Google’s approach to software increasingly centers around personalized recommendations and assistance. It will filter your email and surface only the messages you need to see, pop up alerts if your train line goes out of service, and even chat with you about the latest scores and trades from your favorite sports team. Today that same approach comes to Google’s music service, Google Play, which is being revamped to focus on contextual recommendations.
Using machine learning and clues like location, activity, and weather, the app tries to serve up a smorgasbord of playlists that will match your mood and moment. "At a high level, Google’s mission is to make the world of information really accessible and useful to people," said Elias Roman, the lead product manager for Google Play Music. Google Now might do that by suggesting you leave for the airport early to avoid bad traffic on the highway. "There is a role for Google to play in music that is similar."
Every time you open the app, it completely refreshes what you see, trying to serve the perfect selection. Tapping the app on Saturday night I got options for a cocktail party playlist and some serious dancehall jams. When I opened it again late Sunday afternoon, I got offered a playlist for relaxing at home and another for cooking dinner. The interface, based on cards, feels a lot like Google Now, and so does the approach to just-in-time doses of pertinent information, in this case the tunes you need.
I am not a Google Play user by habit, so the service doesn’t have a ton of information about me. If it did, the machine learning algorithms would try and get even more specific, putting a workout playlist in front of me when it sees my location matches the gym I visit every week, and swapping that out to show me a playlist for focusing when my location matches my office. If I traveled to a new country for the first time, it would pick up on this, and offer music to accompany my adventure or get me into the swing of local culture. Google Play Music relies on a diverse range of datasets connected to my Google account: from search history, to maps, to YouTube, and beyond.
I’m a big fan of Spotify and the products it has put forward in the last year around discovery. But while Discover Weekly and Release Radar have introduced me to a lot of new music I love, they aren’t meant to match a certain mood or activity. They show up every Monday and Friday morning, and then they stay static until they refresh seven days later. Spotify has playlists for different moods, but they aren’t refreshed nearly as often and don’t have as much context about exactly where I am or what I’m doing to draw on.
The team behind Google Play Music came out of Songza, a New York City startup. That company was all about building playlists that matched a certain mood or activity, but users had to tell the app how they were feeling or what they were doing. "I’ve always been interested by the idea, is there a way to scale a personal DJ, someone who follows you around and knows exactly what you want to hear," says Elliott Breece, one of Songza’s co-founders, along with Roman.
Google allowed them to achieve that scale by combining Songza’s approach with a vast trove of information on users and powerful machine learning algorithms. It also made it possible to serve up a recommendation without prompting the user to input their current mood or activity. "If you want to use data to improve people’s lives on a daily basis, music is the perfect fit," says Breece. "So many people spend so much time with it every day."