Inside Spotify’s Hunt for the Perfect Playlist
If you use Spotify, it knows what you like and what you don’t like. It knows you like alt-rock, that you love the country songs you claim you hate, and that for all the game you talk, you never listen to that deep-cut Dylan record.
To prove it, Spotify is introducing a new feature today called Discover Weekly. It’s a completely personalized playlist (updated, you know, weekly) full of songs Spotify is pretty sure you’re going to want to hear. Some hits, some new stuff, that Foreigner song you’ve never been able to get out of your head. My first 29-song list has a little more girl-pop than I might have picked—but I’m digging it.
Discover Weekly is just a playlist, but it’s not just a playlist. It’s the latest move in a long-term plot to turn Spotify from a giant library of songs into something more like the always-on soundtrack of your life.
It’s no longer enough for a music service to offer unfettered access to 30-plus million songs for a monthly fee. That’s table stakes now—it’s assumed. Now the race is on to become the place for music discovery. This problem has spawned a laundry list of solutions. Apple’s music-discovery machine is Zane Lowe, the loud, shouty British man reminding you Beats 1 is always on and worldwide in between spins of Pharrell’s latest inescapable earworm. Apple talks down its nose at companies like Pandora, whose famous “Music Genome” breaks songs into their many component attributes and algorithmically assembles endless playlists of songs based on songs based on songs you like.
It’s easy, but incorrect, to dump these approaches into two buckets: humans and machines. If you do so, you’ll piss off everyone at Spotify, another of the implicit targets of Apple’s “people are better” campaign for Apple Music. At Spotify, the team says, they’re most interested in learning how to perfectly blend man and machine. That’s the only way to get to the real goal: to build better playlists by any means necessary.
The Hunt for the Perfect Mixtape
Spotify has this playlist called “Your Favorite Coffeehouse,” which I listen to constantly. With 1.2 million subscribers, it’s one of Spotify’s most popular playlists. The soothing mix of warm guitars and wistful, someday-life-will-be-perfect lyrics is just the right way to settle in at work, or lean into the mid-afternoon doldrums.
Doug Ford, Spotify’s director of editorial and music programming, made that playlist. He doesn’t want you to know it, though—he wants you to remember artists and songs, not his name and voice. When anyone can make and share playlists, he wonders aloud, “What makes me better than a kid in Ohio? I just don’t get it.” It should be about the music, not the DJ.
Your Favorite Coffeehouse began with the hypothesis (you hear that word a lot when you talk to people at Spotify) that you could replicate the experience of sipping a latte in a comfy chair while reading a book, without having to fight for a seat or spend six bucks on a cup of joe. “I wanted to make it so no matter where you were, you could invoke that feeling,” he says. But he didn’t just want to include the staples, the John Mayers and Norah Joneses of the world.
Playlists are made by people, but they live and die by data.
“It wasn’t driven by known artists,” he says. “In fact, I intentionally tried to include artists I’d never even heard of, and songs that were maybe B-sides or were never released.”
To start building the playlist, Ford turned to an internal Spotify tool called Truffle Pig. Jim Lucchese, CEO of The Echo Nest (which was also acquired by Spotify) refers to Truffle Pig as “Pro Tools for playlists.” It’s part of a version of the Spotify app that’s only available to employees. It lets them build a playlist from almost anything: an artist’s name, a song, a vague adjective or feeling. You tell Truffle Pig you want, say, a twangy alt-country playlist. That’s enough to get started. Then you refine: “Say you want high acousticness,” Lucchese says, “with up-tempo tracks that are aggressive up to a certain value. It’ll generate a bunch of candidates, you can listen to them there, and then drop them in and add them to your playlist.”
The Echo Nest’s job within Spotify is to endlessly categorize and organize tracks. The team applies a huge number of attributes to every single song: Is it happy or sad? Is it guitar-driven? Are the vocals spoken or sung? Is it mellow, aggressive, or dancy? On and on the list goes. Meanwhile, the software is also scanning blogs and social networks—ten million posts a day, Lucchese says—to see the words people use to talk about music. With all this data combined, The Echo Nest can start to figure out what a “crunk” song sounds like, or what we mean when we talk about “dirty south” music.
