Since the birth of rock’n’roll in the 1950s, music has evolved, separated, stratified, and segmented into so many different streams. Today we’re faced by innumerable genres, sub-genres, sub-sub-genres, and sub-sub-sub-genres.
Some, like metal, goth, and punk seem eternal. Others have the half-life of francium-223. But someone has to keep track of all these different types of music because, well, that’s what humans do: we organize things into neat piles and categories.
So how does Spotify do it? New sounds and scenes are emerging all the time with names like “post-teen pop,” “downtempo Swedish endurance,” and “tracestep.” Who comes up with those names? And who formalizes these genres within Spotify?
Take a look at this Spotify blog called “How Spotify Discovers the Genres of Tomorrow.”
Spotify data alchemist Glenn McDonald turns numbers into music experiences.
Ask Glenn McDonald what he does at Spotify, and a wry smile will appear. “My occupation, strangely, is called ‘data alchemist,'” he says over a cup of coffee before a show in Boston—just over the river from Spotify’s Somerville, Massachusetts, outpost. “I worry about anything at Spotify that takes in numbers on one end and tries to produce music experiences on the other—I try to make sure that the numbers make sense, and that the musical experience makes sense.”
McDonald’s work helps power the Related Artists tabs on individual artist pages, as well as Daily Mix—and that’s not all. “I do a lot of strange projects on my own, looking for patterns and generating playlists,” he says. “The website Every Noise at Once is the largest manifestation of that.”
Every Noise at Once is a treat for anyone who likes to fall down online rabbit-holes of music: It’s a massive map of (at this moment) 1,742 genres—umbrella genres like pop and country, as well as smaller niches like Thai hip-hop, German metal, and discofox—that McDonald has identified while examining the data that comes across his desk at Spotify. For music fans, it’s an endlessly fascinating look at how sound and aesthetics flow between regions and eras. For artists, it’s a useful source for understanding where they fit in the overall music universe, a context that can potentially be a valuable source of inspiration.
“It was originally a debugging tool,” says McDonald, who started working on Every Noise at Once when he was at the Echonest, a music intelligence company that Spotify acquired in 2014.
“We tried to use machine learning to evaluate subjective psychoacoustic attributes of songs—so how danceable, or energetic, or happy, or sad,” he says. “It was very hard to tell how such a thing was working—you think, ‘Does this song sound .7 happy?’ But I found that if I aggregated those scores over genres, I could make sense of it.”
Wow. Keep reading.