How Google Searches Itself

Google has become one of the hottest companies in Silicon Valley by helping millions of Internet users search the Web smarter and faster. But how does this wildly popular search engine find the new ideas that will keep its business moving forward? By ''googling'' itself.

Most Fridays at Google, the search-engine company in Mountain View, California, Marissa Mayer and about 50 engineers and other employees sit down to do a search of their own. Mayer, an intense, fast-talking product manager, scribbles rapidly as the engineers race to explain and defend the new ideas that they've posted to an internal Web site. By the end of the hour-long meeting, six, seven, or sometimes even eight new ideas are fleshed out enough to take to the next level of development. Some of those ideas might become new features on Google, new code or search algorithms, or a new way to juice up the Google home page. "We really jam in there," Mayer says.

And jam they must. Google has taken its place as the leader of Internet search engines, answering 40% of the estimated 375 million queries thrown out on the Net each day. But newcomers (such as Teoma and WiseNut) are looking to unseat Google by promising solutions that are better and faster.

To stay on top, Google needed a stream of new ideas. But where could the company look for them? And more important, how? "We always had great ideas, but we didn't have a good way of expressing them or capturing them," says Craig Silverstein, Google's director of technology. Mayer's proposal: Search for ideas in the same way that the Google search engine combs the Web. Google's Web searches succeed because they roam far and wide, scouring billions of documents. Also, search results are ranked by relevance (taking into account how many links a page has, among many other factors), and they come back fast.

Google's idea search starts with an internal Web page that takes minutes to set up. Using a program called Sparrow, even Google employees without Internet savvy (there are a few) can create a page of ideas. That enables the company to cast its net across its 300-plus employees. "We never say, 'This group should innovate, and the rest should just do their jobs,' " says Jonathan Rosenberg, vice president of product management. "Everyone spends a fraction of their day on R&D."

The easy-to-use intranet also benefits those tech-savvy Google employees who aren't always the most vocal participants in meetings. "There's an engineer here who's very quiet," says Mayer. "He has lots of ideas, but he feels better about writing them down. Once his thoughts are out on the intranet, they get discussed, and by the time the forum comes around, I can get him to come out of his shell."

Mayer combs the site daily, searching for relevant ideas. She digs out the ones that generate the most comments and that seem the most doable. Relevance isn't necessarily measured by how much money an idea makes; it's more about making Google searches better. "Sales may say that we need a certain feature," Mayer explains. "But great technology usually comes from somebody who's spent a year hacking a problem. You can't force technical innovation."

In the Friday meetings, Mayer insists on speed. The sessions are kept to one hour, and individual presenters never get more than 10 minutes. But everyone knows that the conversation won't end when the meeting does. Promising ideas are quickly outlined on the intranet site. Usually, the person who came up with the idea is put in charge of turning it into a feature. "I never have to hammer on people," says Rosenberg. "They showcase their ideas and then move on them."

Two recent examples include a news-search feature that debuted earlier this year and a pilot project that keeps track of persistent searches on the Web. "You can take Google's temperature just by going to the intranet site," Rosenberg says. "It's a window to everyone's soul."

Add New Comment

1 Comments