This is Why Spotify Pays So Little & How to Fix It
First, let me introduce myself.
I am an entrepreneur of 19 years. Have been a hardcore full-stack dev for 10. Have worked in the music business for 6. The amount of hours I have spent researching the music business and digital music services likely spans into thousands. From university studies to books to blog posts to news articles. You name it. But my start in the business happened more or less by chance.
In 2006, I started my own website. Nothing more than a quirky personal autobiography, photos of my cat, awkward short movies I made with my high school chaps, and a short playlist of all of my favorite music. I was 16 at the time. By 2007, this poorly designed, Comic Sans-ridden website was receiving hundreds of thousands of hits per day. My website ended up on the top 10 websites for music and new media on StumbleUpon. As it turned out, the music was more popular than my cat photography (though it probably didn’t hurt).
In 2008, I moved away to college. With my newfound technical expertise, I decided to host my own website (because, why not?). It wasn’t long before I was contacted by the university on behalf of the RIAA. By this time, media search engines all over the world had indexed my website. Even with a die down of actual web visitors, HTTP requests for music stayed consistently in the thousands per day. Of course, I wasn’t acutely aware of the difference between “visitors” and “requests” for some time.
“We notice an unusually high amount of traffic going to what appears to be a personal website.”
The RIAA had demanded I remove “Kryptonite” by 3 Doors Down, and in line with their request, the university threatened to disable my internet access and required me to complete a copyright education course. Not doing so could hit me with some steep fines or place me in jail, so of course I complied. Less than 100% on the final was considered failing. I took the final 5 times. I was so jarred by the experience, I removed my entire website from the Internet and didn’t do anything with music again for almost two years.
The New Era of On-Demand Streaming
In 2010, I started an on-demand music streaming service called BluePlaylist. It began as (like my personal website) a single playlist where anyone anywhere in the world could search through our database of ~1 million songs, and those that were played most often moved up into this global most popular playlist for all to enjoy. I envisioned it being a democratized source for music. One that wasn’t decided by media companies but by everyone in the world.
BluePlaylist quickly became what is now known as an “on-demand all-you-can-eat” music service, similar to what Spotify is today. The service acquired 121,644 registered accounts, totaling an estimated 200K+ users including unregistered users on its website and Android app.
In 2011, I wasn’t the only one in the market anymore. Rdio, launching just one month before BluePlaylist, had already secured agreements with the major rights holders. Spotify entered the U.S. market in a flurry over the summer, helped along by Facebook.
To paint a picture of what the landscape had looked like up to this point, piracy had been rampant. The younger generation couldn’t justify paying for music, at least on a per-song basis. The music industry, as a whole, was going down the drain. Profits were plummeting. For the longest time, the major record labels attempted to mitigate the problem by suing music listeners in a fight for their survival.
One of the first major players to make online streaming into a legal business was Pandora. Pandora operated (and still operates) under a special federally controlled statutory license (in the U.S. only) that allows them to play any song they wish legally (without even needing permission from the copyright holder) in exchange for a fixed per-stream or percentage-of-revenue royalty that would be collected and distributed by a single performance rights organization called SoundExchange. iHeartRadio, other internet radio providers, and even Spotify’s free tier use this same license.
Even though Pandora introduced many to online music streaming outside of piracy, it was not without limitations. Under the federal statute, they cannot play songs “on-demand” and still qualify for the government’s reduced royalty rate (today between $0.0017 and $0.0022 per stream). Attempting to do so would not only disqualify them from federal licensing, but it would also require them to get direct agreements with all copyright holders and pay (very likely) a much higher royalty, which they can’t stand to bear.
Circling back around, Spotify next entered the market, along with a slew of other on-demand services that came just before it like Rdio (recently sold to Pandora), MOG (sold to Beats then Apple), and of course my very own BluePlaylist. While Spotify rocketed into the market, there was an excitement about on-demand streaming but also concern about whether on-demand streaming could ever be 1) profitable for the service and 2) profitable for the artists.
The State of the Industry Today
Mo money, mo problems…
On-demand music streaming is now at a record high. While the industry overall has hit a low, on-demand streaming is expected to compensate for a loss in physical sales and downloads over the next several years, potentially raising the entire industry back up by billions.
But with new opportunity comes new issues. On-demand services are unable to sustain themselves. With Rdio selling to Pandora, MOG and Beats going to Apple, and Spotify losing more and more money ever year, music services themselves haven’t yet proven streaming music to be profitable. Not exclusive to on-demand services, Pandora too has struggled. Content acquisition costs (paying for the music) has been the biggest reason, with Spotify paying ~70% of its revenue to rights holders and Pandora ~55%.
On the flip side of this, a war between Spotify and artists has been intensifying. Income inequality between large and small artists is stark. The top 1% of artists and their record labels now earn 77% of all artist recorded music income, and artists and listeners are paying closer attention to how little of their monthly subscription actually goes to artists. Spotify officially claims to pay an average of $0.006 to $0.0084 per stream to rights holders, after payout is calculated based on a percentage of their total revenue. But how much can that differ between artists?
