Anti-scale in music licensing: The Spotify example

More often than not, startups are in pursuit of venture dynamics that exhibit benevolent economies of scale: cost per unit, or per user fall as the number of users increases.

When licensing key inputs, rational startups pursue ceilings and not floors: license terms that place limits on the obligations rather than really high minimums for the obligation.

Music service startups face their own very special sort of “anti-scale.” Really high minimum obligations alongside ceilings that scale 1:1 with user growth or music use.

Eric Eldon of Techcrunch claims to have otherwise “secret” information on the license terms accepted by Spotify:

Its deal structure with the labels requires either a $200 million annual payment, like what it had to do last year, or around 75% of total revenue (whichever is higher)

That “whichever is higher” aspect of the licensing context says it all: Anti-scale.

First, by agreeing to such a significantly large upfront and guarantee, Spotify hopes to develop a barrier to entry– to offer a similar model any new service may have to match the scale of this payment.

Second, rather than license costs quieting with user growth, these costs get louder. In other word, once the risk of whether Spotify can acquire users dissipates a bit such that the guarantee can be paid, the total cost pool proceeds to expand linearly with that user base–a floor for the obligations rather than a ceiling.

In contrast, major radio stations will offer a large fixed pool of money for their royalty pool, enabling costs per user (listener) to decline once revenues exceed the size of this pool (which they do by many multiples).

Perhaps, someday, music service startups will not have to be completely irrational to operate. I guess it’s all for the fame, and they will just make money selling t-shirts and autobiographies.

The verdict is in… Media Pirates are undeniably and economically irrational, just like the rest of us.

A seemingly endless stream of statistics are in play these day suggesting (or proving, in the minds of some) that so-called pirates also buy music, movies, and other forms of digital media. What few if any of these authors discuss openly, however, is the fact that this strange behavior—buying some files and downloading others through unlicensed venues—is, for lack of a better phrase, undeniably and economically irrational.

Two of these recent findings would be:

Where do music collections come from?, from the American Assembly
Downloaden neemt niet af ondanks bestrijding, from the Dutch Institution for Information Law.

First, and to be clear, neither of these studies (and very few if any prior studies) are in fact “before and after” studies. Which is to say that these studies to not sample music consumption patterns before access to pirate sources and then after such access.  Therefore, we are still in lack of the conditions for a “this causes that” conclusion. We cannot say that pirating causes purchase behavior, whether that behavior is a greater or lesser number of purchases.

Second, some of these surveys ask users about their P2P behavior and then also ask about file sharing among friends and free downloads from online. In reality, many people may not be able to distinguish these sources, and some of these sources—such as file sharing among friends—might constitute P2P in small networks.

Finally, and ultimately, we can draw only one key conclusion from these sorts of studies:

People who download media from unlicensed sources are irrational—undeniably and economically irrational.

This observation, or assertion, should probably be the biggest curiosity for researchers and industry mavens. Instead, as the coverage of the two stories above goes to show, we are still stuck in the endless debate over who buys more, and whether piracy causes this or that.

Its time to move on.

Why do people purchase some media and acquire other media through “piratic” channels?

Frankly, understanding this question is to understand why and how the industry will survive going forward.



How the first subscription-based music service was built online, and why it’s time to move on.

Nearly fifteen years ago I began hashing the plans for what would become, as far as I can tell, the first subscription-based music service online. What we built was pretty straightforward and, as a result, the initial business model was nearly identical to that model later adopted by services such as Spotify. (Aside: True old-timers will remember that Patronet would also launch, but it offered music from only one artist, Todd Rundgren—who built Patronet).

Importantly, neither technology nor copyright law were ever really the roadblock to opportunity, even back in the Web 1.0 days. What and How people—meaning artists, lawyers, owners, consumers, developers, and even policy makers—thought about these music services were the speedbumps that would prove to be large enough to block the road, even still today.

