diberry - 2:43 pm on Apr 14, 2013 (gmt 0)
This right here is one of the best pages I've ever seen on WebmasterWorld. TOI's comments on measuring "when the search ends" fill all the gaps in the "Panda is all about Exit Rate" theory. It's not exit rate per se, but whether you come back and search for the same thing again... the two metric would correlate a good bit, but while "exit rate" could be very misleading for sites that satisfy queries like "who won last night's game" in two seconds, the "when the search ends" metric wouldn't be.
I believe brand bias may actually be hurting Google, but not in a way they're going to feel for some time, and maybe not in a way that wasn't inevitable at some point anyway. Anecdotal support based on people I know in a major city who are just average surfers using the internet to shop, socialize, and be entertained:
--They Yelp to find local stuff to do, places to go. They don't think of a search engine as the right job for that task.
--They go directly to websites to shop instead of going through search.
--They hear about entertainment sites through Facebook or word of mouth, and bookmark them without ever going through Google.
--They use Google to research topics.
And the reason they give me for not using Google primarily? "All Google wants to do is sell you stuff." While Google is not directly making money by feeding you transactional sites for queries that may or may not be transactional, that is their perception.
I say this was inevitable because search used to be the best tool we had for everything, but now it's not. Yelp is a much more robust way to find local stuff of interest. Now that people understand how shopping through websites works, there's no need to look "Dillards" up in Google if you know it exists already. (Though some may unintentionally route through Google to get to Dillards, by typing it into the bar without the ".com", thus inflating Google's sense of its - and Dillards' - importance in search.)
I wish there was a way to cook up some quantified data on this. Instead of looking at stats from Google or Facebook or Yelp, just take a large sample pool of people and give them access to the whole net and a task list like, "Find a great new restaurant to go to with friends and book a table" and "Find a new humor site to enjoy anytime you need a short break from work" and "Shop for a new [whatever]" and see what they do.