Future of search rides on relevance

CAMBRIDGE, Mass.--Search engine providers are working to catalog every corner of the Web, but what they really want is to get a better idea of what's going on in your mind.

A panel of industry executives gathered to speak at the Harvard Business School's ongoing Cyberposium conference here Saturday roundly endorsed the idea that making search tools more relevant in customers' lives will be the most important factor in driving their companies' success.

"Our future products will be defined not by what we feel users need, but rather by what sort of information they really want to find," said Rahul Lahiri, vice president of search and product management at Ask Jeeves. "People are usually coming to us looking for a specific answer. The focus for us can't only be on adding new kinds of content, like audio or video, to search--it has to be centered on what the users are really looking for."

Other companies represented at the conference included search heavyweights Google, Yahoo and MSN. And despite their different approaches to developing search tools, the executives here agreed that it will be ordinary folk outside the industry that have the most influence on its continued development.

Initiatives focused on the push to create a more personalized search engine are already under way at all of the companies represented at the conference. The experts said that garnering more involvement from people doing searches, and persuading those people to trust the companies with greater amounts of personal data, will be crucial to future search technology.

Bradley Horowitz, director of media search at Yahoo, said that his company's portal approach, in which it offers an array of personal services such as Web-based e-mail and shopping, in addition to search capabilities, is helping to convince customers that they can help shape new offerings.

"The user is expanding the amount of personal data they share with us in the form of their address books, e-mail accounts, or their shopping habits," Horowitz said. "By gathering this information we already have and studying that behavior, we can see a significant opportunity to apply the existing user relationship into new tools."

And the personalization trend won't apply just to creating new search engines, according to the experts. By allowing people's online habits and preferences to influence advertising, a key source of revenue for search companies, the speakers said, everyone involved will benefit.

To many eyes, Google has led the way in tailoring its strategy to meet the needs of both users and advertisers, with the success of its AdWords and AdSense programs. Even the executives representing Google's rivals in the discussion conceded that by clearly separating its search results from its ads, the market leader made sizable headway in winning users' trust.

"Google led the way in clarity in advertising," said Mark Kroese, general manager of information services at Microsoft's MSN. "We weren't separating results (from ads) a year and a half ago, and since we've begun doing so, the response from both users and advertisers has been huge. Google proved that if you have clarity, people respond."

Most of the executives conceded that the technology to build personalized search tools already exists, and they said the fight to persuade people to share more personal information is what stands in the way of new products.

"(Personalization) isn't an area where the technology isn't ready, where there's a need for a lot of innovation," said Ask Jeeves' Lahiri. "The question is, are people willing to give up (more information) to get a better search engine back in return? Only time will tell."

Barking up the wrong tree with personalized search

In my opinion, if the search engine marketers think people need search engines that know more about their needs, habits, and preferences, then they are totally missing what people really want.

In most cases, search is based on very diverse short term needs. For example, one moment I am looking to buy a house, next a new pair of shoes, the next I need to learn the side effects of a new prescription, the next I need an answer to a legal question. A few days or a few weeks later, my needs are completely different. For any of these example searches, how will it help to know more about me personally? I can tell you now that it won't. Are you going to record my preferences for a 2 bedroom versus a 3 bedroom home? Well, next time I search I may want a 4 bedroom home because there is another baby on the way. You can't keep up with me in a relevant way. You will be wasting your time and my time.

Instead, I suggest that search engines should focus more on understanding what I am looking for in a particular query. I'm suggesting that they need to go way beyond merely matching keywords. The trouble is that this is a very difficult problem and there is no easy answer, like collecting more personal information about me.

There are two places at which understanding must occur:
1.) In my query text there is an idea (a concept) of what I want to find.
2.) In the database that is being searched, there is a massively large collection of ideas (concepts), of which only a small part is relevant to my query. Let's call that small part X.

So, what I really want and need is for the search engine to give me as much of X and as little of non-X as possible.

I'm sure that this is so obvious it is almost not worth stating, but I am mentioning it because I think the search engine marketers are losing sight of this obvious goal. They are looking for something else because they think that this part of the problem has been taken as far as it can be taken. Yet, this is far from the truth. Current search technology barely scratches the surface of the true possibilities. Continue to focus on the real problem and you will deliver real benefits and that will be the measure of success: deliver to me what I ask for (not what you think I need) and save me from having to look through hundreds or maybe even thousands of erroneous search results.

What's lacking is a steering wheel.

You've hit the nail on the head. Immediate information needs cannot be intuited from past activities or personal profiles. Our company's research has indicated that not only are short term search needs unique, but they change as the user learns more about their topic. They may start our wanting to know what ocular melanoma is, but quickly switch to wanting to know about the state of the art in treatment - a subtle shift that implicit personalization will totally miss. Yet that's what people do. What's missing is a relevancy steering wheel - a way to feed back to the search engine what's of interest and what's not.

Google is now too big to hear users voices.

Sadly, the demise of Google as a useful search engine is complete. A victim of its own success, Google was great at returning relevant links when it was a relatively unknown search engine used mainly by academics and IT workers. In the push to become the biggest, Google has indexed a lot of pages, but most of the links returned from any given search will be irrelevant. I now use Yahoo! at least as often as Google, and always check failed Google searches at Yahoo! because it is common that a failed search on Google! will be successful on Yahoo!

I first got clued in to the problem when I realized that Google wasn't indexing a hobby site of mine that had been on the Web for years. Even searching Google with the exact url turned up nothing. I e-mailed Google about it repeatedly, and have never received a reply, and the site has never been indexed. After discovering that problem, I began checking my Google searches against other engines and began to find many such omissions.

Google needs to get back to its roots, focus on relevance, and take a break from cranking out nifty new utilities that are of questionable value.

Word to the wise - if your search fails on Google, try another search engine.