In last week’s article, I explored how the market for recorded music grew with the CD, shrank with piracy and then grew again with the advent of streaming. If we understand the size of the $20 billion cake, now it’s time to look at how that cake is shared out.


Let’s start with the assumption that the money broadly follows the music we’re listening to (the details are tricky, as we’ll find out). The next question is crucial: how do we decide what music we listen to?

Algorithms: the new gatekeepers

Some of the music we stream is chosen when we deliberately seek out a specific track. But that appears to be a decreasing proportion. Much of the time, we listen to tracks in whose direction we have been firmly steered by our streaming platform. There are three ways in which this might happen:

  1. Playlists (general ones, not specific to us)

  2. Playlists or suggestions created for us by a personalised recommendation algorithm (“if you played X, you might like Y”)

  3. Search results from a query we made

There are many types of playlist: genre-related (Blues or Latin), mood-related (Romantic or Mood booster), music for activities from cooking to gym to gaming. Many are uncategorisable, some have big followings: on my UK edition of Spotify, Top Gaming Tracks boasts over 3,000,000 followers; Hot Country over 6,000,000; ¡Viva Latino! is the largest I’ve found at just shy of 11,000,000. Classical Essentials clocks in at a respectable 2,000,000.

Even on an occasion when you’ve chosen to play a specific track, you may be doing so because you first heard it on a playlist (20 years ago, this would have been on the radio). It’s not surprising that music marketers now pay an enormous amount of attention to getting their tracks onto popular playlists.

In the classical world, if you are looking for a particular recording of a particular work, you are the mercy of the platform’s search algorithm. As we’ve seen in my comparative review of platforms, it can be difficult to identify a specific recording if you don’t know what its album artwork looks like.

So who gets to decide?

Anyone can create a Spotify playlist and persuade their fans to “follow” it. But usage of these user-generated playlists is now in the minority: the key real estate in the Spotify user interface is the “Browse” function and that’s changed over the years. It used to be an eclectic mixture of various people’s lists, then it narrowed down to lists created by Spotify and the major labels. Today, my Spotify Browse shows only playlists curated by Spotify themselves.

The choice of what goes onto playlists is therefore a major factor in the success or failure of a recording. Whether the people creating Spotify’s playlists are working manually or with the help of algorithms, you can be certain that they have been tasked with keeping people listening and renewing their subscriptions, just as social media algorithms are tasked with keeping people on the platform above all else: the "attention economy" is king. The methods they use – as well as the algorithms for personalised recommendations and search – are closely guarded secrets and there’s evidence of them being commercially driven: last November, Spotify announced1 a scheme whereby labels can favour specified tracks in the algorithms chosen for personalised recommendations, in return for Spotify paying them a lower royalty: essentially a “pay to get played” scheme whose working is invisible to the end user. And since Spotify's algorithms and processes are secret, we can’t tell how many instances there are of playlists being padded with low-royalty or no-royalty tracks: one researcher has called for playlisters to be regulated by the Advertising Standards Authority.2

One high profile business, Hipgnosis, is buying up rights to hits of the past and actively working to ensure that they are streamed as much as possible in order to get the maximum return on their investment (some of the major record labels are following suit).3 The artists involved forgo future revenues in favour of upfront sums that can be very large indeed, which is great for them but is likely to come at the expense of younger musicians.

And what about classical?

All this should set off serious alarm bells for classical artists, because Spotify’s mood lists include virtually no classical tracks. So if you listen to Duvet Day, Dinner Music, the dubiously entitled The Most Beautiful Songs in the World, or any of dozens of workout or yoga playlists, Spotify have decided that classical isn’t for you. 

In any given country, Spotify’s composer and artist payouts are divided up in proportion to the total number of track plays across the whole platform. Therefore, every time someone puts on the Chill Vibes playlist featuring such artists as Lucy Park, Kevitch or Cigarettes After Sex, a classical artist’s payout is reduced. A dedicated classical fan might listen to a couple of symphonies in an evening, clocking up 8 track plays. That’s a fraction of the several hundred track plays that can be clocked up by a heavy user who leaves a background music playlist on shuffle for most of the day. The UK average listening, according to the BPI, is 2.5 hours per day4, somewhere around 60 track plays.

Even within the classical ambit, new artists and recordings can struggle, because they’re in competition with seventy years or more of back catalogue: my quick look at Spotify’s Easy Classical playlist showed that only a handful of the 68 tracks were recorded in the last ten years.

