Could half of music royalties pay for music we would rather skip over than listen to?

In this post I am going to highlight what may be one of the great ironies of how we pay for music in the “music as a service” era:

Upwards of half the money paid out in royalties by music services may be used to pay for music we would rather skip over than for that music we would rather listen to.*

What we’ll discuss below is if nearly 15 songs are skipped during any hour (according to data released by Paul Lamere from Echonest), then upwards of half the royalties paid by music services could be accounted for by these skipped tracks.

If, however, skips only count as plays—from a royalty perspective—after a certain number of seconds of playback (e.g., after 30 seconds, as suggested by Mr. Jonze at the Guardian and Stuart Dredge, at MusicAlly), then the proportion of royalties paid due to skipped tracks falls to between 20.86%, or just over one-fifth of royalties paid, and 23.5%, or nearly a quarter of total royalties paid.

Regardless, the matter of counting skipped tracks as royalty-bearing performances needs to be taken into account amidst the broader discussion of payouts from these music services. A reasonable person would conclude that skipped tracks are evidence of music transactions that might not otherwise have occurred in a world of Track and Album sales—few people purchase a track just to skip over it!

As a result, browsing is now a crucial piece of the business model for music; As a revenue stream for music stakeholders.

How did I come to this ironic, if not ultimately ridiculous conclusion? Here goes…

* With apologies to the grammar police for ending a sentence with a proposition.

A Song Skipped is a royalty paid

We all skip tracks while listening to music—whether on CD, via Pandora, or via our favorite music service like Spotify, Beats, Rdio, Google Play, Deezer, etc.

What most of listeners don’t realize, however, is that royalties are paid for music that we skip. In fact, in some cases, a full royalty is paid for music that we skip even if we skip a track in less than a few seconds. Whether you think this approach to royalties is fair or not, in order to get the license, services pay for music even when a track is skipped.

For example, webcasters such as Pandora owe a full performance royalty for any portion of a track streamed to a listener, if they choose to operate under the variety of licenses offered via the statutory process in the US. As a result, a track skipped after 1 second is worth as much in royalties as a track listened to for its full duration

In other words, webcasters — and in turn listeners — pay as much for a skipped track as they do for a track heard in its entirety. This little factoid is why the nuances of no royalties owed for skipped and promotional “Heat Seeker” tracks played into the economics of the supposed agreement between Apple/iTunes and record labels.

While the law—or, more clearly, the cost of abiding by the law—leads some webcasters such as Pandora to limit the number of skips in any hour and other webcasters to not permit skipping at all, that law does not limit interactive, or on-demand music streams from services such a Spotify, Beats, Play, Rdio, etc.  who license directly.

Roughly speaking, a service such as Spotify pays royalties based upon any song’s share of total listens over any period. And so, if song X is 10% of all listens, the stakeholders in song X (i.e., record label(s), performing artist(s), musicians, publisher(s), songwriter(s)) will receive—roughly speaking—10% of the royalty pool. Other services, such as Rhapsody, might pay for music on a per performance basis, or at least face a minimum per track payment alongside a percentage of performances rubric for royalty payouts.

The above-described difference in payouts methods is why artists might have seen metronome-like consistency in the dollar value of payouts from services such as Rhapsody, while seeing variation in payments from a service like Spotify. Slight changes to listening hours leads to different payouts (on a per stream basis) from Spotify, while similar changes in listening hours lead to different total payouts but the same per stream payout from Rhapsody.

Nearly 15 Songs are Skipped Per Hour

What if a large proportion of the song plays from any service are accounted for by tracks that users choose to skip? Well, given a skipped track is a played track, the stakeholders in skipped tracks get their full share of the total play pool — a situation that leads to rather ironic outcomes: a skipped track can be worth as much as a track we listen to in full.

Paul Lamere, Director of Developer Platform for EchoNest (owned by Spotify), was kind enough to share, way back in May of 2014, an in-depth look at song skipping behavior within Spotify. This piece was picked up by the Guardian and other sources.

Mr. Lamere found, through an analysis of billions of song plays on Spotify, that we skip, on average, 14.65 tracks per hour.

This number may seem incidental if it were not that case that we can likely only listen to around 15 full tracks in any hour, given the average song is around four minutes in length. At four minutes per song, on average, the best we could do without any skipped tracks is listen to 15 songs per hour. But if some part of that hour is spent listening to a few moments of songs that we skip? Well, it quickly becomes clear that we skip over a greater number of tracks per hour than we listen to in full.

Now, according to Lamere, 24% of song plays are likely skipped within 5 seconds, 35% of are liked skipped within the first 30 seconds, while nearly half (48.6%) are skipped before the song finishes. If 48.6% of songs played are skipped before they finish, and 14.65 tracks are skipped per hour, then 30.14 tracks are played on average each hour (14.65 / 48.6% = 30.144).

As such, and supposedly, there would be 14.65 songs skipped per hour, leaving 15.49 songs to listen to in full each hour.

