A better way to cut a cake and eat it too

On revenue share in music streaming and the value of platform economics


In the wake of the Apple-Beats deal, some of the most interesting comments high-light limitations in the current business model of music streaming, and point to future possibilities as a result of this acquisition. This post discusses how the business model of music streaming can be strengthened by a shift from a ‘pro rata’ to a ‘user-centric distribution’ of income—which the Clouds & Concerts research project at the University of Oslo presented at SXSW in March—and by building on this to develop the platform economics of streaming.

The current streaming model is broken, in the sense that there is no direct link between artist and fans in an economic sense. This is due to the way subscription fees are divided in the current pro rata model model, where income is not decided only by how the fan group streams a release — but what other users are listening to. In other words: No matter how well you serve your dedicated followers, and how much they love your work, your success is in the hands of events and users outside of your control.

The pro rata model’s missing link

The settlement each artist receives is based on a model of distributing the income from ads and subscriptions which is called a pro rata model — which for was chosen as the standard when streaming arrived, over of several other possible models, probably to maximize profits from a few top stars.

In a world of physical albums or tracks, a share (perhaps 10-12%) of each individual purchase goes directly to the rights-holder of the record the listener has bought; hence there is a direct link between the user and the artist in terms of where the money goes, even though the money goes via a a distributor or a label. With streaming, revenue is generated from a subscription fee of typically $10/month (I’ll keep ads out of the discussion here, to keep it simple) and divided according to a pro rata model (or prorated), meaning a proportionate allocation, assigning an amount to a fraction, according to its share of the whole. Put differently: All subscription fees are divided by the collected streams from all users.

To illustrate the fundamental shift the pro rata model introduced from the physical sales model, let me exemplify with two users: Miriam and Arnt.

Let’s say that Miriam streams 1000 songs a month, while Arnt listens to 100. Both contribute the same amount in subscription fees per month. In the pro rata model the collected streams of these two users are summed up (here 1100 streams). When the settlement for labels is calculated (which is the basis for what the labels pay out to artists), Miriam’s 1000 streams is divided by 1100 and Arnt’s 100 streams is divided by 1100.

In other words Miriam’s listening has a 90,9% impact over who gets paid, while Arnt’s listening merely has a 9,1% impact. And out of their collected $20 in subscription, $18,2 goes to the artists Miriam listens to, while only $1,8 goes to the artists Arnt listens to. Listeners who are not streaming a lot hence subsidize the habits and artists of heavy streaming users.

If Arnt had only listened to one stream this month, the difference between Miriam’s and Arnt’s impact on the revenue streams would have been even more skewed: a 99,9% to 0,1% impact. The same logic applies for millions of users: the impact of each listener or fan group has on what the the rights holder (and subsequently the artists) will get, vary according to the quantity of their monthly streams, and not what each contribute to the streaming economy via their subscription fees.

One consequence of this model, is that an artist or label cannot predict the income generated from a core fan group, and count on income from this group when they release their album on a streaming service, because income is not decided only by how the fan group streams the release — but what other users are listening to. In other words: No matter how well you serve your dedicated followers, and how much they love your work, your success is in the hands of events and users outside of your control, because there is no direct link between the streaming of one fan and one artists anymore in terms of where the fan’s subscription fees goes.

A user-centered distribution of streaming income

An alternative way of distributing the same collected income would be a ‘user-centric’ distribution. In such a model, each user has the same impact if they pay the same subscription, regardless of how many streams they stream per month or what and how much other users are streaming. Here the money from each user goes directly to the artist each user has listens to a given month (via their label), and hence fixes the direct link between artists and fans that is broken in the pro rata model.

In other words: the artist which banks on their core fans listening to a new release, will not be dependent on what and how much other fans are streaming — only the fans they are trying to reach. In this model, if you are a die hard fan of band X and only listen to this artist during a month, all your subscription fees goes to this artist — whether you stream a hundred or ten thousand streams by this artist. It also follows, from this model, that each stream will vary in value, according to what each individual listener streams. In the case of Miriam, each stream she plays would be worth $0.01 ($10/1000) while Arnt’s would be ten times more valuable $0.1 ($10/100). As before, both users would be worth the same to the streaming sevice ($10 per month) but their impact on where the money goes would be equal, in contrast to the current pro rata model, where Miriam would have 10x more impact on what artists receive in settlement as Arnt.