Once the editor clicks publish on the playlist, listener data starts flowing in immediately. The playlist changes just as fast, as the team sees which songs people love, which ones they save to other playlists, which ones they skip, and which ones make them ditch the playlist entirely. Are there two songs people particularly love? Maybe those should go up top. “If we see that it’s generally good,” Ford says, “but there are certain areas where it’s faulty—like maybe some of the songs are bad, or the leadoff song isn’t representative of the rest of the playlist—we’ll try to refine that and give it a shot.” Playlists are made by people, but they live and die by data.
The editorial team at Spotify is currently 32 people, spread throughout the world. Armed with this data about music, along with constant feedback about what people are and aren’t listening to, they can build and refine playlists to perfectly match their initial hypothesis, no matter how particular or complex. They’ve created more than 4,500 so far, everything from “Dinner with Friends” to “Late Night R&B” to “Metal Domination.” Every imaginable kind of music, for every imaginable mood and situation.
There’s another part of the equation, though. Let’s assume—and this is a big, not necessarily true assumption—that every time you open Spotify, or any other music app, there is a perfect playlist waiting for you. Somewhere. How do you find it? More importantly, how does it find you?
One Shining Moment
When Shiva Rajaraman joined Spotify after a long run as a product lead at Google and YouTube, he immediately noticed something: people weren’t listening to music the way they used to. They weren’t creating and organizing a huge library; they were building mini-libraries, playlists, each one for a different moment in their lives. It began with things like workout playlists, which people tweaked and updated far more than their overall library.
It got Rajaraman and his team thinking. “What if we took this to its full conclusion,” he asked soon after, “where instead of orienting around this idea of having music which you put in a library, we orient more around your life?” The overall notion of Your Music Taste is actually broadly irrelevant—what really matters is what you want right now. Rajaraman ticks off a few examples: “Music that might kickstart your morning, music that might help you run faster in the afternoon, music that might help fall asleep at night.”
These are what Spotify calls “Moments,” which are the organizing principle for the next phase of Spotify. Spotify is beginning to read your context—your location, the time of day, and more—to make deeply educated guesses about what you might want to listen to. You always run at 7 am, before work; Spotify’s going to start showing you running playlists at 7 am. In the morning, Rajaraman says they’ve found most people like uplifting music, so they’re starting to show users playlists like “Have a Great Day!” End of the day, you’re heading home, maybe you want “Evening Chill” to mellow you out a bit. It’s 2 am and you’re still listening to Spotify? You’re probably drunk, so here comes Avicii.
Spotify is completely redesigning its apps to bring recommendations front and center.
Combined with what the company calls your “taste profile,” an internally kept list of traits and types of music you tend to enjoy, Spotify can automatically refine and perfect recommendations just for you. When you open the app, you’ll see different playlist choices than I do; those playlists themselves could even be customized just for you. That’s where Discover Weekly comes from: It’s your taste profile, brought to life in two dozen or so songs each week.
Spotify is completely redesigning its apps to bring recommendations front and center. It’s even bringing in new content to suit those moments, like video clips from Broad City and podcasts from RadioLab, because sometimes music isn’t what you want at all. It’s even creating new, original content, like a Tiesto track that automatically adjusts its tempo to match your running speed. Rajaraman talks with great excitement about a newscast he’s thinking about, which works equally well whether you choose to watch or listen to it, and keeps playing even as you switch the screen on or off. Spotify is fundamentally a background app, which you start and then exit, but he wants to change that. “When do we lean in to show you something visual?”
They’re asking the same question about showing you something new, too. Everyone at Spotify loves the idea that the company’s role is equal parts push and pull—to give you the music you want, and to show you some you never knew about.
It’s a lovely idea, to think that someday Spotify or Apple Music might be so smart that the service knows exactly what you want to listen to or watch, and all you ever have to do is open the app and press a giant play button. But that doesn’t work. People are finicky, especially with music. Rajaraman knows he can’t be perfect all the time, and knows just as well that there’s nothing worse than playing the wrong song at the wrong time. He and his team just want to be the place you go when you need something to listen to, whether you know what it is or not.
If Spotify can pull it off—or if Apple Music, Slacker, Rdio, Deezer, Pandora, Google Play Music, or whoever else can beat them to it—they’ll have built the first app you open every time you put on headphones, because you trust it to have what you want even when you don’t know what you want. In the increasingly volatile, ever-changing music business, that’s the only powerful place to be.