tedster - 5:51 am on Dec 15, 2012 (gmt 0)
The reason I zeroed in on mobile is that some members in this discussion pulled that difference out of their log files. So we have some members who actually caught that critter in the wild.
I agree that the entire testing mechanism Google apparently uses is worth considering, but testing what exactly would be good to pin down - again, now that we have the idea, let's capture the actual critter. Here are ideas I've had - differences to look at between zombie traffic periods and regular periods.
1. Analyze the number of "not provided" keywords - Google could be testing some searches against personalized, logged in users.
2. Analyze whether the variety of search terms shifts between long tail and short tail.
3. Google could be testing clearly purchase intent searches against informational searches - trying to determine which type of page your page is.
4. Analyze the country of the user's IP address during the two periods.