Social Search: It's Not Who You Know…

Originally published in MediaPost’s Search Insider; it continues in the extended entry

EVERYONE WANTS TO BE social these days. Given the hype around
social networks and social bookmarking, it’s little wonder that search wants to
be social too. The problem is that when you refer to "social search,"
the odds are that whoever you’re talking to will have a very different idea of
social search than you do.

Social search should
be defined as the process by which a site’s community of users influences the
algorithmic search results displayed for any one of those users. A classic
example from several years ago was when Eurekster
powered Friendster as of 2004. Back then, you could see the most popular
searches from your friends, and search results viewed by your friends appeared
at the top of your results page. It was a model for social search, until it was
discontinued.

Over the years, Eurekster has kept up its focus on social search
while taking it in new directions. It has been focusing on its Swicki product,
which is essentially a custom search engine using social search. A Swicki’s
creator, generally a blogger, defines the sites to include or exclude and other
parameters, and then searchers can vote on the results. Eurekster also shares
the most popular searches for that community’s engine. On Read/Write Web, the hot
searches are "mobile," "next generation," and (woefully )
"web 2.0," while on TechCrunch
the hot searches are for "$5 million funding," "aideRSS,"
and "grockit." All the while, the Swicki learns from searchers’
behavior.

I asked Eurekster CEO Steven E. Marder for some color on
Eurekster’s evolution. He said, "When we started, we thought that
leveraging social networks that were built on FOF (friend of a friend) would
provide an active and robust set of user data that we could use to refine and
rerank search results. What we found is that your pure social nets are so broad
(as your friends have many interests — and not necessarily expertise in those
interests), that the search results would be influenced and skew, and would be
of interest to searchers BUT not necessarily be more relevant. However, we have
found that applying our technology in a social network environment makes great
sense if within vertically oriented or geo-targeted oriented groups within those
networks. Friendster was too unstable as a company and a platform for us to
effectively roll this out with them."

That means that with social search, it’s not who you know, but
who you share affinities with. If I’m searching for travel recommendations,
I’ll have more in common with other readers of a blog about exotic travel
destinations than I will with most of my friends. The wisdom of crowds has its
limitations. If I were to pick a honeymoon spot based on where all my friends
are going, I’d wind up in Hawaii. If I were to pick a honeymoon spot based on
the results selected from other people choosing more unusual honeymoons, I’m
more likely to find recommendations for India, Chile, or some other place that
my social network as a whole would not come up with.

That’s what social search is supposed to be about. Yet listings
for social search sites, such as the comprehensive list on
Mashable, are littered with sites that have no business being there, Given
some of the confusion over what social search is, here are some examples of
what social search isn’t:

· Human-powered
search: Human-powered search engines are barely search engines,
let alone social search sites. With sites such as Mahalo, people write the
searchable content, but they have no control over how it ranks in search
results. Additionally, the search results rank the same way for searcher.

· People
search engines: Searching for people has nothing to do with
social search. Spock is a search engine for people where anyone can contribute
to profile pages, but that’s not enough to call it a social search engine.

· Live
assisted search: At ChaCha, you can have a guide help you
search, and there is some social interaction, but it’s not the process of
social search. And for the most part, it’s a waste of time. Assisted search is
slow, guides rarely know more if anything than the searcher, and the major
engines are getting better at offering shortcuts to guide specialized searches.

That being said, there are still other examples of companies
tackling social search in its truest sense. One great example is from Collarity, which runs on publishers’ sites
and learns from users’ search behavior to recommend search terms specific to
that site. In its partnership with Fox Interactive Media, you can see a perfect
demo of how searches differ from communities, as Collarity powers Fox’s local
MyFox sites throughout the country.

When you type "rangers" in the search box at MyFoxDFW.com (for Dallas-Fort Worth),
all the search terms it recommends relate to baseball, given that the Texas
Rangers play in the DFW Metroplex. Yet when you enter "rangers" as a
search at MyFoxNY.com, the first
suggestions are "shanahan" and "lundqvist," players on the
New York Rangers hockey team. Even more generic search terms show differences;
while a query on "crime" brings up terms related to hate crimes for
both cities, DFW’s results also refer to "crime spree" and
"crime scene" while NY’s refer to "crime stoppers" and
"violent crime."

What’s important for Collarity is
what other visitors of your local MyFox site are searching for, not how closely
related those searches are to you. You might be a friendless hermit with no
social network to speak of — but with social search, you’re never alone.