Friday, March 29, 2013

Rankdex vs Google

Great ideas do not emerge in a vacuum. The most famous and commonly mentioned example I can think of is Alexander Graham Bell vs Elisa Gray, the two people that raced to file the patent on the telephone. Bell won the race by a matter of hours; but clearly his idea was in the zeitgeist of contemporary technologists if such a big step forward in technology could have been credited differently in a matter of mere hours.

Google was no different apparently.  PageRank, the often cited key innovation that differentiated Google from all other search engines that came before it, had a contemporary competitor: Rankdex.  I found a wonderful thread in the history of Usenet today that is both Google’s debut in Usenet and also a nice debate of the merits of Google vs Rankdex.  From the source Usenet post, Rankdex is (or “was”):

- Linking to a page means you like it.  Counting these votes gives a
rating for a page.

- Anchortext for the link usually summarizes the content. When your
query is straightforward, like “Debbie Harry” this becomes a highly
reliable way of finding a good Debbie Harry fanpage.

Sound familiar?

One key insight that Rankdex misses in that simple description is the weighting of each link based on that source page’s importance.  This problem of weighting sounds circular on an intuitive level, but it’s computable and is a subset of math called Markov Chains

But even without the weighting insight in Rankdex, it’s a wonderful piece of history to see that the concept of a “link means you like a page” was in the zeitgeist.  It shows that the value of Google is in their execution, not their originality, like nearly every successful startup.


  1. adamp reblogged this from thegongshow
  2. nateberkopec said: Never knew about the link between Markov chains and Pagerank! Fascinating. I just use Markov chains to make dumb twitter bots ;)
  3. gbattle said: Put the last sentence in bold.
  4. thegongshow posted this