By factoring in bounce rate data in calculation of PageRank values, Fabien Mathieu and Mohamed Bouklit create a new and enhanced ranking algorithm.
In their research paper titled “The Effect of the Back Button in a Random Walk: Application for PageRank”, Mathieu and Bouklit (LIRMM) propose an enhancement to Google’s original algorithm. They factor in the bounce rate of a document and refer to it as either “Reversible” or “Irreversible Back”. By doing so they’re giving the traditional PageRank model an interesting and potentially useful addition, a possibility for the random surfer to “return” by hitting the back button.
This is a fundamental shift in the well-established PageRank model which has found its application in numerous areas already, even beyond search. The concept of “backoff process” was first introduced by Fagin in “Random walks with back buttons” in 2000.
Mathieu and Bouklit take this a step further and apply their formulas to a collection containing 8 million documents which has produced results different to those of typical Google results. The research paper was published in 2004 and it is likely that their version of PageRank which offers better modelisation of the web users has been merged with a semantic pertinence-sort and tested out in a practical search situation.
We will attempt to contact the authors of this paper for their comments.
References:
Mathieu F., Bouklit M,. 2004 – The Effect of the Back Button in a Random Walk: Application for PageRank
http://research.deyandarketing.com/The-Effect-of-the-Back-Button-in-a-Random-Walk-Application-for-PageRank.pdf
For the sake of Clarity – as far as I know, Google do Not use Google Analytics data,
thus the “Bounce Rate” that people see in GA is not likely to be the one referenced here.
Instead, it is much more likely to be what I’ve been refering to as
Bounce Back to SERPs (BBtSERP).
Further, it may have a requirement of occuring within a time-window (say withing 6 seconds).
There is also the chance that it may requir follow-up actions (such as clicking the next listing, rather than utterly changing the Search).
I thought to have heard that they use Dwell Time for BBtSERP, but as always in a Googlish way based on other on and off page factors, so it is not a fixed dwell time.
I can’t imagine this happening in the near future. Although I don’t know what the variables stand for, I don’t believe this could consider queries that have been successfully answered. There are tonnes of “answer” sites that you look at 1 page & leave – it’s the nature of these sites.
I think pagerank it has lost its position and is no longer what it was, but its place substitute relevance. Links from relevant trustworthy sites is golden for SEO
And now (10-22-12) I heard they don’t use dwell and/or bounce, . . as such … in their algorithm as a ranking factor (leaving some open ends to do something with that gathered information anyway 🙂
I guess adding Bounce rate in PR algo is not fair thing to do. Bounce rate is a result not cause. Increased bounce rate could be due to many reasons and at times, Good Quality content gets high bounce rate. I wrote an article on this, explaining bounce rate and CTR here: http://socialmediatoday.com/baawraman/1070231/bounce-rate-and-seo