Personalized Search

Google just launched a new version of personal search based on user preferences. In early 2000, we seeded a Xerox PARC spin-out called Outride (formerly called Groupfire) which aimed to bring personal search to the web by learning from a surfer’s prior searches or his workgroup or community’s prior searches. For example, if you searched on the word “Java” how does the engine distinguish Java the computer language, Java the coffee, or Java the island. So based on user behavior if you were a tech geek and into computer programming, it would serve you up the Sun Java and so on and so forth. The business model was to be an “arms merchant” to all of the major search engines like AOL and Yahoo. The problem was that it was very difficult to monetize. How do you get a search engine to pay for a supposedly more personalized search result? So at the end of the day, we ended up selling the assets and patent to Google. Fast forward to now and Google is bringing this back into the market, although it is using its latest acquisition, Kaltix, as the basis for its search. This one is based on profiles rather than behavior. As Jim Pitkow, co-founder of Outride says in a San Jose Mercury News article:

“That’s good because the search engine doesn’t have to try hard to infer anything from the user’s behavior. But it can also be a disadvantage, because a person’s interests will change over time, but they may not update their Google profile to reflect that. It’s really unclear what it’s learning about me,” said Pitkow.

While the idea for Outride was interesting, we were way ahead of the market without a clear business model. It is a long story but the old adage “pioneers get arrows in their backs” certainly applies to this company. Anyway, it seems that Eurekster is doing a very similar job to Outride except that it has created a destination site. It will be interesting to see how personalization and the search wars play out over the next couple of years. I, for one, am a big fan of the original Outride model based on user behavior. Of course, that can open up a whole new issue related to privacy. If you are interested in personalization, I suggest you visit the Eurekster blog for a nice comparison of Google and Eurekster.

Published by Ed Sim

founder boldstart ventures, over 20 years experience seeding and leading first rounds in enterprise startups, @boldstartvc, googlization of IT, SaaS 3.0, security, smart data; cherish family time + enjoy lacrosse + hockey

2 comments on “Personalized Search”

  1. Hi Ed
    A few years ago I had filed a patent titled ” Predicate indexing for locating objects in a distributed directory”, which talks about optimizing searches based on reducing the search query to target the previous result sets. The domain was different and the searches were based on Directory objects. So one would do a search first and the result set would be cached. Any subsequent queries that can be reduced via the previous query (the previous query is cached with the result set as well) can use the reduced target set. A side effect of this is that by applying proper variables the searches can be highly personalized.
    The funncy part was when I did this invention I was at Novell and Dr. Eric Schmidt was the CEO. In 2002 my patent was the most innovative technology awarded in the state of Utah. Ofcourse I had left Novell a year before this happened and did not have the pleasure of getting honored, but Novell did manage to locate me and provided me the patent plaque, which is still with me

Comments are closed.