Why It’s Still Not Fixed
Justin Bieber is only partially to blame
There are 2 main reasons why the income inequality issue from on-demand music streaming continues to build. Hint: it has less to do with the major record companies than you might think.
The math for royalty distribution is wrong
At the moment, Spotify and every other major on-demand music streaming service pays according to a certain formula. Out of the service’s total number of plays, they figure out the percentage of plays that represented each artist. That artist gets that percentage of Spotify’s total revenue minus Spotify’s cut.
That means, if Calvin Harris receives 1 billion plays out of a total 7 billion plays on Spotify, 14% of your $9.99 subscription (after Spotify takes 30%) goes to Calvin Harris’ record label regardless of whether you actually played his music or not. These are not real numbers and Calvin Harris is actually very vocal about fair compensation for artists, but you get the point.
This royalty distribution model does not reward artists for the quality of their fan base but instead their quantity of streams. If one lone listener listens to Calvin Harris 1,001 times, Harris will make more money than if 100 fans listened to him 10 times each. If a niche artist has a very loyal following of listeners that generate 10,000 song plays but Drake gets 1.8 billion plays (actual), their royalty share is literally 0.0005% when only put up against Drake. Now imagine competing against Rihanna, Bieber, and every other artist on the service.
With subscription music surpassing 1 trillion streams and 79% of them going to the top 1% of artists, niche artists are at a grave disadvantage. The current model dilutes payouts to artists relative to the performance of every other artist. This does not reward artists for having a loyal following as long as Justin Bieber gets more song plays. This would be like 80% of my purchase of a Kia going to GM because they sold 5 times as many cars (not fibbing).
Justification for this model is based on some dangerous assumptions. It assumes, even though I have access to a GM vehicle, that I will drive it 80% of the time. The consequence for this entitled line of thought is the rich will become richer, and the economic mobility of smaller artists will diminish. This can help explain the sharpening decline of the longtail economically.
Content curation is skewed
No matter how royalties are distributed, income distribution among artists will remain skewed as long as the process in which listeners discover music favors the top 1% of artists. Royalty distribution is a result of song plays, but song plays are the result of curation.
Every major music service uses a strategy or variation of strategies to assist in music discovery and recommendation. Among these are hand-picked playlists, acoustic analysis, attribute tagging, textual analysis, and collaborative filtering. Each strategy lies on a spectrum at varying degrees of human analysis and machine analysis.
Human analysis involves great care and is used most often in hand-picked playlists. These songs are selected by “curators” whose jobs are to find songs that best fit together, like a radio DJ. But this method moves slowly at scale. Hand-selecting 30 million songs would be a near impossible task, even with crowdsourcing, and it would still not accommodate for individual listening preferences or changes in listening behavior within a short period of time. Machine analysis, in contrast, scales well but the quality of results can seem inauthentic or questionable, since computers don’t understand all musical attributes and the cultural context behind music.
Pandora uses a combination of human labor and machine labor using a strategy called “attribute tagging.” This method involves having several dozen music experts listen to each individual song, tag them with over 400 attributes (like instruments, harmony, melody, rhythm, etc.), and then match the songs listeners enter with songs of similar musical qualities. Because of the manual labor involved in tagging each song, Pandora’s total catalog amounts to only 1 million songs versus Spotify’s 30 million.
Spotify, the largest on-demand music service, uses a different approach called “collaborative filtering,” which scales better. In this approach, Spotify seeks other users who have built playlists featuring the songs and artists you stream most frequently, finds other songs on those playlists that appear most often between similar music listeners but you haven’t heard, and then filters the results based on your affinity to certain artists and the genres you are most likely to explore (source).
But there is a problem with this approach. It is susceptible to popularity bias. Meaning, popular songs are likely to become more popular and those with a lower rate of exposure are pushed further down the longtail. I witnessed this firsthand when using this same approach while building the “Suggested Songs” feature on BluePlaylist 5 years ago, and scientific research from Pompeu Fabra University and Barcelona Music and Audio Technologies (BMAT) — a music recommendation company similar to The Echo Nest — supports this claim:
“The results from the experiments reveal that — as expected by its inherent social component — the collaborative filtering approach is prone to popularity bias. This has some consequences on the discovery ratio as well as in the navigation through the Long Tail.” (source)
In addition to collaborative filtering, the medium in which Spotify recommends music — through a playlist — further exacerbates popularity bias.
“Because the playlist, that explicit act of curation, is both the source of the signal and the final output, the technique can achieve results far more interesting than run of the mill collaborative filtering.” (source)
This is known as a positive feedback loop, and while interesting, it actually worsens artist inequality on the discovery front by encouraging more plays for well-known songs. This can have a direct effect on income distribution, especially when amplified by the current inequitable royalty distribution model.
Oh these are good…
The Curation Problem
First, start at the source of the problem. Song plays. They’re skewed. To fix it, you have to fix curation. This is something I feel no one has done right yet. Spotify is on its way building more advanced recommendation algorithms. They bought The Echo Nest, one of the top players in the industry for music recommendation engines, and the guys there have been doing some impressive work. They’re the ones responsible for the “Discover Weekly” playlist and “Fresh Finds” playlist you will now find on Spotify.