In the end, I am pretty certain the following hold true:

  • The business model of subscription music services is and should be brain dead simple.
  • No amount of “That’s not how it’s been done,” or “Are you sure this won’t cannibalize iTunes?” or “But I make more from a CD sale” is going to change the fact that adapting to this model is both required and will require change.
  • The longer any adaptation to change stalls and the higher the price for that change, the fewer the number of ecstatic fans, the fewer the number of happy customers, and smaller the pile of money that will be on the table for the music artists and owners involved.

It is time, quite frankly, to move on… regardless of the road conditions.

In the beginning…

In 1998 we launched a simple subscription-based music service online, thanks to a number of artists who were willing to experiment—way back then—with something new. That paid site would later shift to a free site, Noisebox, through which we shared ad revenue with artists and copyright owners. The business model was designed to fit within BDS (Brain Dead Simple) constraints.

The pitch to consumers, artists, and copyright owners for this way-too-early-to-be-plausible music service was pretty simple:

(1) Consumers were charged a monthly fee to stream/download music from the site. The initial price was $2.99/$3.99.

(2) Artists, and any other party with a financial interest in the recording and/or the composition/lyrics, would split the subscriber revenue based upon each song’s proportion of all songs streamed each month. The initial pool for royalties was 70-80% of revenues. The initial catalog of music came from self-released artists, which limited the complexity in who/what was in the pool.

For context: Napster had yet to be released. was still for the most part a news site, with only a few music downloads. DimensionMusic, or dMusic, was more actively engaged than in hosting MP3 files for artists. IUMA was still hosting MPEG2 audio files. eMusic was still for CDs, and the digital download store of the day was called NordicDMS. Neither iTunes nor the iTunes download store would exist for nearly half a decade. Spotify would not exist for a decade.

Ironically, points (1) and (2) above basically describe the subscription music service model by which services such as Spotify operate today, nearly 15 years later. Why? Because that model was and is brain dead simple.

This model also happens to be the same one that is debated almost daily on music news and artist sites, as people try to grapple with the implications of not being paid $X for every album sold, or ¢Y for every iTunes download sold, or a guaranteed ¢Z for each stream on these sorts of services. Even though part (2) is essentially the same method through which radio royalties have been distributed (weighted by actual rather than imagined listener numbers) for decades–either as a portion of ad revenue or license monies.

Frankly, its time to move on.

And so I figured I would describe how simple that basic design truly was, and why neither technology nor copyright law were ever really the roadblock. Any roadblocks were in the minds of copyright owners, artists, developers, policy makers, and even consumers.

How does a music service work?

Technically, in the simplest of terms, you really only need three things to build a subscription-based music service:

(A) A number of music files, encoded in a streamable format (with mime type set, back then), and stored on a working storage device that is connected to the internet;

(B) A method for preventing access to those music files unless the user can present appropriate credentials (e.g., show they are a paying customer); and

(C) A means to accept some form of payment (e.g., credit cards) on a regular schedule (e.g., monthly) the result of which is the user acquiring the credentials mentioned in point 2 so they can access and stream the files mentioned in point 1.

Some copyright owners would add an additional lettered point that would require truly bullet-proof security on the files. But that point, in the end, is pointless.

Aside: Negotiating licenses is not a technical problem, its a time/effort problem. However, yes, I realize that acquiring these license can be a real challenge. But, in my opinion, the law was not the ultimate roadblock. It took us less than 30 days to license the music we were looking for in the initial experiment and the law never stopped us. Artists/owners who participated were willing to license and learn—and these were the two dimensions that mattered. The law simply gave these artists/owners the right to say, “No,” if they wanted to.

That was it, and that’s what we built through the absolutely simplest means possible via a traditional LAMP (Linux, Apache, MySQL, and Perl/PHP) setup.

All the music files were placed in a folder that was “protected” through .htaccess (those in the know will know why I used quotes). When a user paid for a subscription they were given a password. That password was added to the .htaccess list. If the user did not pay the next month, the password was removed from the list. Voila, Subscription music service in a box!