And so, since some proportion of songs skipped count as songs played in the pro rata calculation of royalties owed:

Songs we skip over, rather than songs we listen to in full, may account for nearly half of the royalties paid by modern music services.

As an aside, these numbers also suggest that the average song played on Spotify is clearly less than four minutes in length—a topic for another discussion.

 The Thirty-second Rule of Thumb

While webcasters in the US pay full royalties for songs skipped (even after a few seconds), there is some debate out there over whether music services such as Spotify, Rhapsody, etc. pay similarly. For example:

Tim Jonze over at the Guardian, while covering a story on silent songs on Spotify, suggested that songs skipped within the first 30 seconds do not trigger a royalty (thanks @rmirchandani for the link). Stuart Dredge, of Music Ally and the Guardian as well, has also suggested this royalty-free safe haven for skipped tracks.

And so, what proportion of royalties paid would be attributed to skipped tracks if only those tracks skipped after 30 seconds were counted? Let’s dig in:

If 48.6% of songs played are skipped before they finish, and 14.65 tracks are skipped per hour, then 30.14 tracks are played on average each hour (14.65 / 48.6% = 30.144) and 15.49 songs are played in full (30.14 – 14.65 = 15.49).

If 35.05% of songs skipped are skipped prior to 30 seconds, and 48.6% of songs are skipped prior to the finish, then roughly 13.55% of songs played (48.6-35.05), or 27.88% of songs skipped are skipped between 30 seconds and the finish of the song.

27.88% of 14.65 songs would be 4.08 tracks.

Under the 30-second rule, the total number of royalty bearing tracks played would be 15.49 (the full tracks) plus 4.08 (the tracks skipped after 30 seconds : or roughly 19.58 tracks.

4.08 tracks would be 20.86% of these 19.58 royalty-bearing performances.

I also did some maths based upon an assumption of the average song being four minutes in length, but those maths are truly too nerdy to show here.

My guesstimate, under the 30-second rule of thumb:

Under the 30-second rule of thumb, using the Echonest/Spotify ratios for plays/skips, 20.86%, or just over one-fifth of royalties paid would be accounted for by skipped tracks.

Under the 30-second rule, but assuming 4 minutes songs on average, 23.51%, or nearly a quarter of royalties paid, might be accounted for by tracks we skip after 30 seconds of play, but before the track is finished.

Browsing as a business model

A reasonable person would conclude that skipped tracks are evidence of music transactions that might not otherwise have occurred in a world of Track and Album sales. When we consider the apparent fact that a non-trivial proportion of royalties paid by music services are paid to the stakeholders of tracks we would rather not listen to than to those we would rather listen to, a new dimension to the royalty discussion emerges:

Browsing is now a crucial part of the business model for music.

Within this business model (browse over to the the business model canvas for a framework), browsing is a contributor to the revenue stream of music stakeholders, or the cost structure of music services. When, for music retail, the situation is reversed: browsing is part of the cost structure for music stakeholders—paid out in marketing dollars for endcaps, listening stations, etc.— while a contributor to the revenue stream for music retailers.

As a result, the dollars paid and received for these skipped tracks should probably be included within the broad discussion of music royalties, whether fair or unfair.

Tracks versus Albums versus Streams: The time value of money and music

This article is a rehashing of one from years ago, given the “streams versus sales” debate continues to rage.

A greater number of artists are posting their payouts from streaming services (e.g., Spotify, Deezer, Rhapsody, rdio, MOG etc.).

And artists (as well as labels), such as Taylor Swift and Big Machine, continue to pull their music from streaming services — calling the payouts “worth-less,” given download sales seem to be worth-more — at least in the short term.

Folks like myselfMark Mulligan, Philippe Astor, and others (the list is getting long, and now spanning over years rather than months) have taken the time to compare streaming service payouts to those payouts from radio, track downloads, webcasting, and even CD sales.

In this nerdy post, I am hoping to take this conversation up another notch by taking into account the so-called “time value of money (and music),” given the fact that streaming payouts will arrive over long periods of time while downloads and CD sales happen today.

Foremost, the streams versus sales debate often confuses two facts related to how revenue arrives differently from these two models. To be clear:

When you get paid for a download, you are being paid in advance for all subsequent listens to that track by some fan.

On a streaming service, you will be paid over time for each subsequent listen to a track by some fan.

And so, a sale is not the same thing as a stream. Instead, a record sale is akin to a whole series of streams experienced – and paid for – over the life of the customer if not also the copyright.

If this difference in payment model is not taken into account, the money from sales and those from streams are not in the same currency.

Warning: I am treating a recording and the underlying song like assets. I believe those assets have value. Sometimes, that value is earned today (as with download sales today). At other times that value is earned over time (as with streaming services).

The Maths:

The value of a track stream is equal to the value of a track sale at around $0.00737 payout per stream, given only a single play during each of the 95 years of the copyright. The maths are simply $0.70 (the royalty pool from a track sale) divided by 95 listens.

