Google Universal Search - The New Battle for Verticals

The way people find information on the web has been changing. First it was directories. Then it was algorithmic search. Now we have vertical search-engines. Finally we have "universal search" from Google and Ask. What is going on here? Is there some other trend underlying these changes?

Directories and human-edited encyclopedias are great for finding pages on any single topic. If you want to learn about "mutual funds", you will get excellent information from Yahoo's directory, dmoz, Answers.com, and Wikipedia. But if you wanted "no-load mutual funds with emphasis on small-caps" you might want to use Google, Yahoo-search, or Relona Calculus.  

Initially users entered one or two word queries, and in those situations human-edited tools performed well. But over time, users started entering longer queries. Yahoo reported in a 2006 analyst call that the length of the average search-query has increased from 1.2 in 1998 to 3.3 in 2006. That is a huge increase! 

By indexing dmoz directory, Wikipedia and Answers.com, an algorithmic search-engine like Google could effortlessly "envelope" the single-topic tools. This meant that users remained with Google for both simple queries as well as more complex ones. Tools like AskJeeves (the human edited version of 1998) that excelled at single-topic queries could not compete against the "universal-access" provided by Google. In other words, tools built for complex queries were able to subsume the tools designed for simple queries.

If users had continued to enter short queries, an engine like Google would have provided no advantage over something like AskJeeves. But users did want to enter longer queries, and that has led to the present dominance of algorithmic search-engines. 

But recently, a new class of tools has appeared. These are the "vertical information portals" such as maps, yellow-pages, book-databases, restaurant reviews, people-finders, and job-search. To maintain dominance in the presence of such engines, Google/Yahoo/Ask/Microsoft will need to "envelope" these databases in the manner that they now envelope wikipedia and dmoz. But these are not simple to envelope - indexing a map and figuring out exactly when a map should be embedded within a search-results page is not as simple as indexing wikipedia.

In terms of “short-head”, “medium-body” and “long-tail”: For the one or two word "short-head" search-queries, wikipedia and answers.com do very well - but because algorithmic search-engines envelope these portals, most users stick with their favorite search-engine. For complex "long-tail" queries, Google performs very well. But a battle for dominance over the "medium-body" queries is now beginning.  

Vertical portals currently excel at "medium-body" queries. Users have been willing to bookmark and use a few of these verticals directly. Universal-Search is Google's first attempt to envelope the verticals.

My guess is that the vertical portals will eventually get enveloped by the search-engines. But it is not clear whether Yahoo/Ask/Microsoft might be able to provide better quality "universal search" than Google. The future will tell who wins this battle. 

References: There has been a lot of discussion about Google Universal Search in other blogs. Check out the following:

1. Ask's approach to universal search (adotas)

2. About the relevance of Google's universal results (MarketingPilgrim) and (search marketing sage)

3. Ask and Google's approaches ("threeminds organic")

4. Maps in Google (Small Business SEM)

Comments

[sujith]

Kumar is bang on when he says long tail is the key when it comes to differentiate between similar search engines. Chris anderson has been shouting from roof tops about the significance of long tail for quite some time now. I would rather go through the last 5% of a search dump to understand how fairly my search is doing. I don't see a reason why it was difficult for search engineers to understand why the intent of the user is more important than what we derive from the user input. Intent may not even come from what a user types in the small search box. If i were to derive the intent of the user, I would start concentrating on collective intelligence (or wisdom-of-crowd), pattern matching, machine learning and NLP. I am sure if relona touches any or all of these subjects, it is on the right track.