How Should Music Streaming Royalties Work?
A genuine question, I have no idea, and this post doesn't give an answer š
Like all good substacks, this started from āI was listening to a podcast and I heard something that annoyed me.ā
In this case, it was a (good, by and large!) episode of the (good, by and large!) podcast Decoder, where fellow podcaster Charlie Harding and host Nilay Patel were talking about how artists make money, and how the current streaming model is ābroken.ā
They talked about how the model works for Spotify and others, and how that model seems like a confounding mess, totally different and making way less sense than the more straightforwardly āfairā approaches that seemed like obvious alternatives.
My take isā¦.. I donāt think things are so simple. While I agree that the end state is something like āmany artists donāt seem to be making as much money as I feel like they shouldā, I donāt think there is any one model that is obviously conceptual āfairā, or āfairerā than others. I think there are a lot of options that all have trade-offs, and picking the best one is a matter of identifying what are the things we want to prioritize and then figuring out which model does the best at the things we care about (or does so while best balancing the trade-offs and externalities.)
Now spoiler, I donāt know what that option is! I donāt even know what to prioritize, or at least what the collective āWeā should prioritize!
Instead, this is more of a summary, a posing of questions, kind of a lit-review but before the research has been done. In this I want to lay out some of the models I see as options (including the one most dominantly used in streaming today), argue thereās no obvious prima facie ābestā, and then highlight the questions I think would need to be answered in order to be able to reason your way into an actual ābest.ā
Cool? Cool.
The ecosystem
First letās look at the overall (simplified) ecosystem, and a (also wildly simplified) version of the value chain.
In one direction, you have artists who make music. That music is funded and promoted and curated by labels. Labels work with distributors to get music into the hands of consumers, who listen to it.
And in reverse, consumers pay distributors to get music. Distributors take some cut, and pay the rest back to labels. And labels, depending on how the deals are structured, pay some money back to artists.
The reality is a trillion times more complicated than that, but thatās good enough for now.
For the streaming case, letās abstract even further and collapse artists and labels together into artists. This is obviously not true, and a huge part of the complaints around artists not getting their āfairā share is specifically related to how money is distributed between labels and artists. But I think thatās reasonably a separate discussion, and if weāre only talking about streaming this can help us focus just on the parts streaming services touch more directly.
To that end, letās say that streaming services are the only distributors, which again is obviously false, but also again it can let us focus.
So in that world, artists make music that is put on streaming services and is served to consumers. In reverse, consumers pay streaming services, who take a cut and pay royalties back to artists.
Also, for the sake of this, lets assume that the basic model between consumers and streaming services stays the same: consumers pay a flat subscription fee1 per month to streaming services, and get unlimited access to all the music in the catalog.
Now: how does that money actually flow?
Revenue Share
The primary model that basically every major streaming service uses is what Iāll call a ārevenue shareā model. So named because in this world, the amount that is paid to artists in the form of royalties is a percentage of the revenue that the streaming service takes in.
In the case of Spotify, this is allegedly about a 30/70 split, with Spotify taking about 30% of the incoming revenue from subscriptions or ads, and artists getting about 70%.
Within this model the next question is: how do you split that 70% among the artists on the platform. There are a lot of ways you could do it (maybe itās just split totally evenly! Everyone makes the exact same amount!) but thereās really one model in use, and one alternative thatās been proposed and lobbied.
Pro-rata model
This is the dominant model today, and the one used by basically every music streaming service, Spotify included. To demonstrate, letās ignore ads and just assume everyone is a subscriber and pays $10 a month.2
In the pro-rata model, every subscriberās $10 gets put into one big pool of money. Then every subscriberās listening behavior, or usage as Iāll call it, gets put into another big pool. Then the pool of money gets divided by the total number of streams in the pool of usage. That gives you a per-stream dollar amount, and that amount per-stream is the same across all users and all artists.3 Artists then get paid by multiplying the total number of streams they got by the per-stream dollar amount.
User-centric model
In contrast, activists and lobbyists and the streaming service Deezer4 argue for another model: the user-centric model. In this model, rather than one big pool of money and one big pool of usage that spans all users, each user gets their own individual pool. And then based on those pools, a per-stream dollar amount is calculated per-user. And the artist is paid based on multiplying each stream by its per-user-per-stream dollar value.
As an example of the two, letās say Alice and Briana are the only subscribers to Spotify, and both pay $10 a month. In a given month, Alice listens to 10 Blackpink songs, and Briana listens to 10 Blackpink songs and 10 Taylor Swift songs.
