The image won every poll. Advertising CTR said it was worse. Both were telling t
By Mansour Norouzi · September 29, 2026 · Curated by George's Blog
The image won every poll. Advertising CTR said it was worse. Both were telling the truth.
A client changed their main image. The polls were not close. It beat their old image and it beat the competition. I looked at it myself and thought, yes, obviously better.
Then we looked at click-through rate in advertising. Not up. Actually a little lower. The client wanted to kill it and go back to the old image. So the question on the call was, if this image is clearly better, why is the data not showing it?
Here is the thing I keep repeating to my team. Do not take one metric and call it a conclusion.
Advertising CTR is not reliable for this. Your ad shows on product detail pages and rest of search, and you have no idea if the shopper ever saw it. A no-click there is not a rejection. It is an impression that never really happened. You are dividing by a number you cannot trust.
So we asked a different question. What metric could actually prove this change worked?
Search query performance. Click share, at the keyword level. Out of all the clicks that search term handed out, how many came to us?
Our team pulled four weeks before and four weeks after. Click share went up and stayed up, even after we set the deal week aside.
Nothing else changed. Just the image. The new image stayed.
And no, I do not look at the CTR column inside search query performance either. Same problem.
Before you run a test, decide what metric could prove it worked. Otherwise you will kill a winner and never know.