It's a vastly inferior method to Pandora's song analysis method IMO. I'd definitely use Spotify more if it had that, maybe significantly more - I can use it for 6 hours on a given day but then I often give it breaks because I'm on a discovery rut.
Maybe that's just me, and they know on very certain statistical terms that for most users that's not the case and music filter bubbles work much better.
Or maybe they're completely wrong in their approach and they're making a massive mistake at scale. We've come to develop this bias that tech giants "must know what they're doing" but I think it's often not the case. In ten years we could be reading how a competing service took over by offering the same catalogue with much better recommendation tech or other bells and whistles layered on top.
The story I linked also has a similar feel, the music is basically avant pop and neo soul, and instead of forwarding this cluster to someone with a musical/cultural background, a techbro comes up with their own label that comes off as basically "it reminds me of my favourite thing".
From the outside it seems they are really only approaching it from a purely technical point with a subconscious arrogance, ie they are adding the value not the musicians.
Their programmers probably make more money than most musicians on their platform and that's frustrating as fuck.
People who might in the past have gotten their new music from such deeply passionate music lovers and experts as John Peel, or any number of other DJs, are now often using algorithms in their place. And there is a sense that those algorithms are created by people with no real love for music or personal investment in it as part of our culture, which is kind of sad.
I think this is a big part of why music discovery through Spotify feels so soulless, especially compared to something like Bandcamp Weekly.
I guess their next step would be to start creating the music itself with algorithms. You would simply turn on Spotify and it would start playing a continuous stream of algorithmically created music to match your profiled tastes.
Spotify by comparison will also often recycle stuff I've already heard, but there's still always some new stuff in their weekly recommendations. Sometimes it's a hit, sometimes not, but at least it's something I haven't heard before.
I wouldn't be surprised if the whole thing collapsed under the weight of some stupid policy of the rightholders (eg, make sure it plays Taylor Swift every 5 songs, or else!) or some other commercial consideration.
They're merely "good enough" IMO. I can go to "Song Radio" or "Artist Radio", and browse through related artists and the like. But the latter still requires a significant amount of effort on my part, and the autogenerated radio playlists are nothing to write home about. I think it's fine that things like the "related artists" features exist, they emulate to some extent the old ways of discovering music, but they're also obvious and they would be part of a music service MVP nowadays. What I miss are solid automatic recommendations, more intelligent than those that are available.
As I remember, Pandora blew Spotify's automatic recommendations out of the water as far back as the mid-2000's. The idea of analyzing the song itself is so powerful and I think Spotify barely does it or weighs it way too little, favoring instead social aspects (such as the preference of other users). I discovered insane amounts of music with Pandora when I was able to use it, like dozens to hundreds of songs per week. Whereas with Spotify I go through these bad droughts that last weeks where I'm not discovering anything new that really clicks. I'd definitely pay for Pandora if it was available here in Spain.
It's weird to me that Spotify being the kingpin is so far behind Pandora in this regard. Is it that they don't care because users don't know any better or demand it? Or that what Pandora did was a technical miracle and can't be replicated at Spotify scale? Maybe it's some business model decision (eg, intelligent recommendations would stray users too far away from Taylor Swift)?
Maybe someone on HN can shed some light on this. And if anyone has any ideas for music services with intelligent recommendations available in Spain I'd be sure to give them a try.