The following two articles about Spotify are pretty interesting which can be found here and here. Part of the equation that another member had brought up in this forum a year or two ago (as well as the first article) is the user base. The greater amount of users = the greater amount of data that Ai can pull which generates recommendations. I believe there’s likely a variable of personal usage duration that help steer these recommendations even more - quality of recommendations may vary from a user who streams an hour/day vs a user who streams three hours/day.
Ai and its implementation doesn’t appear standardized, but lets say it was across all of the streaming platforms and we were strictly looking at user base and available data:
Spotify - 236 million users (2023) here
Apple Music - 93 million users (2023)
Tidal - 2.1 million (does not disclose/lawsuit claimed)
Qobuz - 200,000 users (2019, somewhat murky info)
Spotify and its massive user base would appear to have an upper hand on data and perhaps its recommendations to its users.
If you’re a Roon user:
Roon's general explanation of Valence can be found here.
Interesting discussions on Roon Radio within Roon Forums can be found here and Valence here.
* feel free to correct the numbers above. Used Ai to locate the majority of the stats :)