|
Quiver: Inktomi on
Steroids?
By
Andrew Goodman - July 10, 2000
I recently
had the opportunity to talk with Scott Potter, CEO of Quiver,
a hot new directory infrastructure company which seeks to capitalize
on the increasing importance of vertical or topical searching. Quiver
was founded by two algorithm and data experts, Ofer Mendelevitch
and Avi Segal, whose collaboration began when they served together
in the Israeli Army's elite technology unit. The company is now
located in San Francisco (of course).
Human
guidance plus popularity ranking algorithms
Every new web
navigation service seeks to improve on the limitations of existing
search engines and directories. The hottest trends in the past couple
of years have arguably been notions of human-guided search, as well
as technologically-driven methods of filtering or rating web content
by popularity. Direct Hit, a "popularity engine," tracks the time
users spend on sites under popular search terms. In practice, however,
the technology hasn't lived up to its promise. Google is a more
compelling search service, beloved by many for its uncanny ability
to point to relevant pages. It is designed to assess the reputation
of web content by assigning a PageRank score which is based on a
complex algorithm which focuses on measuring the number and reputation
of links to a page.
Like
Google, Quiver leverages scientific muscle
As the Quiver
folks pointed out to me, there is nothing straightforward about
what Google is doing. Google's work is the culmination and commercialization
of some high-level theoretical ideas about "spectral filtering"
developed by computer scientists such as John Kleinberg and Amos
Fiat (the former was a mentor for the founders of Google; the latter
mentored the founders of Quiver). There is some complex math involved
in ranking pages to determine whether they are either good "hubs"
(point to a lot of important content) or good "authorities" (get
linked to by major hubs). The problem is, there can be a certain
circularity to such measurements, so there is a need for complex
math to break out of the circularity problem. With companies like
Google and now Quiver, we are starting to see the beginnings of
a real commercial use of advanced search and filtering techologies.
In other words, in the ongoing pursuit of more reliable Internet
navigation and more relevant search results, we're just getting
warmed up. The best is yet to come.
Quiver has
chosen to go in a specific direction - choosing to become an "infrastructure
company" which will power the directory offerings of vertical and
niche sites. This makes sense in two powerful ways: (1) in a navigational
sense, because the reputation of web content is being rated by the
RELEVANT community, not by a diffuse collection of general web surfers,
and not by one editor assigned to a category; and (2) in a business
sense, focusing on the importance of vertical and affinity markets
and the demand for custom content and directory services for the
many niche portals now being built.
Beyond
the "single editor" model
In the first
place, then, we have yet another claim for enhanced relevance. It's
a powerful claim, because Quiver has chosen not to rely on a general
faith in "human guides" nor a technologically-driven solution that
tries to master the entire Internet universe. It's a hybrid model.
One of the first major Quiver partners is an example of the type
of vertical or affinity portal which seeks a means of having the
community of users recommend web content: Gay.com. As Potter put
it, "what are users of this portal going to trust more if they are
looking for a vacation spot, the recommendation of one 27-year-old
Looksmart Editor, or the recommendation of the community of people
who are members of Gay.com?" In the second place, Quiver hopes to
find many partners willing to pay for Quiver as a solution to providing
vertical-focused directory services in a similar vein to a company
like Looksmart. Whereas Looksmart builds numerous custom directories
using a team of professional editors, Quiver's custom directories
will use a combination of technology and the surfing habits and
recommendations of members of vertical communities. As a result,
niche portals may find the custom Quiver service more affordable
than Looksmart's.
Only
as good as its partners
Of course,
Quiver is a much newer company, facing the growing pains of any
newer company. It has promise, but no one yet knows if Quiver can
deliver. One potential drawback is that some of the Quiver measurement
will require the downloading of a special toolbar. It will be up
to vertical partners to offer their members incentives to use the
"Qbar."
The advantage
of the Qbar is that it tracks users' surfing habits more closely.
Therefore Quiver is likely to deliver more reliable rankings than
the popularity engine Direct Hit, which to some extent can only
guess at users' patterns. Quiver has made some great strides of
late, securing two rounds of funding from (among others) Hummer
Winblad Venture Partners, based on the strength of its ideas and
its ability to map out a unique strategy. The community-powered
directory concept is a little like Backflip (the online bookmarking
service which allows users to share collections) and a little like
the combo of Open Directory Project and Google (a directory using
a filter to rank sites by reputation or popularity). The difference
is that the Quiver ranking technology is being applied by the most
relevant community, and topic-specific versions of the directory
will be deployed in many verticals rather than being touted as a
comprehensive guide to the web.
Inktomi
on steroids... they hope
Because of
this "infrastructure" focus, Quiver and its VC investors hope that
this community-powered directory service has the potential to be
"Inktomi on steroids." I can't help but hope it realizes that potential.
Quiver has similar appeal to some major players in the navigation
and search infrastructure field, but it has cleverly chosen its
own path - one which shouldn't overlap too much with others. In
the end, Quiver will only be as good as its partners. May it find
many.
Maybe it's
just my overactive "spider sense," but I suspect something big might
happen to Quiver. It could come in very handy to a vertical player
like About.com, or you name it. Human guides plus a system for tracking
the habits of a particular community of members could form a powerful
combination. Ask Jeeves acquired the popularity engine Direct Hit
for what was then $507 million in stock. The early indication is
that Quiver is offering something considerably more powerful than
Direct Hit.
Update: since
this article appeared in the Traffick Monthly for June, Quiver is
close to closing deals with five partners, and has many more in
the works. An updated version of the software is slated for release
in September.
|
..........
Can you afford
not to know this stuff?
Sign
up for the weekly Traffick Newsletter
and stay informed!
|
|
