March 05, 2003
40% random growth : A PageRank incident

February was another blowout month on classy.dk with 40% traffic growth. This time the growth was caused by one very specific reference to a a joke from "The Simpsons" about the french. The phrase in question was hard to find on the net and therefore my lowly weblog turned up as number 4 on Google for this high profile search term. That single weblog entry accounted for almost 20% of all hits on classy.dk.
What is interesting about this is first of all Google's efficiency even for low change low rank sites like mine. 4 days after I posted the entry the traffic doubled because of this log entry.
It would appear that a viable Google marketing strategy is very specific multi-word text which uses infrequent words. I.e. if you can find a phrase, some meme that is flying around currently, then you can harness that as a cheap traffic boosting scam by publishing the exact phrase. It is more difficult to get a good ranking for very general searches or single word importance - but for very specific searches you can get lucky. This is truly targeted advertising.
It is not surprising that it works of course and taking advantage of the effect doesn't even have to be malicious, in fact it can be hard to be malicious about it. You have to define a very specific audience that you're targeting and if you're being specific enough people might not even mind, nor consider what you're doing advertising. On the other hand, one can easily envision a new kind of link farm generating entries for search terms crawled automatically from Google Zeitgeist, like e.g. Norah Jones 'Come Away With Me'.
It would be interesting to study how possible this is considering the distribution of PageRank on the internet. The simpler your search term, the more popular it will be. But the simple term will also mean that the subset of URLs meeting the search criteria grows, and this of course means that the PageRank of the most popular URL in the subset is higher.

Incidentally, from the case of Norah Jones - one of the top 5 searches for the week ending Feb 25 - one can see that Google is ordering by, but not displaying on the google toolbar (which I use), alternatives to the pure PageRank of the URLs displayed. As an example the Amazon.com page for Norah Jones' album is listed in the top ten. Obviously a good choice - but of course that particular Amazon search URL for that particular album doesn't have a significant PageRank (it is 2).
I'm not sure if a simple page equivalence based on similarity of content fixes that or if it takes something more sophisticated than that to correct the Rank for this effect, but it might just be really simple to do: If you do a site specific search on Norah Jones you get approximately 23000 amazon pages. The top twenty on that list are different URL's showing essentially the same data - either a Norah Jones search page or the product page for her debut album.

Posted by Claus at March 05, 2003 10:18 PM