Modern services use a wider and more sophisticated range of technical means to accomplish their objectives, but the A, B, C’s listed above are ultimately “the basics” that inform any subscription service—whether in music, film, text, or otherwise.

Why it’s time to move on.

Since those early days, very little technical or legal change has intervened. Technology still provides the means to accomplish the A, B, C’s listed above—the formats and acronyms have simply changed. Copyright law still spells out the necessity of negotiating licenses with the appropriate parties. Most importantly, after years of music service licensing deals that are highly-punctuated with sophisticated protections (e.g., per stream minima) the underlying business model is returning to its origin:

(1) Consumers are charged a monthly fee to stream/store music from the site.

(2) Artists and/or those parties with a direct financial interest in the recording and the composition/lyrics, split the pool of subscriber revenue based upon each song’s proportion of all songs streamed/accessed each month.

Any substantive change that has occurred over the last decade and a half—meaning the sort of change that made this BDS business model more acceptable—has occurred in the minds of people, not the granite or silicon pillars of institutions or technology within which these minds make decisions or with which they take (virtual) action.

Change has been slow. Glacier, slow.

What I have found is that over the last 10-15 years I keep hearing the same questions, having the same conversations, and meeting with many of the same people (who simply work at different companies, have new haircuts, or wear different hats). From conferences panels to research chats to board meetings, the discussions usually leave me in a nearly perpetual state of deja vu. Bookmarked by entrenched positions at extremes. Punctuated by far too many egos (one of which is my own).

What was once thought of as “renting music,” is now thought of as “renting music.” Except, fewer consumers are freaked out about that idea. What was thought of as “the end of CD sales” is still in part “the end of CD sales,” and so what. Executives who used to think $19.99/month for music was a bargain now think $9.99 is a bargain (that’s a change), regardless of what consumer behavior suggests (that’s not a change). What was thought of as a challenging licensing market is still a challenging licensing market—except the incumbent firms now see that challenge as a barrier to competition.

In the end, and as I said at the beginning of this post, I am pretty certain the following hold true:

  • The business model of subscription music services is and should be brain dead simple.
  • No amount of “That’s not how it’s been done,” or “Are you sure this won’t cannibalize iTunes?” or “But I make more from a CD sale” is going to change the fact that adapting to this model is both required and will require change.
  • The longer any adaptation to change stalls and the higher the price of that change, the fewer the number of ecstatic fans, the fewer the number of happy customers, and smaller the pile of money that will be on the table for the music artists and owners involved.

Frankly, it is time to move on.

The ‘Lather, Rinse, Repeat’ model for understanding the copyright licensing paradox

Since the season of “he said she said” in copyright licensing is now in full bloom, I thought it wise to reveal a model I have developed for understanding the copyright licensing paradox.

To be clear, the paradox is not that the parties involved don’t understand this model.  The paradox leads to confusion over the right wrench to throw into the machine to make the model stop.

The model:

Note: The following steps can be performed in any order.  In particular, steps One and Two are often performed in reverse order. 

Step One

One party wants to obtain a license for as small a fee as plausible, under the most flexible of terms. The other party wants to grant a license for as large a fee as plausible, under the most stringent of terms.

Step Two

One party accuses the other of benefitting from copyright infringement and holding down the work of creators. The other party accuses the one of benefiting from innovation infringement impingement and holding up the work of technologists.

Step Three

One party claims the dollars on the table are too large, but pays. The other party claims the dollars on the table are too small, but accepts payment.

Lather, rinse, repeat.

The above models plays out ad infinitum for a few very simple reasons:

(A) Both parties have different interpretations of key concepts., such as “plausible,” “flexible,” and “stringent.”

(B) Both parties are correct. Most new technologies can and do “infringe” upon copyrights.  Most licensing deals will “infringe”  “impinge” upon innovation in some way.

(C) Speaks for itself… pure paradox.

Innovation at the edge: An investigation of music licensing efforts and the process of opportunity development

I have been finishing up a rather lengthy working paper, “Innovation at the edge: Making sense of opportunity at the boundary of technology and copyright.” I reckoned the raw findings from the work were better suited for a summary brief, however, in the form of a two-pager PDF and the post below. The academic working paper is, well, rather academic — making it not the easiest sort of thing for everyone (including myself) to digest.