Dividing the cake: the user-centric versus platform-centric debate

Clearly, the “every track play pays the same” model (known in the jargon as “platform-centric”) has the potential to seriously damage classical music and any other minority genre or one not thought appropriate for background listening. An alternative payment model has been proposed and has been trialled by French platform Deezer: the “user-centric” model.

To understand how it works, let’s imagine a platform with just two subscribers. Alice and Bob pay €10 each; Alice listens to 10 classical tracks and Bob listens to 90 non-classical ones. In the platform-centric model, Alice and Bob’s €20 are divided equally among the 100 streams; each artist receives €0.20. In other words, Alice’s 10 artists receive a total of €2, only around a fifth of the subscription she paid: the overwhelming majority of her subscription has gone to Bob’s 90 artists.

But in the user-centric model, Alice’s €10 goes only to the artists Alice listened to: in our example, they get five times the payout. The model seems manifestly fairer to artists and composers from genres outside the mass mainstream and the user-centric model has a very large number of advocates, mainly amongst individual musicians and representatives of those smaller genres.

A study from France’s Centre National de la Musique (CNM) confirmed that the user-centric model “would limit the effect of income being channeled towards streams by heavy users” and “could favour a redistribution of income for the artists, titles and musical genres with the lowest audiences” but stopped short of making a clear recommendation in favour.5 The major record labels, who span most genres and therefore have little to gain, aren’t so keen on the idea: Warner Music argue that it would be “far more complex and administratively burdensome for digital services to implement” and that overall payments to artists might therefore be reduced.6  

Having participated in the CNM study, Spotify have declared themselves to be considering alternative models and that “the user-centric model is one that we are open to”.7

Sorry, is something missing?

To state the obvious, a composer or artist is only going to receive royalties if they have been correctly credited in the track’s metadata on the streaming platform. Sadly, that’s a lot less certain than one might have hoped for. Complaints abound of tracks being uploaded with missing artist or composer information.8 In some cases, that’s because labels, platforms and rights management organisations are suffering from antiquated systems. In others, it’s simple piracy: Bucks Music Group found that up to 30% of the Spotify uploads of one of their tracks was unauthorised.9

The problem has been going back a long time, and is severe. In 2018, the US Music Modernization Act obliged platforms to pay royalties from unmatched tracks (known in the trade as “Black Box Revenues”) to a specially set up fund, which is tasked with finding the rightsholders and making payouts. Last month, The Mechanical Collective announced that it had received a total of $424,384,787 in these payments.10 And that only covers tracks known to be missing metadata: one can only speculate as to the value of streamed tracks with metadata that’s wrong or incomplete.

The spectre of piracy has not yet been laid to rest. It’s easy to find apps to download a YouTube clip, getting rid of those annoying ads while depriving artists and composers of royalties. It’s also easy to upload a track to YouTube even if its rightsholders have forbidden this. The Music Publishers Association describes issuing takedown requests to YouTube as being like playing whack-a-mole: every time YouTube takes down a clip, someone else uploads it.11

But if platforms pay out 70% and artists get 10%, where is the rest going?

We’ve now talked about what music we’re listening to, how much we pay for it and how that’s shared between tracks. But that leaves a third giant piece of the jigsaw.

Royalties paid out by the platforms are shared out between five parties: performing artists, their record labels, composers or songwriters, their publishers, and the rights management organisations who represent them. The laws and the commercial landscape surrounding this process are incredibly complex and contentious.

We’ll be telling you about them next week.


1. Amplifying Artist Input in Your Personalized Recommendations, Spotify newsroom, 2 November 2020

2. DCMS Committee on Economics of music streaming: Written evidence submitted by Dr Hayleigh Bosher

3. Dorian Lynskey, ‘Record companies have me on a dartboard’: the man making millions buying classic hits, The Guardian, 27 February 2021

4. DCMS Committee on Economics of music streaming: Written evidence submitted by the BPI

5. Impact of online music streaming services adopting the UCPS, Centre National de la Musique, 27 January 2021

6. DCMS Committee on Economics of music streaming: Written evidence submitted by Warner Music UK

7. DCMS Committee on Economics of music streaming: Oral evidence, 23 February 2021

8. DCMS Committee on Economics of music streaming: Written evidence submitted by Helienne Lindvall

9. DCMS Committee on Economics of music streaming: Written evidence submitted by Bucks Music Group Ltd

10. The Mechanical Licensing Collective Receives $424 Million in Historical Unmatched Royalties from Digital Service Providers, 16 February 2021

11. DCMS Committee on Economics of music streaming: Oral evidence, 10 February 2021