E-commerce
Your ad dashboard says one number. Your store says another.
Raghav’s work at a Canadian sports memorabilia retailer, where he works now.
Showed that Google Ads was claiming credit for 130.8% of the store’s web retail sales, and worked out a realistic return range (4.42x to 22.96x) to budget against.
- Share of web retail sales Google Ads claimed credit for
- 130.8%
- Store-matched return, against the 22.96x Google reported
- 4.42x
- Of listing spend on unbuyable products, which ruled out a theory
- 1.17%
The situation
The retailer spends heavily on Google Ads, and budget decisions were based on the return Google reported. Nobody had checked those numbers against what the store actually sold.
What was hard
Every ad platform counts its own conversions generously, and the store, email and automation tools each tag traffic differently. A theory also needed testing: that listings customers couldn’t buy were behind a drop in orders per click.
What Raghav did
- Set up account-wide tracking tags, and checked 30 email flows and 24 store automations for tag conflicts.
- Built a census of every unbuyable product listing receiving ad spend, run by read-only AI agents. Adversarial reviewers found two ways around its read-only guard, which were closed before the results were trusted.
- Compared Google’s reported results with store-matched sales to produce a floor and a ceiling for real return on ad spend.
- Added 12 negative keywords to the live account, each with a readback to confirm it and a rollback ready.
- Set the working rules: agents run only dry runs and read-only queries, a different AI model reviews every change, and a person runs every live write.
What changed
Google’s reported return was 22.96x, but the store-matched floor was 4.42x, and Google was claiming 130.8% of all web retail sales.
The census showed unbuyable listings took only 1.17% of listing spend, which ruled them out as the cause of the drop and kept the team from chasing the wrong fix.
How this was measured. Google’s figures come from the ad account; the store-matched floor uses the store’s own recorded web retail sales over the same period.
Bring us in when you’re spending real money on ads and want to know what they actually return before you raise or cut the budget.