By woodjr on Mar 20, 2007
A lot of bloggers are talking about Google's patent application for a method of ranking blog Search results. As Bill Slawski and Alex Chitu have noted, these break down into a set of factors which provide positive and negative scoring influences. I won't repeat them all here, but I did find one of the positive factors particularly interesting: the implied popularity of a blog, as determined from click stream analysis in search results.
In other words, if users consistently click on a result from Blog A more often than one from Blog B when both show up in the results for a given search (such as on blogsearch.google.com), it can be seen as an indication that Blog A is more popular and/or of higher quality than Blog B. Pretty obvious stuff. Right?
Sure. And it's also pretty obvious that the same idea can be applied to non-blog resources (such as general web results returned by www.google.com or image results from images.google.com).
The question is... How would Google actually obtain this data?
So when you click on a Google search result, Google should never know it.
But wait... There is a good chance that they do know it. If you use Google's toolbar and enable the "PageRank Display" feature, they'll know about this click (and all of your others, for that matter). Of if the final destination happens to use certain of Google's server-side services (such as AdSense or Google Analytics), they'll likewise know about it (and all other access to that site).
So does this imperfect but growing view of users' behavior on non-Google sites provide enough data to plug into their search ranking algorithms? Probably. And it's one more example of how a web giant such as Google is gaining a "moat" of data which guards against smaller competitors.