Yahoo Expands Research Labs in Search of Personalized, Mobile Experiences
The rejuvenated research labs at Yahoo are investigating new forms of mobile hardware and new ways to predict what users want.
There are many conflicting opinions about what troubled Web giant Yahoo must do to turn itself around, but critics and company leaders at least agree on one thing: fresh ideas must be part of the solution. As a result, Ron Brachman, head of the research division Yahoo Labs, finds himself working closely with new CEO Marissa Mayer, who took charge of the company in 2012. “There is a lot of two-way dialogue with Marissa and her senior staff,” Brachman says. “They expect us to be the thoughtful leaders in innovation that can tell you where the world and technology are going.”
Yahoo Labs has significantly expanded since Mayer left a senior position at Google to take on the job of CEO in late 2012. She inherited a group that had withered during the five-month tenure of her predecessor, Scott Thompson, a period that saw some leading figures of Yahoo’s New York City lab quit to join Microsoft (see “Microsoft’s New Lab Hunts Value in User Data”).
Mayer sought to rebuild the research effort, asking that 50 new PhDs be hired in 2013. Brachman says the company passed that figure in August and “significantly” exceeded it by the end of the year. “Like any organization, you adapt to the changing circumstances,” he says. “In the last year and a half, we’ve done another round of adapting.” Hires have been made at the Sunnyvale headquarters and established locations such as Barcelona, Spain, and Santiago, Chile;, the New York group is being rebuilt; and a new research team has been established in London.
Brachman stresses his group’s independence, but he has also shifted the Yahoo Labs research emphasis to align with Mayer’s priorities for the company. Mobile computing—an area that Mayer has publicly labeled a weak point for Yahoo and made the focus of her product teams and several acquisitions—is now a major preoccupation for the group. Yahoo Labs researchers also work more closely with the main business, says Brachman, with some working alongside engineers on products on a day-to-day basis. “We’re really in the trenches,” says Brachman.
Technology developed at Yahoo Labs is at work inside the well-received weather app that launched last April—it was the first new product released under Mayer, and she cited it as proof that Yahoo is on a new course. The app uses striking photos from Yahoo’s photo website, Flickr, as a background to weather forecast information; and it selects images that match both the location and conditions of that forecast.
The human editors who chose those images are aided by machine-learning software that originated with Yahoo Labs researchers, says Brachman. The software processes photos and associated metadata to look for signals about their quality and content; an editor makes the final selection. “Humans are in the loop but they’re aided by machines,” he says.
Other research is less likely to appear in a Yahoo product anytime soon. Some of the company’s researchers are thinking about what new forms of hardware might take off in the near and far future, including devices worn on the body. “Even though we’re not a hardware company, we need to be thinking about it, and even inventing it ourselves,” says Brachman. The idea is that if a new type of device does suddenly hit the mainstream, Yahoo won’t be slow to realize its importance—as it arguably was for smartphones.
The idea of dedicated academic-style research labs has fell from favor in the generation of large Internet companies that came after Yahoo. Google, for example, avoids that model and embeds fundamental research into its product groups. But Gary Flake, who was the first leader of Yahoo’s research effort, says that approach won’t work for the company because it doesn’t have Google’s reputation as an engineers’ playground.
“If Yahoo wasn’t doing research this way they’d be doomed,” says Flake, who left Yahoo to join Microsoft’s research labs, and later founded the online bookmarking service Clipboard, which was acquired by Salesforce in 2013. Without a separate research organization, Yahoo would be a dim prospect for leading computer science thinkers, who would fear losing their status as researchers or academics, says Flake. “It’s a long time since anyone has said that Yahoo is an engineering-led company,” he says. “A research group means they have a place to attract talent; the proposition is you can keep your academic career alive and have access to data you would never have in academia.”
Indeed, the data generated by the over 800 million monthly visitors to Yahoo’s sprawling Web properties, from e-mail to news and fantasy football sites, underpins another major preoccupation of Brachman’s. “The fact that Yahoo has so many people willing to share with us gives us a fantastic sandbox to develop highly personalized experiences that can anticipate your information needs,” he says.
Such technology is attractive to Yahoo as a business because it can serve two purposes: matching people to content, and matching ads to people. It also gives the company a toehold in the future of search, argues Brachman, despite the fact that Yahoo currently licenses the technology behind its search engine from Microsoft. “Once the world moved from directories to search engines, it looked like we had solved the problem of finding information, but we still have a long way to go,” he says.
Brachman has been thinking beyond the search box for longer than most in the computer industry. Prior to joining Yahoo Labs in 2005, he worked at the Pentagon’s DARPA research agency, where he was involved in the project that eventually spawned Siri (see “Intelligent Software Assistant”).
Yahoo Labs is already reshaping Yahoo’s homepage and other sites to show people news stories and other content that the data suggests they will like the most. Brachman says that a Siri-style approach is also in the works. “A Holy Grail would be an environment where all of your information and shopping needs are anticipated and you have something like a very smart, conversational virtual assistant,” he says. “That is missing from Siri.”
Although a Yahoo-branded virtual assistant seems far off, some of that thinking will show through in forthcoming Yahoo products. When Mayer was asked in an onstage interview late last year how Yahoo would attract new users, she replied, “We’re a personalization company.”