Every day 268 million people use Google (GOOG) to search for something. The query goes in, the company's software delivers back the most relevant links. The interaction is so simple—and the hidden calculation behind the results so complex—that it's no wonder people tend not to notice much about the process. Who bothers to ask the ingredients of a magic formula?
For all of its experiments with maps, books, e-mail, and social networking, Google is still an empire built on search. Ninety-seven percent of the company's $23.7 billion haul in 2009 came from advertising. While Google doesn't break out what each of its individual ad products such as AdWords or AdSense generates, multiple people within the company concede that Google is as dependent on the "Sponsored Links" generated by search queries as an oil nation is on its wells.
Since Google's 1998 debut, the search results page—where a home page query is returned with 10 suggested links on the left and multiple advertiser links on the right—has been through seven subtle redesigns. The most recent, in May 2007, saw the addition of images and video in what was dubbed "universal" search. On May 5, Google unveiled its eighth iteration, which Marissa Mayer, vice-president of search products and user experience, calls "particularly large and particularly important."
Google has long had advanced search capabilities, but they were difficult to find. The goal of redesign eight was to surface them and integrate them into the main results page. Users now get results with an extra column of tools to drill deeper into information. That means a query can be quickly refined to show only results from shopping sites, say, or just videos on a topic, or the latest news results. Add in a new logo and a splash of colorful icons on the left side of the page that guide users through the new options, and the look is noticeably different.
Given that the shift of a few pixels can affect Google's profits, why would the company ever mess with the most successful product in the history of the Internet? "The Web is always changing, evolving, and innovating," says Mayer. "It's important even for sites that people use every day and are very familiar with, like Google, to update their look."
It's not just the look that's been updated. Microsoft's Bing and social media sites such as Facebook and Twitter have made a case for the past few years that Google's search, based on ranking the overall relevance of a Web page, is outmoded, and that the future lies in an integration of relevance with real-time search. In December, Google conceded the point and announced it would begin indexing the Web in real time to help users organize the cacophony emanating from social media.
Some in the search industry are skeptical of introducing such complexity to the results pages. "Advanced search almost never works," says Jakob Nielsen, principal of the Norman Nielsen Group and co-author of the book Eyetracking Web Usability. "People don't want to use a search engine. They want to get away from a search engine. That's the reason the advertising works so well—the search engine is one site they want to get away from so they might go to an advertiser." In other words, Google has thrived precisely because it hasn't tried to envelop its users in a full-frills experience.
Still, the competitive landscape is changing. In March, Facebook surpassed Google as the most visited Web site in the U.S. Microsoft has unveiled its own plans for Bing, which currently comes in a distant third in the battle for eyeballs but is making a play in searches along topics such as travel and shopping. Bing recently rolled out a design that adjusts the user interface of search. So if it's clear a user is looking to buy a plane ticket, Bing morphs into a travel engine. "It's a very radical departure from the dominant paradigm of the link-based system on the Web," says Stefan Weitz, director of Bing. "For every query, presenting results back with the same user experience doesn't make sense. You want to book a flight or track a flight, why display a bunch of links? That's illogical."
All the tweaks Google has made complicate its algorithm—and none of them can be allowed to complicate its interface. "They have to try to continue to innovate and evolve the product but stick to keeping things simple," says Andy Miedler, senior technology analyst at St. Louis investment bank Edward Jones. "People have grown to love the simplicity of Google." Every previous redesign has negotiated the demands of simplicity and complexity while improving Google's bottom line.
Google allowed Bloomberg Businessweek access to the core design team of 16 staffers in March, as the company was fixing bugs and gearing up to roll out the redesign—a process marked by collaboration between people who create wild illustrations with dry-erase markers and mathematical types who write in precise code. The atmosphere was loose, though the pressure of the job—engage the new decentralized, socialized Web while still driving users to the advertisers paying the bills—generated its share of stress. "There's a lot of weight associated with the project," says product manager Nundu Janakiram, the point person between engineering, design, and the five other departments involved in the project. "Everyone feels invested in the search engine; it ties together the company. We have to get this right."
Every Google page must work seamlessly on countless different computer makes, connection speeds, browsers, and screen sizes. It also must be optimized to keep people searching and clicking as fast as possible. Just thinking about the technical requirements can drive designers mad. So when the team assigned to Search Results V8 first gathered in the summer of 2009 in a small room at Google's Mountain View, Calif., headquarters, they opted not to think about requirements at all.
Janakiram, a 24-year-old Stanford grad, Web designer Jon Wiley, a 34-year-old ex-improv artist, and two visual designers mocked up hundreds of different results pages that turned everything users know about Google inside out. "Anything we could move around, we moved around," says Janakiram. Navigation tools were switched from the left to the right side of the screen. Colors changed from Google's muted blue to variations with Crayola-like vibrancy. Link styles got tweaked and modules were inverted. One proposal crammed every pixel on the page with information. Another added display advertising.
The wild-ideas phase was a good mind-clearing exercise—and mostly just that. When it came time for the designers and engineers to move into a larger space together and get serious about the new look, the exigencies of reworking one of the world's most profitable Web pages couldn't be held off. "We tend to focus on an evolutionary model where you can bring users along with you vs. dropping them into some completely radical experience," says Wiley, who once ran the Web operation for Texas Governor Rick Perry and who joined Google in November 2006. The engineers try to view themselves as enablers rather than wet blankets. "Sometimes a designer will tell you what they want, and you'll tell them you can give them 90 percent of that with a quarter of the work," says Nate Gaylinn, the technical lead on the project. "If it's really worth it to have the right experience, then you have to put in the extra work. But if it's a minor detail that you can argue either way, we'll take the easier route."
