To most Internet users, if Google's dominant search market share can be said to represent a yardstick of satisfaction, Google's search service works just fine. But Google nonetheless feels the heat from startups and competitors that claim to have built a better mousetrap. It can't afford to rest on its laurels.
So Google constantly strives to make its crown jewels -- its search algorithms -- shine more brightly. As Udi Manber, VP of engineering at Google in charge of search quality, explained in a blog post last year, Google maintains search teams that focus separately on core ranking, user experience improvements, new features and interfaces, fighting Web spam, and special projects.
In 2007, according to Manber, Google introduced more than 450 search improvements, an average of nine per week. A company spokesperson said Google hasn't released a comparable figure for 2008. Nonetheless, search research never sleeps at Google.
"We're improving Google all the time and most of the time you don't even hear about it," said Ori Allon, technical lead on Google's search quality team.
The latest improvements come in the form of better search-term association and longer search results snippets.
The search-term association improvement helps Google understand when search terms are related to other concepts that don't necessarily contain the same words. Google makes use of this knowledge by providing searches related to the keywords entered at the bottom of its search results page.
Allon said that Google's better understanding of how search terms relate isn't so much semantic technology -- systems for understanding meaning -- as it is a matter of data mining.
As an example of how this search improvement might work, someone searching for "principles of physics" would see related search query suggestions that use the words "physics special relativity," "physics angular momentum," and "quantum mechanics physics," among others.
While IAC's Ask, Microsoft's Live Search, and Yahoo Search offer related search suggestions for the search "principles of physics," their suggestions cover less conceptual ground and, arguably, are less useful. Live, for example, offers as related search suggestions "laws of physics," "principles of science," "fundamentals of physics," "definition of physics," and "principles of chemistry," among others.
The lengthening of search results snippets for searches with lots of keywords represents an attempt by Google to provide searchers with more context. The goal is to help searchers understand what the pages at the end of search results links are about. Reducing visits to pages that don't really address a query means more satisfied users.
Allon said he couldn't discuss specific metrics that quantify how much these two search improvements affect user happiness, but he said they had indeed helped more Google users find what they're looking for. "We witnessed a significant increase in people who get to a page and stay on the page," he said.
Google was recently criticized by a departing designer, Douglas Bowman, who lamented the company's reliance on data to evaluate design decisions. While Google's dependence on data as a driver of product features may not lead to the most inspired aesthetics, data clearly plays an indispensable role in making Google's search responsive, effective, and innovative.
And Google is likely to keep relying on data as it continues to refine its search technology and other products. "We're doing a pretty good job [with search], but there's a really long way to go," said Allon.
InformationWeek has published an in-depth report on predictive analysis. Download the report here (registration required).