Now, this model is in itself not new, and labels, artists and others have speculated about how things would look if this model were implemented. Still, no one has until now calculated how it would actually work in practice and in detail, based on the concrete listening patterns for a whole population of listeners in a streaming service. The Clouds & Concerts research project recently did this based on all Norwegian users of WiMP Music streaming service (which roughly 21% of the Norwegian internet population has access to) and presented the results (pdf) at SXSW in March. Just to recap three of our main findings here:

  1. On a macro level, little changes. The three major labels still get the lion’s share, with 75% of revenues; down 1 %point from 76% in the pro rata model.
  2. 75% or the top streamed 5000 artists (who account for roughly 90% of all income) gain or loose ± 20% with this new model (the two models have exactly the same income to distribute, of course).
  3. The winners are local artists (gaining +13% on average) and artists with a high passion index (a term coined by Paul Lamere i.e., the total artist plays divided by the total number of listeners each artist has) — which is a useful metric to identify artists with especially dedicated fans.

The last finding is the most interesting here, and ties the ‘user-centric’ case to the mentioned platform idea that has been used in discussing the Apple-Beats deal. Let’s recount at a few of these arguments:

Lady Gaga and streaming as an economic platform

Michael Vakulenko frames the Apple-Beats deal in the light of ‘Lady Gaga as a business model’, an in the terms of platform economics, where “two distinct user groups that provide each other with network benefits” (Wikipedia). In this perspective, Gaga and her team has built “a platform on top of music—because music […] sells everything but music.” (Troy Carter quoted by Vakulenko).

“The business people behind Gaga […] saw the missing link in the digital music value chain. This missing link is also a huge opportunity to fill the void going much further that “selling the artist” to the loyal fan base.” (Vakulenko, Medium, May 29, 2014)

Horace Dediu makes a similar case, and sees the aquisition as a possibility for Apple to:

“reinnvigorate music as a business around the artists as curators, [and] as the force of the industry that they traditionally held, before the labels became the overwhelming power-brokers through distribution” (Dediu, The Critical Path podcast, transcribed from post cast May 28, at ca. 1 hours 09 minutes).

Both Vakulenko and Dediu point to the skills and clout of Jimmy Iovine and Dr. Dre as key to this interpretation, as does John Maples:

None of the streaming-music companies have a leader who built one of the most dominant labels in the past 20 years, like Iovine did. Nor do they have someone who invented a new music genre, developed the talent, and then blew it up for a commercial audience, like Dre did (Maples, <re/code>, June 5 2014)

Maples also brings Ian Rogers, Beats Music CEO, into the mix:

Rogers understands technology and how it can be used to help artists market to fans. From his days building the Beastie Boys’ website to his time as CEO of Topspin, he has been on the leading edge of new ways to communicate directly with fans. And there has never been a better potential platform for artists and fans to connect than streaming services. (Maples, <re/code>, June 5 2014)

The basis for the latter claim is, perhaps not, obvious to all. But put short: until streaming, there has never existed a method of studying in detail how a large population of music lovers actually use music in their everyday life. Earlier, researchers or labels at best knew how many items of an LP or CD was sold in each market, or gained insight from surveys with a sample of users, or from observations and interviews with groups of users. With music streaming, all services log each individual stream down to a millisecond, every search queery within a service, how playlists are made and shared, which device each user plays the music on, how listening sessions vary with day, week or season, the zip code of fans of any particular artist, and a host of other details by each user and user session, which can be used to analyze relationships and correlations between different fan groups, or develop algorithms for discovering new artists, predict outcome to concerts at a particular venues and whatnot.

This is a gold mine of information about the way music is used for streaming services, labels, — and for a research project like Clouds & Concerts, who have been fortunate enough to be given access to anonymized user data from WiMP Music since they launched in 2010. Unfortunately, fairly little of this wealth of data has until recently trickled down to the millions of artists that make out the sought after content that fans access through the different streaming services. Most artists only get a vague sense of their fans’ listening patterns via the settlement check they receive via their label each month — be it large or (more often) small.

Building upon a stable platform

Commentator discussing the Apple-Beats deal from the vantage point of the US or UK, risk assuming that there is little or no money in streaming for artists, and that streaming is primarily about the platform (cf. the quote from Troy Carter above). However, in advanced markets such as Norway and Sweden, where streaming now is the main medium for music listening, revenue has come back and are now at a pre-Napster levels, looking to have a full recovery somewhere around 2020 — according to WiMP Music’s projections (cf. figure below).

Music sales in Norway 1977 to 2013, with projection for 2014-2020. Source: WiMP Music.

Still, even in successfull markets like the Nordics, the streaming model raises two main challenges, as I see it:

1) The liquidity challenge: In streaming, most newly produced albums can only break even after a fairly long period of streaming, compared to with puchase of a physical album or iTunes downloads, where the consumer makes a purchase up front for future music listening. Hence, small labels and artists struggle with generating enough short term income to make the next album or go on tour.