In the pro-rata model, Alice and Brianaās $10 each get put into a pool of $20 total, and their 10 and 20 streams get put into a pool of 30 total streams. Each stream is then worth $20 / 30 = $0.66. Blackpink got 20 streams, so they make 20 * $0.66 = $13.20. Taylor Swift got 10, so she makes 10 * $0.66 = $6.605
Now the user-centric model. In this case, we treat each user as a separate pool, so first Alice: she paid $10 and had 10 streams, so the per-user-per-stream rate for Alice is $1. Blackpink had 10 Alice streams so they make 10 * $1 = $10 from Alice.
Now Briana paid $10 and had 20 streams, so the per-user-per-stream rate for Briana is $0.50. Blackpink had 10 Briana streams, so they make 10 * $0.50 = $5 from Briana. And the same for Taylor Swift ($5).
So in the user-centric model, Blackpink now makes $15 total, while Taylor Swift makes $5. Remember, in the pro-rata model, it was $13.20 and $6.60, respectively.
On the podcast, Charlie and Nilay were saying that the user-centric model makes more sense and is the more obvious approach. But Iām not so sure about that!
At the very least, I think this shows that āfairā is a very messy and hard-to-define term, or at least a term with conflicting definitions.
For example, does āfairā mean that every stream is worth the same amount? If so, then the pro-rata model is clearly the fairer one, because in the user-centric model the value of a stream will vary wildly depending on the listening behavior of each user. A stream from Alice is now worth more than a stream from Briana.
Or does āfairā mean that as a user, I only give money to artists I personally listen to? If so, user-centric is fairer, because in the pro-rata model some of Aliceās money is subsidizing Brianaās Taylor Swift listening habit.
Or, as is a pretty common talking point, does āfairā mean a more equal distribution of money between artists, i.e. that we redistribute money from the most popular artists to ones that are less popular? If so, then itās unclear which model is the fair one, at least at a conceptual level, because that totally depends on looking at the actual data! For instance, in our toy example, switching from pro-rata to user-centric meant that the most popular artist (Blackpink) was making more money.
Weāll talk more about this later, but for now my point is: just looking at the models conceptually, I donāt think itās clear that thereās one āobviousā moral winner.
Other models
Revenue share (particularly with a pro-rata distribution model) is the primary way streaming services pay artists. But itās not the only way that could theoretically be used.
Flat per-stream royalty
Another one mentioned on the podcast is the idea of a flat per-stream royalty. In this case, rather than the royalty rate being determined as a percentage of the revenue, itās some fixed amount: making it up but say $0.01 per stream.
On the surface, it seems like maybe this is the āfairestā; itās an equitable, easy-to-understand flat cost per stream.
But if we go back to our Alice and Briana example, lets say Spotify says āGood news, weāre abandoning the pro-rata model and weāre switching to a flat per-stream royalty, and weāre setting it at the per-stream rate from last month.ā In that case, the new flat royalty is still $1 per stream. And that means many complaints about the pro-rata model also apply to flat per-stream royalties. If pro-rata advantages big artists, so will flat per-stream. If pro-rata means the listening of a low-usage user is subsidizing the listening of a high-usage user, then same here.
Margins
The other side to this, and to most non-revenue-share-based models, is that it shifts margin risk and margin upside from artists to the streaming services.
If we go back to our simple model of consumer pays streaming service pays artist, for an artist lets assume that their costs are fixed: it costs some amount of money to make a song, but there isnāt any marginal cost to each new stream.6 Artists' revenue comes from the streaming services. And therefore their margins are the revenue from streaming services - fixed costs, so their margin scales with their revenue.
For streaming services, their revenue comes from consumers, and their costs are whatever their royalty obligations are to artists. Their margins are the difference between them.
In the revenue share models, for streaming services their costs (royalties) scale with their revenue according to whatever the agreement is (again, for Spotify itās ~30/70) But their margins are fixed. Spotify will always make 30% margin, no matter how much money they bring in. They also will always make 30% margin no matter how much usage there is on the platform, because the overall royalty pool is only a function of revenue, not usage.
