Blog1 min read
On music recommendations from friends
Three or four people, narrowly calibrated. The size of the network doesn't matter; the calibration does.
A friend's recommendation is the most valuable music referral I get. Better than any algorithm, better than the music press, better than the radio sessions a streaming service builds out of a track. The reason is straightforward: a friend who knows me chooses a track for me, not for an aggregate of people slightly like me. The conversion rate is real.
What's interesting is how much of the music on the playlists here came from a handful of specific friends. Three or four people whose taste overlaps with mine in narrow ways. None of them are music critics; one of them barely listens to music at all but happens to send me tracks from films. The size of the network doesn't matter; the calibration of the network does.
The flip side is the friend whose recommendations consistently miss. There's nothing wrong with them as a friend, or with the music; their taste is just calibrated for a different room. I've stopped feeling guilty about not following up on their tracks. The reciprocal favour is to send them recommendations that match their room, which most of mine don't.
For the rooms here, every track has a chain. Mostly the chain ends at a friend, or a friend of a friend, or an artist a friend introduced me to. The algorithm contributes a few tracks; the café at four in the afternoon contributes a few; the friends contribute the rest. The playlists are, in that sense, made by more people than just me.