Inside Look At Collarity's Personalized Search

Written by Emre Sokullu and edited by Richard MacManus

When we did a poll
asking what search approach stands the best chance of beating Google, personalized search
came out tops with 22%. Collarity was the company we used as an example of such a service,
so we wanted to find out more about it. Last week, I had the chance to meet Collarity
founder Levy Cohen and his work-mates in Palo Alto and get some insight about their
2-year old, 15 person, angel funded company. We also managed to get some exclusive
screenshots of the as yet unreleased Collarity Discovery, which you’ll find at the
bottom of this post.

Firstly a quick overview of Collarity. It’s described as “community based search
technology that automatically tailors each search to the specific interests of the
individual.” Let’s explain this with an example. When we search, do our queries always
return relevant results? Let’s say you do a search on “execution”. If you’re a financial
type of person, “execution” might refer to the management of a company. If you’re a
programmer you may mean running an application. Or perhaps you’re looking for information
on Saddam!

In search, putting social and personal variables into the equation is one way
to improve results. Everyone should not get the same results; your results should be
personalized. Also, searching on specific websites should return results that are
influenced by the site’s community. This is what Collarity is aiming to do. Collarity
uses Yahoo in the background as its underlying search engine, but their patented features
are:

Focus on “Search In Site” instead of Global Search
Compass: an innovative User Interface that allows you to easily distinguish Global,
Personal and Community Search
Ability for Web Site Owners to get useful information about the people who visit
their site

A Screenshot of Compass

Personalized Search Market

The Personalized search market is highly crowded, with well-funded startups like Wink,
Rollyo, Swicki. But Collarity has a unique position – which we will call Behavioral
Personalized Search. Collarity tracks your behaviors via javascript snippets placed
in partner sites, to extract your personal tastes and your communities automatically.
Other services (Wink, Swicki et al) require you to define your search engine manually, by
entering your favorite keywords and web sites or by voting.

To summarize the other approaches to personalized search:

So Collarity works on the ‘collective intelligence’ principle. All your search history
is saved, so that they can find your internet usage attitudes and like-minded people (aka
your communities). But also via Javascript, they track your behaviours on partner sites.
The sites that Collarity has made deals with are currently confidential. While I can’t
disclose them, I recommend you the look at the sources of the sites you frequently visit,
as you may find Collarity bits in the HTML. [Editor’s Note: before you go
looking, I can confirm that R/WW isn’t one of the test sites!]

The advantage of Collarity over other approaches is the fact that it does not require
you to change your behaviours. Your behaviours are tracked and clustered in the
background by Collarity. Making the extra effort to enter keywords is actually a big
burden for users. But think how you would feel if all your search results were magically
customized for you without needing you to change your attitudes. For example, as an
automobile collector you’d never see the animal Jaguar while you search the newest
models.

The disadvantage of Collarity is the potential privacy implications of having your
behaviors tracked and recorded, without your knowledge.

Collarity Discovery for Site Owners

Collarity also has something new to offer site owners. You can essentially outsource
the hassle of building and maintaining a ‘search in site’ system to Collarity, as well as
track important data like who visits your site, what they search for. This might give you
a better picture of your community and the textual dynamics driving your site. At the
bottom of this post we have some exclusive screenshots from Collarity Discovery, a
site-specific analytics interface.

Competition

Collarity’s competition is the companies we mentioned above. But their biggest
competition may end up coming from Google. Google has the advantage of dominating most of
your internet sessions, so it has the biggest access to your personal data. Also Kaltix,
which was snapped up by Google in 2003, was a pioneer in search personalization.
Co-founded by PageRank guru Taher Haveliwala, Kaltix
innovation allowed Google to compute pageranks 1000 times faster. But Google seems not to
have focused on personalization since then, leaving a gap for Collarity and others to
explore. Also don’t forget that Collarity mostly focuses on search-in-site functionality
instead of global search, and is more about innovative interfaces and personal/community
searches.

The new free Yahoo powered OmniFind
search-in-site solution of IBM may also be a threat to Collarity. Some site owners may
want to rely on their own servers and not outsource search to another party. However OmniFind does not have the personalization, community clustering and site
dynamics tracking capabilities of Collarity.

Conclusion

Collarity, along with Yahoo MindSet, SearchMash and the stealth mode Powerset, is one
of the most promising search 2.0
companies. However Collarity needs to keep up the pace and announce their big deals
as soon as possible. Then we’ll have more opportunity to evaluate their technology.