For artists then, itās a little more complicated, because the total revenue to all artists scales with revenue to the streaming service, but the per-stream revenue scales complicatedly-but-close-to-inversely with usage: if all of a sudden Alice and Briana listen to 10x as many songs, then the per-stream rate in both cases will go down7. Since we pretended artistsā costs are fixed and their margins scale with the revenue they bring in, that means margins scale proportionally with streaming service revenue, and scale (messily approximated as) inversely with streaming service usage.
In this flat rate model, the costs to the streaming service now no longer depend on revenue: they depend entirely on usage, since the streaming services now have to pay per-stream. So now for the streamers, if their costs grow with usage then instead of their margins being flat, their margins scale with revenue, and scale inversely with usage.
For the artists, since their revenue is the streamersā costs, their revenue scales with streaming service usage (and scales purely, since thereās no longer any dividing up a limited pool of money) meaning their margin scales with streaming service usage, and is uncorrelated to streaming service revenue.
To summarize:
Revenue share model:
streaming services have flat margins
artists margins scale with streaming service revenue, and inversely with streaming service usage
Flat rate model
streaming service margins scale with revenue and inversely with usage
artists margins are flat with respect to streaming service revenue, and scale with streaming service usage
Again, I donāt think itās obviously clear that one is fairer than the other!
In the revenue share model, streamers have no margin risk and no margin upside; itāll always be flat. Artists have margin risk related to usage increasing, but they also capture all the margin upside of streaming service revenue increasing.
If it seems unfair that streamers donāt have to take any margin risks, then we could switch to the flat rate version. But in this case streamers get all of the margin upside from any increase in revenues; if all of a sudden a ton of people sign up for Spotify but they listen less, then Spotify keeps all of that extra margin.
Maybe one is better than the other on deeper analysis, but same as before itās not obvious to me that one is clearly fairer. Itās a trade-off!
Other others
Those are the models mentioned in the podcast, and the main ones Iāve seen discussed elsewhere, but in theory there could be still others.
One I will call out that I think maybe is somewhat āobviouslyā less fair than those three is the Tiktok-style model of a āCreator fund.ā Hank Green explains it quite eloquently, but the basic idea is similar to the pro-rata revenue share model, but instead of the amount of money in the fund scaling directly with revenue, itās a fixed amount. So this would be if Spotify said āweāre gonna put a billion dollars in the pool, you guys will split it up by usage, and weāll take all the revenue upside.ā To me, for artists that has all of the downsides around usage-based margin pressure of the revenue share models, but with none of the revenue-based margin upsides, so thatās a pretty raw deal.
There also are things one could apply to all of these models that may tweak things.
You could imagine an approach that doesnāt split things per-stream, but instead per-minute-listened. Most of the same dynamics would play out, but maybe instead of the alleged current dynamic where streaming pressures songs to be shorter so you can get more streams in the same amount of time, maybe it would pressure songs to be longer so you can get more minutes from a given song.
You could imagine applying tiers, where perhaps royalty rates scale with the number of streams an artist gets. Maybe this involves scaling proportionally, as a way to reward artists that users most enjoy and bring the most value to platforms. Or maybe scaling inversely, to redistribute from the most popular and successful artists down to the less popular.
You could also imagine some sort of minimum floor, where if you get some number of streams to be considered a ārealā artists, you have some guaranteed payout thatās enough to support some minimum standard of living, and then any streams above that are a bonus.8
Or, I mean I joked earlier, but likeā¦ you could just say āEveryone with at least one stream gets paid equally.ā That is certainly āfairā in one sense, although obviously not in others.
How to make streaming āfairā?
This whistlestop tour of models is certainly not exhaustive, and thereās probably lots of clever ideas out there that I donāt know about. And this is still limited to the world where we assume all-you-can-stream music services funded by monthly subscriptions is the model to use, which maybe itās not!
But the point is that I think most of those options even in the brief tour have legitimate pros and cons for artists, and itās not immediately obvious that one model is the āfairā one.
So what could be done? I think the answer is to 1) clarify the goals, and then 2) get more data
Clarify the goals
At the simplest level, what do we mean by āfairā???? What does a āfairerā world for artists look like? Some specific questions that I think would be helpful are:
Do we want more money for all artists?
If so, do we want money to come from a redistribution from streaming services to artists, or from consumers to artists (aka consumers should pay more?) Or both?9
Do we want redistribution of money from some artists to others? If so how do we want that redistribution to go?
Is the important thing redistribution/more money in and of itself, or is the real goal more about giving artists a minimum standard of living?
What are the criteria for distinguishing an artist from a hobbyist? Should we distinguish between them at all?