Essentially, in an effort to understand the process of opportunity development — the academic interest — I investigated the music licensing efforts of interactive music services — the policy and practitioner interest. Many thanks to my university (Washington and Lee) for supporting this work, and to the many people who provided their time and insights through interviews yet shall remain anonymous.

Here goes:

The purpose of this research was to understand the process of innovation and opportunity development at the edge of two industries. The specific edge upon which the investigation focused was that at the intersection of copyright and technology, while the process of innovation and opportunity development under study manifested in the licensing discussions and negotiations undergone for new music services. Importantly, technology and copyright meet at a complex intersection, one that has confounded technologists, copyright holders, and scholars.

A combination of publicly available and privately obtained data for interactive music services launched or attempted to be launched in the United States were employed for analyses.  Published news stories and company statements were inputs for those data obtained through publicly available sources. For privately obtained data, semi-structured interviews were conducted with individuals who had been directly involved in music service licensing efforts.

In all, relevant data for greater than twenty music services were collected, resulting in a representative sample of firms whose lifecycles spanned more than a decade of licensing efforts and include: (a) services that launched and still operate, (b) services that launched yet have since closed down, and (c) services that failed to ever launch. The result of these efforts were a set of comparable case studies — comprised of core data types such as licensing timelines, process maps, and logics — that could be analyzed to gain insights into the process of innovation and opportunity production.

What I find is the following:

  • This licensing process unfolds as it does for reasons far more complex than one party “gets it” while another party does not — where it is the internet, innovation, copyright, or creativity. Instead, this process plays out at a nexus of frames of law, mind, and practice.

  • At the median, the directly-negotiated licensing activities of interactive music services have required roughly eighteen months of effort. About ⅚ of licensing time is spent completing deals with major record companies who are also major publishers, while the remaining ⅙ of time is spent completing deals with rights aggregators and collectives.

  • Between ten and fifteen sound recording deals, across major owners and aggregators of these rights, are believed to be necessary to offer upwards of 10-12 million recordings.

  • The time it takes to obtain direct and voluntary licenses from the set of sound recording copyright owners considered crucial for service launch has decreased only slightly over the last decade, a decrease on the order of approximately 3 months.

  • Blanket licenses and rates from the appropriate collectives covering the right of public performance have been obtained, in many cases, in less than 45 days if not less than a few weeks — substantial outliers do exist, however.  When services are not similarly situated to previously licensed services or service characteristics do not match those defined for statutory terms, these negotiations (or proceedings) have extended for years.

  • The amount of time it has taken to obtain a sufficient collection of licenses covering the use of musical works has decreased over the last decade — from years to less than 90 days in some cases — as long as that use fits within the categories outlined by the 2008 compulsory mechanical licensing agreement negotiated among the NMPA/RIAA/DiMA, and the service chooses to license via the notice of intent under section 115 process.

  • Estimates of the number of points of contact for licensing musical (work) copyrights vary substantially: From as few as 500, to as many as 6000, to a higher estimate of 30,000 points. This variety depends upon how these rights holders are aggregated and how licenses are pursued — via direct negotiation (low numbers) or NOI (high numbers).

  • As the size of the licensed catalog increases, each additional deal leads to a less reliable link between the effort required to license that additional catalog, user demand for that increase in available tracks, and the increase in legal certainty.

  • At any point in time, it appears that no greater than 2-3 law firms, or individual lawyers and their staff, are most central in brokering directly negotiated licensing transactions.

  • While new services face an expectation to be novel, variation in service characteristics among competing services after launch seems homologous. The service characteristics that arrive at the table to be licensed are, more often than not, at least somewhat different from those service characteristics that leave the table as licensed.

  • The pathway through which innovation unfolds is largely similar across the services studied. Whether there is a right way or a wrong way to travel through the licensing pipeline there is little variation in the way in which services make this trip.

