How Spotify (Almost) Always Recommends Songs You Love | A Thread
If you're a Spotify user, you'll know about the Discover Weekly playlist. Every Monday you get a playlist of 50 songs that Spotify thinks you'll like. But there are around 50 million songs on Spotify, so how does it know which ones you'll like?
Spotify makes use of two main factors to accurately suggest songs that fit your taste: "Likeability Metrics" and advanced Machine Learning algorithms.
Spotify analyses a variety of aspects of each song to determine how much you like it. The song's tempo, length, artists and featured artists, how long you listen to the song before you skip it (if you choose to), & many more elements all teach Spotify how much you like that song.
With that data, Spotify can find similar songs to recommend to you. But that's not the impressive part.
Spotify owes much of it's success to the underlying artificial intelligence and machine learning engines that it runs on.
Spotify has ~250 million active monthly users and among them, there are a total of 3 billion playlists. Spotify looks at the songs you have in your playlist and recommends songs you haven't heard that appear in similar playlists to yours.
That means if your playlist with 10 songs has 9 songs in common with another playlist, then the algorithm will recommend you that one song from the other playlist which doesn't appear in your playlist.
That's just one part of the multifaceted approach Spotify has to machine learning. With the extensive amounts of playlists and data that Spotify has to its disposal, it's only a matter of time before Spotify recommends songs you wish you knew earlier.
Which streaming service do you use?
Treasure | Multimedia Student
@treasure__n
Imperfect child of God | Unoriginal Thinker | Studying Multimedia @UPTuks 👨🎓