I mean, the article is likely correct, but only if the current distribution of royalties in Spotify isn't proportional to the play count (which was presumed in the top comment). And it is totally believable that Spotify adjusts them to entice bigger names to the platform.
But if the distribution would have been equal, then there will ne zero difference with the user based distribution. Because it would be the exact same number counted differently.
User A listens to artists 1, 2, 3 in a 50%, 25%, 25% proportion on a 15$ sub (same below)
User B listens to artists 1, 2, 4
User C listens to artists 4, 5, 6
So by current system, artist 1 has 4 plays, 2 has 2 plays, 4 has 3 plays, 3 and 5 and 6 have 1 play. So 1 gets 15$, 2 gets 7.5$, 4 gets 11.25$, and 3,5,6 all get 3.75$
If we recalculate to the per user scheme, then from user A, 1 gets 7.5$, 2,3 get 3.75$. From B, 1 gets 7.5$, 2,4 get 3.75$. And from C, 4 gets 7.5$, 5, get 3.75$. Sum it up and we get 1 get 15$, 2 gets 7.5$, 4 gets 11.25$, 3,5,6 get 3.75$.
It's very basic arithmetic, from shuffling numbers and then summing them up differently, the total won't change.
So what needs to be changed is a skew artificially made towards big names, that is the real root cause.
You should really double check your "very basic arithmetic", because it's wrong. If a user listens to fewer total songs than the average Spotify user, then the artists they listen to will get a larger payout under a per-user model compared to an aggregated model, and the study I linked shows that this results in the top 10 artists earning ~12% less, which causes a larger distribution to smaller artists.
user-centric (pay for what people listen to) is explicitly not proportional to play count. that's why your arithmetic does not work.
to make it simple
-----
userA: 999 plays of artist A: pays 5$
userB: 1 play of artist B: pays 5$
pool-centric: A gets 9.99$, B gets 0.01$
user-centric: A gets 5$, B gets 5$
-----
bonus: in spotify system artists under a certain threshold get 0$. which gives guaranteed distortions even if you suggest ideal conditions for the pool/user thing up there to not matter (independence and large numbers).
extra bonus: spotify also dilute the pool by running mass music production sweat shops aaaaand by running more and more AI music for which they run the money back to themselves.
in practice small artists lose money on spotify, and a significant section of spotify artists would earn more selling a few cds a year on bandcamp than on their spotify ever.
But if the distribution would have been equal, then there will ne zero difference with the user based distribution. Because it would be the exact same number counted differently.
User A listens to artists 1, 2, 3 in a 50%, 25%, 25% proportion on a 15$ sub (same below)
User B listens to artists 1, 2, 4
User C listens to artists 4, 5, 6
So by current system, artist 1 has 4 plays, 2 has 2 plays, 4 has 3 plays, 3 and 5 and 6 have 1 play. So 1 gets 15$, 2 gets 7.5$, 4 gets 11.25$, and 3,5,6 all get 3.75$
If we recalculate to the per user scheme, then from user A, 1 gets 7.5$, 2,3 get 3.75$. From B, 1 gets 7.5$, 2,4 get 3.75$. And from C, 4 gets 7.5$, 5, get 3.75$. Sum it up and we get 1 get 15$, 2 gets 7.5$, 4 gets 11.25$, 3,5,6 get 3.75$.
It's very basic arithmetic, from shuffling numbers and then summing them up differently, the total won't change.
So what needs to be changed is a skew artificially made towards big names, that is the real root cause.