If you haven’t heard about Eurekster, you should check out their Swicki search engine. Swicki provides a way of applying the wisdom of crowds to search engines. If you have a blog or Web site devoted to a subject, you can easily use Swicki to create a specialized, vertical search facility on that subject alone. What’s more, you can allow your searchers to improve the search results because Eurekster automatically boosts results that people click on. Beyond that, Eurekster offers a wiki-style suggestion box where searchers request changes to search results for you to accept or reject. I spoke with Grant Ryan, Eurekster’s Founder and Chief Scientist, to learn more.
Mike: How did you get the idea for Eurekster?
Grant: I saw the merging of search and social networking, so two years ago I formed Eurekster. We initially used friends (and friends of friends) and did a deal with Friendster. But we like what is going on in communities, so we flipped the idea of a search engine on its head by letting publishers change things any way you want. People are concerned about quality when you do that, but honestly, if you create rubbish, then no one will use it.
If you look at how any new media forms, you start with the big guys, such as TV networks, and then get more and more vertical, [such as cable or YouTube]. Think about the fact that businesses sell billions of dollars in ball bearings—suppose you had a search engine that did nothing but ball bearings?
Mike: What’s your business model?
Grant: We want the publishers to have control over monetizing their searches—pay-per-click, banners, subscriptions, pay-per-call, whatever. Different publishers have different business models and we want to give them flexibility. If you have a search engine on patents, you might charge a subscription fee [whereas] a digital camera search engine might use pay-per-click. When you have a targeted audience, the value goes up to the advertiser—you already see that with travel search engines.
Mike: Do you work differently with larger publishers than smaller ones?
Grant: We hold your hand if you’re a larger publisher and help you with the templates. For example, we think Fox Sports can extend their brand into search by creating a sports search engine. Some publishers get it right away—in the old days, many Web sites started as adjuncts to magazine publishers, for example, but they quickly became valuable in their own right. The same will happen if Playboy decides to do a search engine.
Mike: How does the technology work?
Grant: It’s really all these experts against the mainstream search engines. The publisher starts by providing information to us—the more provided, the better our first version of their search is. Then we observe the searchers—if someone clicks on the #11 result, then we might move it up on the list [the next time]. We experiment with several different search filters and see which ones work better over time. We also have the concept of a wiki where anyone can edit or delete a search result and the moderator [who works for the publisher] can choose whether to accept the change. We love the wiki model for a search engine rather than the old centrally-planned model. We think that if you give searchers control, they’ll do cool, great things.
Mike: I’ve played with Swicki at the Popular Mechanics site. Does this technology get used as a Web site search also or only as a subject-based search?
Grant: We don’t really sell a site search engine—it’s used to do a subject for a community, a vertical search for a lot of different sites. Searchers don’t see what sites are searched (or change them) but very shortly we’ll allow publishers to grant moderator authority to searchers and allow input into more than just the search results—the way filters are set up, for example.
Mike: Do you think of yourselves as the anti-Google?
Grant: Well, you are completely into it with your line of questioning. We’ve thought about that a lot but we don’t want to promote ourselves as a destination in competition with our publishers. I also think that [community approaches] work better for our focused search than for generalized search—link analysis may work better for that.
We can solve the 10% of searches that can’t be solved with mainstream search by creating a vertical knowledge base. One of the advantages is that when you are on a rugby Web site, you have your rugby hat on, even though you might be a parent, too. Mainstream search engines don’t know what hat you are wearing [for every search].
Mike: Thanks for taking the time to talk to us, Grant.