2) The predicability challenge: With the pro-rata model, it is difficult to predict the streaming income from a new release, because the income is influenced not only by how well it is received by a loyal listening group, but also by other events, other releases and other fans. If you release an album the same month as Beatles reunites, Elvis returns, and Justin Bieber releases a double album with Rihanna, you’re pretty much screwed.

The user-centric model presented above minimizes both these challenges, albeit not removing them. As an artists or label, you will still need time to get return on your investment, and even your fans may find the second coming of Elvis more worthwhile than listening to you. Nevertheless, given that the money in a user-centric model follows from users to artist, the chances are much better for artists to build a long term relationship with fans.

Instead of the streaming model being a shifting and uncertain platform to build fan-relationship, and to create value on top of the model, the user-centric model can create a firm foundation and level playground for artists to compete for users’ attention on.

With the gold-mine of user behavior that streaming services can provide, individual artists can (with some help) develop services, up-sale and extra income from their core fans, that fits both the needs of the individual artist and fan-group. Many excellent cases exists for creating value in a platform logic — especially outside of the music business. Think of Facebook, where developers and service providers are given insight into the user patterns on their apps via the developer platform Parse, or Apple’s app store, where developers market and sell their product and services to hundred of millions iOS users. Or think of oDesk, where you can find and hire freelancers for any job. Here costumers get insight into rich data about freelancers, and the other way around: freelancers get insight into user patterns and user groups, enabling them to offer better services; a true ‘two-sided market’.

Using the data on streaming behavior better, and improving transparency for all parties with interests in streaming (artists, labels, developers, the streaming services themselves, users etc), many new opportunities for value creation on top of the revenue generated from streaming itself can be envisioned, such as:

  • Fan donations: When each users gets an invoice with a pie chart or list of which artists they have streamed last month and their share in percent or $, fans can choose to donate ‘top up’ the artists they think deserve more, through a donate button in their electronic invoice or in the streaming UI (like freeware, donation-based or pay-what-you-want businesses models like Wikipedia, Twitch, Radiohead’s In Rainbows etc.).
  • Crowdfunding of recordings or tours, similar to what is already happening on Kickstarter, but with the benefit of being accessible from within the streaming service itself. As on Kickstarter, artists could enable differentiated commitment from fans, such as contributing $20 and receiving a band t-shirt, or $2000 and get your name in the album liner notes.
Screen shot of official merch offered in Spotify when listening to Led Zeppelin.
  • Upselling merchandise or premium content, from within the streaming service, such as artists like Led Zeppelin is already doing on Spotify (cf. screen shot).
  • Booking concerts, festivals or special fan events from within the service, and getting recommendations for concerts and events based on correlation with similar artists on tour (with the possibilities of service providers to take a cut from ticket vendors, tour managers etc).
  • Insight into artist-fan-relation can also be a powerful basis for creating native advertising providing individual fans content that they may find useful and beneficial in a concrete context. Different artists can device different offerings that make sense to their particular fans—or opt out of advertising to their fans entirely, if it does not serve their image.

Opening up for open innovation

Given the nature of open innovation, many other ways of creating extra value on top of the streaming platform will appear, that perhaps no one has the fantacy to envision today. The basis for this to happen, however, is that streaming services share more of their (non-sensitive) data on usage with artists, developers and fans via informative dash boards and APIs , and open up for apps and add ons to their platform. Much in the same way that Facebook works for developers today, where a Facebook provide analytics about their users to help them continue grow or monetize better.

All this is possible, of course, within the current pro rata distribution model. But my argument is that a user-centric distribution model will create a more direct and firm link between artists and fans, create better incentives for artists (and labels) to capitalize on this relation, and increase predictability for coming streaming revenues for artists and labels. In this way, I believe a user-centric model for settlement provides the most sound platform for the future business model in music streaming.

For this to happen, however, all labels within a given market, must agree to collectively change over from the ‘left’ to the ‘right’ side of the road, and a new model for calculating settlement to labels and artists on the same date. In Norway, WiMP Music, and several labels have already stated that they are willing to do this, using this service as a sandbox to see how the model works out in practice over time. Perhaps Apple Beats them to it?


For more information about the research project, and for further results, please visit www.cloudsandconcerts.com at the University of Oslo.

A big thanks to Michael Vakulenko and Tom Christian Gotschalksen for providing thoughtful comments and examples to the initial draft. Gotschalksen was also instrumental in helping Clouds & Concerts establish the initial contact with WiMP Music in 2010, which made it possible to study Norwegian streaming data.


This article also appears as a comment in Record of the Day magazine 26.6.14.