Google's "Reasonable Surfer" - Bill Slawski popularizes the concept

[tedster]

Many SEOs know and follow Bill Slawski's "SEO By The Sea". It's the best resource I know for following the information retrieval patents from Google and others. The best part? Bill does a lot of the heavy lifting for you, removing some of the pain that wading through a patent's twist and turns can cause.

In a recent article, Bill helped to popularize a phrase that seems to be catching on around the web. And it should catch on - it sums up the change from the original PageRank formula's "random surfer" to Google's new Reasonable Surfer Model [seobythesea.com] very nicely, spanning many patents.

In short, there's no longer anything "random" about the way link juice flows. Instead Google is working toward a "reasonable" surfer model. I don't think they've got it nailed yet, but they've made progress.

Would a reasonable surfer click on footer link just as often as a link in the first sentence? Or a link in the main navigation? Of course not. Bill lays out the case from Google patents, and it's a good read.

< edited to correct my original inaccuracy - see below >

[edited by: tedster at 5:54 am (utc) on Jun 2, 2010]

[buckworks]

Did Bill coin it? The phrase "reasonable surfer model" appears in the Google patent here:

[patft.uspto.gov...]

[tedster]

Ah, mea culpa! I should say that Bill is POPULARIZING the phrase. Thanks for the correction - and I'll be editing my first post so we don't continue to publish the inaccuracy.

[jimbeetle]

Wow, so everything old is new again ;-)

I remember we talked about varying link values here and there a few years ago, though basically just whether a link's ability to pass PR was on or off. And at that time we -- or at least I -- didn't envision the number of variable that might be looked at. (And, as usual for a Google patent, the U & D part of FUD is ever present.) I was thinking at the time of page segmentation, for example, where blog roll and footer links wouldn't carry the same value as links in the content well.

A couple of other interesting points:

Examples of features associated with a source document might include:
.
.
.
6. A topical cluster with which the source document is associated; and/or
7. A degree to which a topical cluster associated with the source document matches a topical cluster associated with anchor text of a link.

Again, a (relatively) old concept for Google (Was there a specific patent on the process?). Can still see the effect on the SERP side where the more powerful page of a cluster is returned -- say a category page -- rather than the page that actually better answers the query.

User behavior data associated with documents and links may also be considered, such as:
.
.
.
3. Interests of the users,
4. Query terms entered,
5. How often a link is selected,
6. How often links aren’t selected when one link is chosen

That's right out of a bunch of discussions about how user interaction with the SERPs can affect ranking. It's very interesting to me how many of the same factors are being used in different parts of Google's ranking system. (I can no longer call it an algorithm; with all the discrete pieces feeding it, it feels more like an ecosystem, a beast of an ecosystem.)

Very interesting stuff, and as always thanks to Bill for tearing it apart.