Every Monday, Spotify hands you 30 songs you've never heard—and most of them hit. Here’s exactly how.
No surveys. No “rate this song.”
Spotify silently logs every save, skip, replay, and playlist add. That becomes your taste fingerprint.
But someone had to decide: which actions actually mean you liked something? A save is obvious. But what about listening to 80% of a song? That’s a product decision—not a technical one.
🧠 Implicit feedback — using behavior as labels instead of asking users directly.
Spotify looks across millions of users and asks: who else has a fingerprint like yours?
Maybe 500 people share your exact mix of genres. Those are your taste twins. No music analysis needed—just find people who love what you love.
🧠 Collaborative filtering — matching users by shared behavior. Spotify trained this on 700M+ user playlists. (Chartlex, 2026)
What have your taste twins discovered that you haven’t heard yet?
That gap is your playlist.
But the algorithm just finds songs. Humans decided everything around it—30 songs, two hours long, every Monday. Why Monday? “It cheers up Mondays.” — Matt Ogle, Spotify PM. (Music Ally, 2016)
🧠 Matrix factorization — maps users and songs into the same mathematical space to find your gap.