We need a far better understanding of not only how opportunity unfolds at the intersection of two or more industries, but also how technology firms and copyright holders navigate the intersection at which these two parties meet, why the navigation plays out as it does, and how the intersection might be better designed so as to allow the value of this traffic of ideas-given-form to flow as effectively and fairly as possible.

Most of the conversation around the costs of licensing has centered upon the specifics of royalty rates and the dollar value of advances. Significant costs to these negotiations also exist, however, in the time, effort, and resources expended negotiating these deals, considering both real and opportunity costs for all parties involved — rights holders or service providers.

Without considering more broadly the frames — of law, of mind, and of practice — within which negotiated innovation takes place we lack a full understanding of whether and how any changes in copyright law might translate to changes in licensing processes or outcomes.


This work was funded by a Lenfest Grant from Washington and Lee University and by myself.  Over the years, however, I have worked on various projects for entities on both so-called “sides” of the licensing challenge — that means copyright holders/representatives and technology firms.

At no time while collecting data did I request or was I provided with access to legal contracts or private conversations/emails. I will only speak of personally collected data in aggregate and without attribution.  Any data discussed in this paper that are directly attributed to any service were accessible and obtained from public sources. The direct mention within this paper of any service does not imply that anyone affiliated with that service provided private data for this project.

Another déjà vu in digital music: Spotify offers discounts to university students, à la Napster 2003

Spotify announced today that university students in the UK who have purchased an NUS Extra Card, will be offered a 50% discount on the price of Spotify Premium — that’s £4.99 on a service for which post-college folks pay £9.99.

This sort of “price discrimination of demand” — to use a dismal term from Economics — has been applied before on music services, but in the US. Via this method, a product or service is priced based upon characteristics of the buyer not the product/service itself.  Here, students get a discount — the price is a function of the buyer.

Importantly, the wider context within which these discounts are being offered has changed significant.  And so, this time around, its likely a more positive outcome might result from this appeal to university students.

Back in 2003 and 2004, Napster began offering discounts of various kinds to University students in the US.  These discounts came in the form of (a) bundling Napster within university activity fees (e.g., Penn State), or (b) significantly discounting the streaming service down to $2 (e.g., USC, Middlebury, Vanderbilt, and others). Some of these methods were considered controversial at the time, with Napster requesting that schools not share partnership details and students, at times, complaining about the use of their fees.

Fast Forward eight years, and the dynamics in the marketplace have changed.  First, the service on offer at a discount is mobile this time around, not simply streaming only.  That means, students might actually be interested.

Second, yet related to the first dynamic, millions upon millions of mobile devices exist in the marketplace upon which this mobile service can reasonably and reliably operate.  Mobile services in 2004 could only run on a certain set of portable media devices, and the iPod was not in that set.

Third, I think attitudes have changed around paying for music on a subscription basis, such that the model “makes sense” to many people.  When I first attempted a subscription-based service, the first response from potential customers and partners: Who rents music?  Steve Jobs famously asked this same question.  Second response, “Is this a music club?”  The positive or negative opinion for licensing such a service was contingent upon how record club financials, as prescribed in record deals, were quite different for labels than for artists.

Finally, price.  I have presented research in a few venues the results of which suggest the market for music services could expand dramatically if prices were lowered.  And importantly, for every $1 in price drop the industry would gain additional customers such that the total pool of money on the table is larger overall — a 25% price drop could lead to greater than a 25% increase total dollars.

For example, Muve Music, offered by Cricket Wireless, has signed up between 600,000 and 700,000 subscribers to its service at an effective price of less than $9.99/month (the extra $10 for music includes a bump up to 3G speeds, a music service, as well as unlimited ringtones and ringbacks).  Cricket wireless has around 6 million total customers, so the service has signed up between 10% to 12% of its customer base.  If an equivalent base existed across all wireless customers in the US, the mobile music subscriber base would be between 25 million and 35 million accounts.

Only time will tell whether these sorts of discounts will expand the market, however.