But it’s not enough. The way in which music is presented needs to be changed. Adding a playlist or two designated for music “discovery” makes music discovery a feature, not the focus. Maybe this is just what Spotify is for. Wholesale music. Not music discovery. In that case, they should focus on what they do best. But what they do best may not be best for the music industry and for artists. Especially when it comes to leveling the playing field and giving all artists an equal opportunity.
The Royalty Distribution Problem
There is an alternative way to distribute royalties. It doesn’t have to be done the way it’s being done now. Those familiar with this issue call the current model the “pro-rata” model. Here is a new model: the “user-centric” model. Here’s how it works:
Instead of every paying listener throwing their money in a pool and divvying up the percentage of royalties that goes to each artist based on their amount of song plays, you take the songs each listener listens to, calculate the percentage of time they spend listening to each song, and then distribute their monthly subscription based on those songs they actually listen to.
So, when I listen to a song from an independent niche artist, 80% of my subscription isn’t going to Universal because 80% of everyone else listens to Fergie. If I were to only listen to my favorite niche artist 100% of the time, they would get 100% of my subscription, presuming it’s about $10. This really isn’t a complicated concept, and in fact, many people assume this is how it works already. But it doesn’t.
Changing to the user-centric royalty distribution model would reward those artists with a loyal following versus just a high play count. But most importantly, it removes the competitive factor in royalty distribution. Indie artists are no longer competing with Fergie, Rihanna, or Justin Timberlake. They are betting on themselves, and other artists can’t drive down the purchase power of a single listener. There is no inflationary effect. Support is again tied directly to the listener, as it would be if you purchased an artist’s music. And it is about the quality of music versus quantity.
Why no one is fixing it
The bit about major record labels destroying the music business is half true. Spotify has very complicated deals with the major record labels. For those who don’t know, there are 3 of them: Universal, Sony, and Warner and they control almost 2/3 of the music business. They don’t tend to compensate their own artists very well, with the average artist only receiving about 7% of streaming royalties. But their effect on artist inequality is indirect.
As part of Spotify’s deals with the labels, Spotify must pay hefty advances, meet growth requirements, give away advertising spots for free, give away portions of their company, yada yada. These most undoubtedly always work towards the labels’ favor. But I want to bring your attention to the advances and growth requirements.
In a nutshell, Spotify must grow at a rapid rate in order to pay off the labels and make any sliver of profit. The way the deals are structured, Spotify will continue to pay minimum royalty rates until a percentage-of-revenue model can kick in and *gosh-willing* they can actually make a profit. But to get to that point, they have to grow their paid subscriber numbers fast!
At no surprise to anyone, the way you appeal to a massive audience is to play the music they want to hear. It’s safe. It attracts the most people. It’s the top 40’s mainstream listeners. That’s how Spotify is going to grow as quickly as they need to. Spotify has no incentive to change business practices because 1) the current curation works in their favor to grow their subscriber base and 2) royalty distribution has absolutely no effect on Spotify because they take 30% off the top regardless. Who gets what portion of royalties is not their concern, and the major labels have no incentive to change a distribution model that is weighted in their favor.
‘Bout time this ended
There’s a shit ton of issues facing the music industry today. But for those of you who are artists, know that it is a combination of issues with both Spotify and (if you are with a label) possibly your label. Hardly anything is black and white, and this is no exception.
If you are an artist, you ultimately have to choose whether being signed on a label is right for you or not. I am not anti-label. I believe there are many independent labels out there that brighten the marketplace and are the reason why the industry is doing as well as it is. But as far as the big guys go, times have changed. Technology allows you to distribute, promote, and produce music easier now than ever. If you’re willing to put in 20% of the effort, you can get 80% of the result without signing your rights away.
As far as the issues with Spotify go, I don’t believe Spotify is in a position where they can make life altering decisions. They have become defined as the source for wholesale music. They will do the bare minimum needed to curate music in such a way that they’re still discernible from AM/FM radio, but we know where their incentives lie. And we know hell will freeze over before the major labels agree to a royalty distribution model that doesn’t work in their favor.
I think it might be worth mentioning what happened to BluePlaylist after all these years. In 2012, after I graduated from UC Santa Cruz with a bachelor’s in Legal Studies, I did my own analysis on the music business and concluded the all-you-can-eat on-demand music streaming model was unsustainable because of the high cost of royalties to record labels.
I will admit the numbers and data I had access to were limited, but the fact remained that the success of the business would be completely reliant on negotiated terms with mega companies hungry to get their money back for a decade of piracy. I would have been happy to pay royalties if 1) I thought the majority of it was actually going back to the artists and 2) if I could do so sustainably.
I have a very deep respect for artists and I want to see them succeed. They are the ones who make the music. Without them, there is no industry. Art and innovation are cut from the same fabric. But we cannot progress as a society by leaving the arts behind in lieu of money.
I want to fix that.