Most of those details are hashed out at a daily "stand-up" meeting of 10 people or so, held at 4:07 p.m. (Google co-founder Sergey Brin once estimated that it took seven minutes to walk across the Google campus; it's now company tradition that meetings end on the hour and new ones start seven minutes later.) Everyone working on the project gathers standing up, to make sure no one gets too comfortable and no time is wasted during the rapid-fire update. Janakiram, balancing a keyboard on the back of his chair, takes notes throughout. "The whole concept of the stand-up is to talk through what everyone's doing," says Gaylinn. "So if someone's working on what you're working on, you can discover and collaborate, not duplicate." Every proposed feature is put through a user interface review conducted by members of the larger search team, who work nearby and police proposed changes to make sure they won't have an unfavorable impact on the search-engine infrastructure. "They're making sure we don't screw up," says Gaylinn.
By October 2009 the team had its first functional prototype ready for "dogfooding"—in which all of the company's 20,000 employees can opt to test a new product. The internal focus group members can log bugs or simply comment on the design. Google won't say how many staffers took part, nor will it release the text of any comments. Janakiram will reveal that at its peak the team received a piece of feedback every 10 seconds and that not all of the responses were helpful. Even Google employees, he says, will send a message to let designers know they think their work is terrible.
One immediate focus of the feedback was the search button: The designers had turned it blue. "We got a lot of people telling us they found the button distracting, or they usually hit enter to search, or their gaze was going to the box, not the results," says Janakiram.
In addition to dogfooding, Google researchers ran 19 test subjects through hour-long eyetracking sessions to monitor responses to various aspects of the new design. Eyetracking has come a long way since the early 20th century, when objects were glued to a test subject's eyeballs. Now multiple cameras are built into computer screens, and the technique provides detailed analysis of where someone browsing a Web page is actually looking. Users often read Web pages as two lines across the top followed by a vertical stripe down the left side of the screen, in what's known as the "F-pattern."
During eyetracking, the engineers and designers sat in an adjoining room and gawked through a two-way mirror while users tried out various incarnations of the new search results page. Each participant was given prescribed tasks to search for so that the Google team could see the page in action. In a test staged during Bloomberg Businessweek's visit, a user was told to imagine his niece was writing a report about Michael Jackson, and he was tasked with helping her find specific information. When users got distracted—as they were by the same blue search button the dogfooders objected to—or failed to grasp how they were supposed to interact with the page, it meant it was time to scrap an idea. In other instances eyetracking data helped designers decide between different colors and type sizes.
While the engineers and designers pored over eyetracking results, Google statisticians conducted months-long experiments with hundreds of thousands of live users. (One advantage of having 268 million users a day is that you can roll out new products to a fraction of them and still have the benefits of a large sample size.) Google wants more searches and more clicks on its Sponsored Links, and anything that slows the process or inhibits clicking can have disastrous effects on profits.
Toward the end of the redesign process the statisticians focused on the slightly expanded gap between the Google search box and the first result. "No human being will notice they spend on average 10 milliseconds longer to click on a search result," says software engineer Patrick Riley. "But if all users take 10 extra milliseconds, we need to be concerned." This analysis was fed back to the designers, who adjusted accordingly. "We take the quantitative and qualitative feedback together with our own intuitions about how the search results pages should be working, and we iterate on the design," says Janakiram.
The focus on metrics is a deeply ingrained part of Google's culture. Brin and co-founder Larry Page are both engineers, and they've been consistent in expressing a belief that form must always follow function. (Page has gone so far as to wonder why the home page needs anything except an empty search box.) That doesn't mean designers have to like it. Douglas Bowman, Google's first visual designer, left the company in March 2009—and lit the Web equivalent of a bonfire on his way out the door. In a post on stopdesign.com, Bowman claimed that at Google, data "becomes a crutch for every decision, paralyzing the company and preventing it from making any daring design decisions." He noted that one project was delayed when a team couldn't decide between two blues—so it tested 41 different shades between them. "I had a recent debate over whether a border should be three, four, or five pixels wide, and was asked to prove my case," wrote Bowman, who is now the creative director at Twitter.
Other ex-Google design stars acknowledge that the emphasis on data can be grating. Jeffrey Veen, who joined Google as design director in 2006, says "the designers I worked with were fantastic but very formally trained in human computer interaction rather than having MFAs. That frames how design happens at Google." Veen, who left the company in 2008, adds: "None of the colleagues I would want to hire would be able to get a job at Google because of the computer-science-based requirements."
Mayer, the vice-president of search products, doesn't buy the criticism. "We're not curtailing instinct at all," she says. "We need that to fuel the innovation engine that yields great designs from Google. What we can do is use the power of the Internet and the measurability of it to figure out in a scientific way if something is a good idea or not." Wiley, V8's chief designer, insists he's comfortable with the mission and understands that search design is not intended to wow users. "It's like lighting design or sound design in a movie," he says. "If you're sitting watching a movie and you say, 'that's really great sound design!' then that's probably not right.…I don't want someone to say 'look at this gorgeous search results page!' I want them to say 'where's my…?' And click, they've gone."
In the period leading up to the rollout, changes continued. Some were major: The launch date slipped five times. A plan to include a widget underneath the search query box so users could tailor results to their location was dropped after engineers decided the technology didn't perform consistently enough. A feature offering users a way to look for similar types of results was renamed from "Not Entirely Unlike" (an oblique homage to sci-fi author Douglas Adams) to the less whimsical, more intuitive "Something Different."
As the redesign was being unveiled to the world via a blog post penned by Mayer, the team continued working. For the engineers, launch is the most critical and nerve-racking time. "We'll definitely be biting our fingernails. You can't just turn this on with a switch," says Gaylinn. Even as the team breaks up and moves onto different projects, changes will continue. "In many ways the project never really completely ends," says Janakiram. "There are always improvements to be made."