Is it fair for the money from some users to subsidize the listening habits of others?
Who (artists vs streaming platforms) should bear risks to revenue and margin? Who should get upsides?
And should these risks and upsides vary based on how much consumers pay (revenue) vs how much consumers listen (usage)?
There are probably many more, but these at least are the ones that first come to mind for me.
Get more data
If we have a sense of what the goal is and what the end state of a āfairerā streaming world looks like, then I would want to know more about how close each of these options get us.
Some data is already available, relating to things like how much do artists currently make per-stream on average10, and how does that translate into revenue for certain artists willing to share the data.
I would also want to know some consumer level data such as what are the ranges consumers would be willing to pay monthly for a streaming service, and project out from that what a theoretical maximum total pool of streaming money could look like. Iād also be interested, though maybe more out of theoretical curiosity than real practicality, in what both consumers and artists would think a āfairā per-stream dollar amount looks like.
The big dream data set would be to get a representative sample of actual consumer usage data. As with above in the Alice and Briana examples, how people actually listen across users and across artists has a huge impact on what money distribution looks like under various models. If hypothetically the goal is redistribution from popular to less popular artists, we canāt know whether changing from pro-rata to user-centric, or applying a progressive tiered royalty structure would have the impacts we want without looking at actual representative data.
Unfortunately this is (naturally) private data from streaming companies and record labels, but the more that we can get streamers to share data with researchers or push them to do their own research and back up claims with numbers, the better. If Spotify is going to say that the pro-rata model is best for artists, then why not prove it!
If I went through this exercise myself, Iād say my own personal priorities would be:
Consumers should probably pay more, but I donāt necessarily think streamers should (aka, about 30% margins seems a reasonable ballpark for a fair steady state.)
I care more about artists having some minimal standard of living than some minimal standard of money coming in from their arts.
I wouldnāt hate some level of redistribution from popular artists to less popular ones, but I feel more inclined to have that redistribution happen via taxes and social programs rather than overlaid on the streaming market (and again, I think I care more about ācan artists surviveā than ācan artists make money directly from their art.ā)
I think its totally fine for the money of a low-usage user to subsidize the listening of a high-usage one.
I think I lean towards streaming platforms getting both margin risk and margin upside, but this is a loosely-held leaning.
Based on that and my guesses at data, I would probably imagine my ideal solution being a combo of a flat per-stream royalty thatās meaningfully higher than the current average, and that being covered by raising prices on consumers (and then also implementing a stronger social safety net paid for by taxes.) But that totally may be the wrong idea! Maybe consumers wont pay more than $10! Maybe real listening is so concentrated in a tiny number of superstar artists that thereās no flat royalty rate that could ever raise meaningful money for a non-superstar!
And of course, those are my first stabs at priorities; I wouldnāt be surprised if those vary greatly between people invested in this. Which I suppose is the real goal of this whole piece: a simple statement like āstreaming isnāt fair to artistsā is so much more complicated, and to me itās worth taking those complications seriously.
Iām gonna ignore ads for now, but if you imagine ads as the advertising companies paying the āsubscription feeā instead of the consumers, then it behaves similarly to the subscription model.
Not true both because ads exist, and because Spotify has different price tiers and runs sales etc etc. But for simplicity, lets say its $10 for everyone.
Also not true in reality, in the case of Spotify for instance they do have separate pools of money and usage per country, and I believe Iāve heard that Spotify does negotiate slightly different deals with slightly different labels, so itās not exactly the same per-stream for every artist and every user. But this example is the pro-rata model in its purest form.
Also, not a music streaming service, but this is how Youtube Premium works! Since Premium has no ads, Youtubers get no ad revenue from watches from Premium members. Instead, they get a cut of the subscription revenue, and it just so happens Youtube chose the user-centric model (though I donāt think it was called that at the time) over the pro-rata one.
This was a bad math example because the actual split would be $0.66666666666 repeating, which means the total math isnāt adding up, but whatever, close enough.
You know the drill, not true, but close enough blah blah blah
How exactly this would scale would depend very deeply on the particulars of actual listening behavior, as well as whether itās the pro-rata or user-centric model.
You could also imagine this function being served byā¦ā¦ a reasonable welfare state funded by taxes š¤
Again, this leaves out labels, but thereās the separate argument that money should be redistributed from labels to artists.
Fluctuates, best I can tell from googling is between 0.5 and 1 cents per stream, at least in the US.