Eight times a day Google asks test subjects about their information needs. Their replies can be sobering.
Google search really isn’t threatened by competition from other search engines. But the people on the search team constantly worry that they may be falling short in satisfying the needs of their users. To address that problem, of course, Google needs to know what those needs are. One way to do this is by examining the logs to see what queries are unsatisfied. But there are lots of things people want to know that they aren’t asking Google about.
How does Google know what those needs are?
It asks them.
Every year since 2011 Google has run an annual study to learn what people really, really want to know, whether it’s something Google provides or not. It’s called Daily Information Needs, but the psychologists at Google involved with the project just call it DIN.
Here’s how the DIN study works: Google recruits subjects who agree to report their information needs to Google on demand. Eight times a day Google randomly pings them, and they instantly respond with the questions they want answered at that moment. At the end of the day, subjects compile a summary of their needs, noting what, if anything, they did to get their questions answered and whether they were successful. The study began with 50 people in 2011, grew to 1200 in 2012, and this year has a similar number. In addition to those US numbers, Google runs the study in a number of other countries.
It turns out that people have a wide variety of needs. In last year’s study, Google wound up with over 25,000 different needs. It organized those into a thousand more generalized information needs, and further refined that number into 21 broad categories. Ranking those categories, Google learns where it should focus its search improvements.
To be sure, some of the things the DIN subjects want to know may not be answered by Google for a long time: a fairly frequent information need is posed along the lines of, “Why are men such jerks?” or “Why won’t my husband grow up?” But other needs are already providing valuable pointers to Google as it strives to make search more useful. In the most recent study, transit needs helped rocket that category to the top of Google’s product priorities. Another big information need is, unsurprisingly, weather.
But even such clearly defined categories can be a little tricky, as people in different countries have different needs, even when it comes to weather. Cultural differences weren’t hard to perceive between the 1200 Americans studied in the last DIN and the 300 Japanese subjects. Take the transit category: In the US, people want to know when the next train will show up. In Japan, people know the trains are reliable but need more information on negotiating the complex system itself. For weather needs, Americans want the information to help them dress appropriately and plan their commute. The Japanese want to know how much moisture is in the air, and when sunlight will dominate—because they more often hang laundry outside.
Identifying and analyzing those results has helped Google prioritize its projects. For instance, in the first study, one of the highest ranked general categories was a desire to know “how to” perform certain tasks. So Google made it easier to surface how-to videos from YouTube and other sources, featuring them more prominently in search. That move helped drop that category in subsequent DINs. Overall, the DIN has led, according to Google psychologists John Boyd and Kathy Baxter, to two new voice actions, 17 new Google Now cards, and 22 improvements on existing cards. (Since our interview, Boyd has left the company.)
One peripheral outcome of the study was a reminder of how full a picture of the human situation comes into focus by what people ask Google. Running the DIN in recent years, the researchers saw that the needs of their subjects all too accurately reflected a troubled economy. “This study spoke so much to me about people living on the edge,” says Baxter. “There were so many queries of How do I find a job?” One in particular seems to haunt her, a series of DINs that began with a question, How can I make a quick $200? followed by Where is the courthouse? And then How do I find a bail bondsman?”
It was a human train wreck written in information needs. “She was turning to Google to answer what were clearly important questions in her life,” says Baxter.
Another information need struck the researchers: an abundance of shift workers, who were perpetually in doubt about which hours they were working, or whether someone was covering for them when they had to switch hours or miss a shift. “For us, the question is, when is my next meeting? For them, it’s, when do I go to work next? It’s not something they can pull up—they have paper calendars!” says Baxter. In this case, the researchers met with a Google Calendar team in Zurich and brainstormed new features that would meet the needs of these people.
And the woman who needed $200, fast? Neither Boyd or Baxter recall whether she got the money. “How to make two hundred dollars is probably tough for Google,” says Baxter. “But how to get to a courthouse and a bail bonds person, we can handle that quite well.”
Inside Google’s massive effort in Deep Learning, which could make already-smart search into scary-smart searchmedium.com
Behind the scenes as the Internet giant launched big initiatives to keep its flagship product viablemedium.com