Thursday, July 12, 2007

Random Surfer Theory

Part of the Google organic algorithm involves what has been called Random Surfer Theory. It is the unstructured nature a person uses the web to find things. A click here and a click there, follow a text link, try a banner ad, do a search... the connections between sites is tenuous and haphazard.

Where did it come from? Math and in particular graphs.

Throw in a little of this and a little Erdos Numbers.... and eventually you get a feel for it.

A person - Orion - who's opinion I value and who I have seen give a panel session to thousands at SES over a cell phone feed through the room's sound system - with his Spanish accent - and just a very small number left the room through his discussion of advanced search theory - details it.

Orion aka Dr Garcia, in an interview with Rand Fishkin over at SEOMoz, said: " By 1995, I heard about something called “search engines” and “web crawlers”. I immediately realized that the Web is a dynamical system. Many processes on information flow, retrieval and web surfing respond to equations of the same form to those modeling random walk, diffusion and markovian processes. For instance, in DLA we talk about “sticking probabilities” and the “probability that a random walker visit a site of a growing cluster”, In “Web terms” we talk about the probability that a random surfer will visit a site part of a web graph. However, let’s keep in mind that this graph is not static but is constantly growing.

Another analogy, a random walker moving through a self-similar DLA structure and a web surfer moving through a self-similar link structure are similar scenarios. The structure can affect the behavior of the walker and the walker can affect the growth of the structure. Depending on conditions, a system can exhibit different degrees of surfer-surfer, surfer-structure and structure-structure interactions so the walker no longer is a pure random or pure deterministic walker. Certainly these are crude analogies."

It is a little complicated when it is attempted to be mapped out. Basically it implies that web users do not follow very structured lines to get where they want to go.

If you overuse repeated anchor text, use the same landing page for the text links, buy links that archive and end up being thousands all in the same spot on a page ad using the same words.... you violate Random Surfer Theory...

Or as Shakespeare wrote: Oh what "an untangled web" we weave when first we practice to deceive.... well he should have!


  1. Frank,

    I follow Orion's work too and I highly respect him. My island (Santo Domingo) is right next to his (Puerto Rico), so hopefully I will have the chance to meet him soon.

    The math concept that best describes this is the Markov Chain. You are right when you say that when we use predictable patterns in our SEO work, we raise all sort of flags that scream: SPAM.

  2. The Markov Chain is another of the interesting reads...

    So many people have their perception of the algorithm in terms of title tags and links and see it all in verbalized components, when in reality the elements are never as important as the math used in the background to weigh them all.

  3. And Orion is a great guy outside of the classroom setting as well.

  4. Thank you for this information. I don't follow all the points addressed here, especially on the first read-through, but that is likely due to my less-advanced knowledge of this particular aspect of web surfer behavior.

    This may be the first time I've been at your weblog... Incidentally, I was only recently looking for info about keywords, and I was just (a few minutes ago) thinking about optimization issues, in the context of garnering traffic, and earning money... Incidentally however, I stumbled onto this weblog, when searching for info about another topic. Keep up the good work!

  5. Read as much as you can find and you will get there. is a solid resource... articles, news and forums