Future Search

Imagine a world where you type something into a search engine and the result is a uniquely indexed, real-time display of links generated just for you. Alongside them are, of course, superbly targeted paid listings as tempting, if not more so, than the algorithmic results.

Judging by the public statements of some of the biggest search players, that’s exactly what they’re planning. Surely there’s more to the personalization plans than I’ve heard, but the gist of what’s endlessly discussed these days is new technology, such as what Google acquired with Outride and Kaltix, will enable search engines to “learn” from people’s previous searches. That learning would presumably allow the engines to infer searchers’ stage in the purchase cycle, along with their deepest, truest desires.

If only it could be so. The reality is no one — not even with the gee-whizziest technology — can read minds.

“The risk you take by being too presumptuous and not having the right technology is not giving the right answer,” said Jeff Weiner, senior vice president of search and marketing at Yahoo, speaking at AD:TECH last week.

If magical technology won’t solve the problem, what will? The only way to avoid being presumptuous is to ask searchers what they’re looking for. The tough thing about that is asking seems to mean abandoning the one-button interface Google pioneered — and the searching public loves.

My colleague and fellow ClickZ writer, Gary Stein of Jupiter Research, said he recently spoke with Tim Armstrong, vice president of advertising sales at Google, about the issue.

“He stressed that that [one-button] interface is critical to their success,” Gary told me. “They need to maintain that.”

None of the big players will divulge their plans, but a couple smaller companies I came in contact with this week have developed technologies that at least spark some thought about the subject.

Vivisimo, which this week made a deal with Infospace, provides “clustering” technology that groups similar search results together. Next to the regular search results, a set of clusters, represented as folders, appear on the left of the page. For an illustration, check out a demo on the company’s site that lays the technology over MSN search results.

“The challenge that you have right now is not information overload,” said Tony Philipp, executive VP of Vivisimo, “it’s information overlook.”

What’s especially interesting for search marketers is Vivisimo claims users click on 30 to 100 percent more paid listings when presented with search results organized in its hierarchical folders. The company came to the conclusion by conducting before-and-after tests on its own Web site and those of two small partners. Vivisimo didn’t divulge its methodology for the tests, but the results raise eyebrows.

One customer saw a 5.6 percent CTR before clustering and a 13.5 percent CTR afterward. The second saw a 4.2 percent CTR before and an 11.3 percent CTR after.

The second player I ran across is SLI Systems. The company counts NBC among its investors and customers. It’s working on something called “Learning Search.” In addition to all the usual ways to come up with search results, it learns from experience what users click on and moves these up in the results. It’s pretty similar to the way Google determines rank for listings in its AdWords program.

The latest interface enhancement from the major portals is the introduction of tabs, long a Google staple. Yahoo recently introduced a longer search bar on its front page and added tabs for images, the yellow pages, and products. AOL followed suit with a new “In Your Area” search, which searches AOL’s Yellow Pages. Its tabs read: Web, images, news, in your area, and people. Google’s tabs are the most numerous of the bunch, featuring Web, images, groups, directory, and news. (It hasn’t yet added Froogle to the mix, even though we’re beginning to hit holiday shopping season. The shopping engine is still a beta release.)

How about a “my” version of search results? Setting aside the privacy and security headaches involved in getting users to register, then keeping track of their preferences, it’s an interesting concept. One could at least ask users whether they’re male or female. Direct Hit reportedly did testing with personalized results back in 1999 and found, in a search for “flowers,” men wanted to buy and send flowers, while women wanted to order flower seeds or plants.

Will any of these ideas show up on the search engines of the future? Stay tuned. By then, Google may well have bought Microsoft and integrated the whole search function, including paid listings, into the operating system. But that’s another column.

Related reading