I see on twitter people asking how much is the part of these google-logged visitors to mesure the impact of this change. Guess what ? Everybody can now how much logged-in-google user visit your website.
Actually, when a logued-in-google visitor arrive on your site from a google search he will have a special parameter « sig2=XXXX » in the referrer, a parameter a not-logged user won’t have. This parameter is leaked by the « www.google.XXX/url?sa= » URL which is reponsible to track the click in the SERP.
So by running this simple command on your apache logs, you can have an approximation of the part of logged-in-goolge visitor you have:
http://www.google.com/search?q=query&… (no js)
http://www.google.com/webhp?sourceid=navclient&ie=UTF-8 (no js)
http://www.google.com/cse?cx= (sponsored search, assume not logged)
http://www.google.com/#.. (what the fuck, it’s extremly log ratio, no impact)
But, in my opinion we can consider the ratio of logged-in ratio from « /url?sa » referrer is the same for others google referrers. Or if you’re not OK with this assertion, you can consider they’re not logged in Google and recalc the ratio.
It’ is sure that you won’t have the same ratio of logged-in-google visitor on a IT/tech website and on a « how to loss weight » website. SEL reported that Matt Cutt said it will only affect a single digit of your user, I confirm BUT :
If Google decides to encrypt the search of all logged-in user (so including people doing search on http://www.google.fr/ and so) the stats would be totally differents, you could except something between 20-40% of your google traffic, but I repeat, now it should only affect google.com domain, but maybe latter… To check that you do :
# Amount of logged-in people coming from google, all google TLD included. egrep 'google.[a-z.]+/url\?sa[^"]+sig2' /tmp/google-urlsa-referrer.txt | wc -l