Hi Barry, Dave here. Nice find, I was just yesterday talking about this. The problem with engagement type data (behavioral metrics) is that not only are they easy to mess with (as John mentioned) but it isn't always an indication or interest or disinterest (a noisy signal).

Even in a personalized setting there is every chance for false/positives in the data analysis. If I go to a listing from the SERP and wander off for 10 minutes, does this mean I like the site? If I had to go 4 pages deep into a site to find what I wanted...(and still was disappointed)does this show I liked it? COnversely if I bookmarked the site and went back to the SERP does it mean I didn't like the listing? On and On and On... simply to problematic...

Thanks for the find though, a topic I am keen on...

Dave

"..it is not clear if it is done a few hundred thousand times if it will make a difference or not..." -- haha you are funny!

Well, certainly they take that data into consideration. The most obvious one is probably CTR. Say if you are ranking no #2 for a keyword and a vast majority who searches for that keyword clicks on your result...Google might actually promote you to #1 as based on the 'user data', your site seem to be more popular/relevant than the site which holds the no #1 spot.

My take on it is that John said 'dozens of times' because the questioner said, "They did so dozens of times." I think that JohnMu's statement was directed primarily at answering this specific question and not meant to be considered secret code that back clicks over 763 times will hurt. As Barry said, "But who knows."

This again brings to mind not only what happens with the direct clicks from the SERPs but what happens on a site with Google Analytics installed. Google should have a really good idea what the user did or did not do on the site if their analytics are in place, possibly shedding light on some of the issues Dave referenced.

Well Google has been using these types of data points to calculate quality score in paid search for a long time now. I think that they understand what each of these metrics means by now.

Another problem with using these metrics is that there simply is not enough searches on enough keywords to get a good idea of what site is bad and what one is good. I believe that you need a certain number of clicks in order to get enough data. And JohnMu suggested that it takes more than a few dozen clicks. Some queries are just not heavy enough in traffic/interest. Therefore that makes using these types of stats useless since they cannot implement them effectively universally.

You can't detect a back button browser click on a website. It's not something that is possible to do with 100% accuracy.