A Brief Interview With Udi Manber, Google, On Universal Search

If you've been reading a while, you've seen my coverage of Udi's career, from A9 to Google (and before, though I did not cover his work at Yahoo or prior to that…). I pinged him a while back and he got back to me after the universal search announcement…

Udi

If you’ve been reading a while, you’ve seen my coverage of Udi’s career, from A9 to Google (and before, though I did not cover his work at Yahoo or prior to that…).

I pinged him a while back and he got back to me after the universal search announcement had passed (he had a lot to do with it…).

Here’s our brief interview:

How did Google make the decision to do universal search? What got you comfortable with the approach? It reminded me of the things we spoke of when you were at A9…

The project started way before I arrived at Google. What you see today is just the beginning, and it’s a culmination of many different pieces that came together recently. What got us comfortable are three things: First, the design of the infrastructure is solid. It’s scalable, measurable, and efficient. Second, we solved some interesting ranking problems, which allows us to mix results from many sources in the right way. Third, and most important, we put together a wonderful team that got it done. David Bailey and Dan Belov (from Engineering) and Johanna Wright (from Product Management) ( see http://googleblog.blogspot.com/2007/05/behind-scenes-with-universal-search.html ) lead the core team, but this was an effort of many teams over a period of time. Again, this is just the first step.

I’m very interested in the next steps. Without telling us too much (if you would like to, why, please do), what are the interesting problems in search right now that you feel well positioned to address?

As search gets better, user expectations rise even higher, and we need to improve at a faster rate. Most of our work still focuses on the fundamentals — making results more relevant, more comprehensive, for more users, in more languages. Much of this work involves pure algorithms, deep understanding of search and of the web, and just plain hard work. Just the way we like it. It is not sexy to the outside world and it doesn’t make headlines, but it has the highest impact. Most of the advances in pure ranking that we’re making aren’t obvious to users — they just find what they’re looking for more often and they take it for granted. Just the way it should be.

In addition to the fundamentals, we are involved in dozens of efforts in new areas. I wish I could brag about all the cool stuff we’re doing, but I won’t until it’s done.

Your colleague Adam Bosworth and I had a long chat recently about his work on Health. It strikes me one of the more interesting areas of search might be in domain specific search, like health. You can do things with structured results and deep knowledge of a person’s information needs in that kind of environment. Would you agree that this is an area where devoting significant resources makes sense?



Of course. What’s more important than health?

As you know, we improved the search results for health-related queries with Google Co-op, which launched last year. If you search, for example, for diabetes, we’ll offer refinements at the top of the page (treatment, symptoms, tests, risk factors, for patients, for health professionals, etc.). We are working with the major health organizations and we put special emphasis on ranking of health-related queries. A lot more work on health is being done by Adam and his team.

Given the approach of universal search, how does domain specific fit in? I mean, given that Google is bringing all search results to one place ( google.com) how does domain specific stuff like Health fit into the Google master plan?

The universal search vision is about including results of all types in the best possible way on one page — the main search results page. It does not mean that we will not invest in improving search for particular types of data, in fact it’s the opposite. By building the infrastructure for universal search and by giving every type of result the prominence it deserves we make it easier for the search group and other groups within Google to highlight the best results. We are also working on giving users specialized access to different types of data. The ranking and presentation of local information are different from that of, say, video or news. We now give users the option to play a video directly from the search results page and see a map directly on the search results page. Once you follow the link into our maps area, for example, we provide a very rich experience (with maps, satellite images, and now street view images, all integrated). One of the most important goals of the universal search project is to build new ways to handle specific data types, like maps, images, scholarly articles, or health.

Do you see Google creating stand alone destination sites around search separate from google.com?



My job is to improve search on Google.com.

Well defelected, Mr. Manber. Thanks!

3 thoughts on “A Brief Interview With Udi Manber, Google, On Universal Search”

  1. Mr. Manbar is an interesting study in the search space. I know many people that know him and the feedback’s always the same: “not” a nice guy, and a very “spotty” track record of tangible accomplishment. A9 is an unmitigated failure. Let’s hope he does better this time.

  2. Thanks for the interview, John. My biggest issue/concern with Google is not they will no longer give the best results (my guess is they will for some years to come) but that they will more and more act like a large corporation. I already cannot stand seeing the very visible Google Checkout images next to the businesses that “play ball with them.” This should have nothing to do with search results. But they know that eyes may be drawn to position three of the results even before position one do to the large visible image. I strongly disagree with this.

    They are also trying to direct people to their own list of vehicles and other areas in search. If we see more of them trying to use search to sell their other products, I think this will soon be their downfall. I don’t mind them having other products (hell, I use and like them) but don’t try to sell or promote them in search results. This will be why I use another tool, and I guess others. End of rant.

  3. Sorlen – good points. Time was that Google knew where you wanted to go (to find information), and Google sent you there. More and more, Google is integrating info / answers directly into the SERP, so that you stay at Google instead of leaving.

    Trouble is, I like the other tools out there, and I like that there are tools outside of Google’s direct control (inasmuchas they can be these days, anyway). I don’t want Google to answer the question “what should I do tomorrow” and I’m creeped out that they even want to be able to.

Leave a Reply

Your email address will not be published. Required fields are marked *