaign somewhere and watch what happens to total sales. Everything else, including the return on ad spend that ad platforms report, tells you who gets credit for the sale, not whether it would have happened without ads. An ad platform will never tell you it billed you for a sale you already had.
The benchmark
The simplest test is geographic. You stop a campaign in a few regions, keep it running elsewhere, and compare how total orders change between the two groups.
The gap is what the campaign was really bringing in. Google recommends at least four to six weeks, longer if your customers take time to decide, and its own researchers' protocol requires regions that are comparable in size and history.
For a brand making a few million euros, this test is often out of reach: there aren't enough orders per region to read a gap. A scrappier version still works and is easy to read: turn off one campaign entirely for four weeks, then look at the store's total orders, plus the share of direct and organic traffic.
At a sports brand, we asked to pull brand terms out of the automated campaigns. Success showed up in the cost per click on the brand name and in the return of direct and organic traffic. The ad platform's report couldn't show it.
The rule
Test the most suspect campaigns first: ads on your own brand name and retargeting. They reach people who have already made up their minds.
The steps
- Scope. One campaign, four full weeks, outside any promotion and any seasonal peak.
- Measurement. Total store orders in Shopify, never attributed orders.
- Direct and organic traffic. If it rises when the campaign stops, the campaign was taking its place.
- The result. Record it in the logbook, then rerun the test once a year.
- The line for leadership:
"We're turning off retargeting for four weeks. If total orders don't move, that budget was a toll."
The trap
The campaign with the best reported return is often the least useful. An acquisition consultant, in a meeting with a lifestyle brand, showed a return on ad spend of 8.48 on retargeting, "which is pretty cool, but actually we're clearly not bringing in enough traffic."
Then, a few minutes later: "It's not normal to go and like a retargeting campaign."
Retargeting shows an ad to someone who already has a cart in progress, or who already likes the brand: they buy, the platform claims the sale, and the ratio looks great. A campaign that introduces people to the brand shows a mediocre ratio and still creates customers.
Steering by ratio alone means paying for what would have sold on its own.
