The Art (and Science) of Music Playlists
In a world where access to music is no longer an issue—nearly every musical artist (big or small) is featured on a streaming service—curation has become a key differentiator for music streaming services. The method of curation, however, is much more varied—some platforms rely on people to lead the effort, others rely on algorithms. Here’s a quick look at what these automated curation features do, how they work, and what they can teach us about the intersection of art and technology.
“The only song as important as the one you’re listening to at that moment, is the one that follows it.”–Jimmy Iovine, Head of Apple Music[1]
What is automated music curation?
Automated music curation is a set of rules used to predict and deliver music you will enjoy. And while there are a number of different approaches, automated music curation methods generally fall into one of two camps:
- Human-driven: Experts prescribe connections between certain songs, artists, or musical styles, asserting that if you like “A” you’ll probably like “B.”
- Algorithm-driven: Uses tags from songs you’ve already listened to—such as the songs’ classification, era, and musical characteristics—and matches them up with similarly tagged songs.
While most music curation methods rely on a mix of human-driven and algorithm-driven approaches, the terms are still useful for identifying which of the two takes the lead.
Human-driven music curation
“Algorithms alone can’t do that emotional task. You need a human touch.”–Jimmy Iovine[1]
One of the best examples of this approach to automated music curation is Apple Music, which launched in 2015.
Apple Music’s approach to music curation harkens back to radio, with a number of full-time employees working to produce playlists. In less than two years Apple has already delivered more than 14,000 playlists.2 For comparison, Spotify has been in business for almost nine years and has only delivered around 4,500 playlists.[2]
The reason Apple Music chooses to focus so heavily on playlists? It’s estimated that playlists account for almost 20% of music streaming and growing.[2]
Apple’s approach to human-driven curation is to employ around a dozen fulltime lead curators, each with a designated genre expertise, to elect their favorite artists and songs and place them in prominent locations across the service to ensure people encounter them. To find new music, curators often turn to music blogs, early performances, artist managers, producers, label representatives, and more.
Services using human-driven music curation:
- Apple Music: Curated by leading music experts who were handpicked by Apple
- Google Play: Much like Apple, Google provides a swathe of human-curated playlists, then employs massive algorithms to decide when to show a playlist to you, taking into account time of day, time of year, current events, etc.
- Pandora: One of the most familiar automated music curation systems, Pandora employs music and genre experts to tag songs and write the algorithms. Pandora is one of the only services to tell you why a song is showing up in your playlist (e.g., you like bass-heavy music).
Algorithm-driven music curation
In contrast to human-driven music curation, which is bound to the ebbs and flows of individual opinions, algorithm-driven music curation relies on the collective opinions of millions of users to provide its suggestions.
For example, Spotify’s algorithm takes the historical data from its more than 75 million users and uses it to form intricate connections between artists and songs. The benefit of this approach is that the algorithm can be programmed to take bigger jumps or risks than a human-curated playlist might—delivering radically different (though equally enjoyable) results. Additionally, by focusing on the objectivity of user data, algorithm-driven efforts are also able to overcome the main problem with human-driven curation: the need to satisfy everyone.
While you can try to fine tune human-driven curation to a number of different audiences (such as Apple Music with its 14,000+ playlists), the goal is still to appeal to as many users as possible. With algorithm-driven curation, every playlist is unique to you—and any risks it takes are a result of your own behavior.
Services using algorithm-driven music curation:
- Spotify: Led by a relatively new service known as Discover Weekly, which provides each user with a unique playlist of a few dozen songs each week, drawn from a data layer formed from its 75+ million users
- Google Play: Yes, Google plays both sides. With the release of their Google Music Intelligence this year, Google now employs machines to listen to music and pick out specific qualities for curating playlists—automating the human-driven processes of other services.
Human vs. Algorithm: Which one is best?
Music curation is as much an art as it is a science—and both human- and algorithm-driven methods will have their successes and failures. As a result, neither is necessarily better than the other. They simply provide two means to the same end: an expanding musical palette.
What makes them so important is the way they’re attempting to manage the subjectivity of art and the objectivity of automation. While the two may seem irreconcilable on the surface, the success of both methods helps to highlight the similarities between the way we experience the world on our own and the way technology can be used to classify and share the pieces that make up those experiences.
[1] Apple, June 2015 Live Event
[2] 9 to 5 Mac, Behind the scenes of human curation at Apple Music, as companies up the ante on licensing
Apple Music is a registered trademark of Apple Computer, Inc.
Spotify is a registered trademark of Spotify USA Inc.
Pandora is a registered trademark of Pandora Media, Inc.
Google Play is a trademark of Google, Inc.