If you think the useful life of a recording is only 50 years, however, then one play a year over the next 50 years equals a track sale today at around $0.014 payout per stream. The maths are simply $0.70 (the royalty pool from a track sale) divided by 50 listens.

People usually listen greater than once a year to music they appreciate. And so, the sale versus stream equivalent hovers at simply $0.70 divided by the number of lifetime listens to any track, or tracks on average.

Ironically, we pay less per listen when it comes to albums or tracks we purchase and listen to the most. As a result, the per stream equivalent for any sale is lowest for those tracks we listen to the most and highest for those tracks we listen to the least.

And stream revenue from tracks a listener would have never purchased is revenue an artist would not have otherwise received.

The value of money over time most likely does matter, however. So here goes…

Time Value of Money

The time value of money is a concept from Finance that basically suggests that $5 you receive years from now is “worth less” in today’s dollars than $5 you might receive today. I won’t go into great detail here, but this concept of time value is at the core of just about any modern and functioning financial model.

Given this time value of money, we can think of a very simple “trade” that artists should consider. And, as often happens in an economic sort of example, I am about to really simplify the world in order make this trade as direct as possible:

Imagine that the world is comprised of only two possible fans, and you get to pick only one of them: One of those fans will buy a download today. The other fan will enjoy your music through a streaming service for years to come.

You would be paid $0.70 — today— from the fan who buys the download today. From the other fan, you would receive payments-per-stream at the end of each year — over time — over the life of your copyright (95 years).

At what price per stream, and number of streams per year, would the two fans — the streamer and the downloader — be paying you effectively the same amount of money in today’s dollars?

And so, I am going to (a) consider the value of streaming service payments over the life of a copyright (95 years, depending upon where you are sitting/standing right now), taking into account the time value of money, and (b) compare that value to the payment for a download today.

The Short Story:

Across the following range of payouts per stream, given the value of money over time does matter, the present value of all future payments from streaming services would equal the value of a download sale today at the associated number of streams per year. Most important to these estimates would be the appreciation for risk in these future payments.

UPDATE: I have added and additional estimate in here, at a 5% discount rate, while the original post only considered a 10% discount rate. Just so folks can see how much this “risk” component matters.

At a 5% discount rate

For example: At a payment of $0.0025 per stream, 14.1 streams per year would be the equivalent of a download sale in today’s dollars. At a payment of $0.0075 per stream (3/4 of a penny), 4.7 streams per year would be the equivalent of a download sale.

$0.0025/stream =>; 14.1 streams per year
$0.0050/stream =>; 7.1 streams per year
$0.0075/stream =>; 4.7 streams per year
$0.0100/stream =>; 3.55 streams per year
$0.0125/stream =>; 2.8 streams per year
$0.0150/stream =>; 2.35 streams per year

At a 10% discount rate

For example: In this case, at a payment of $0.0025 per stream, 28 streams per year would be the equivalent of a download sale in today’s dollars. At a payment of $0.0075 per stream (3/4 of a penny), 9.3 streams per year would be the equivalent of a download sale.

$0.0025/stream =>; 28 streams per year
$0.0050/stream =>; 14 streams per year
$0.0075/stream =>; 9.3 streams per year
$0.0100/stream =>; 7 streams per year
$0.0125/stream =>; 5.6 streams per year
$0.0150/stream =>; 4.7 streams per year

On the flip-side, across the following range of streams per year, the present value of all future payments from streaming services would equal the value of a download sale today at the associated payouts per stream.

At a 5% discount rate

For example: At an average of 12 streams a year, the streaming revenue would be equivalent (in today’s dollars) to the download revenue at $0.00295 per stream (i.e., just over 1/4-penny per stream).

01 streams per year =>; $0.03550 per stream
06 streams per year =>; $0.00590 per stream
12 streams per year =>; $0.00295 per stream
24 streams per year =>; $0.00147 per stream
36 streams per year =>; $0.00098 per stream
48 streams per year =>; $0.00074 per stream

At a 10% discount rate

For example: At an average of 12 streams a year, the streaming revenue would be equivalent (in today’s dollars) to the download revenue at $0.0058 per stream (i.e., just over a half-penny per stream).

01 streams per year =>; $0.07 per stream
06 streams per year =>; $0.0116 per stream
12 streams per year =>; $0.0058 per stream
24 streams per year =>; $0.0029 per stream
36 streams per year =>; $0.00195 per stream
48 streams per year =>; $0.00145 per stream

NOTE: In future versions of this analysis I plan on altering (alongside other authors) the conditions to take into account things like: (a) a shorter valuable life of a song, (b) the changing shape of streams over time (high demand early, less demand later), and (c) the implications for different sorts of artist careers (e.g., one-hit wonders versus late-bloomers).

The Untold Story

For those who are familiar with time value, or net present value, the numbers above are a function of (really) only a single outside input: the discount rate. The rest of the inputs are presented in the prior section.

Since the “risk” in the project (the recording) is actually captured by the volatility of expectations for the number of plays on streaming services and the prices per play on those services, I could (if I wanted to) make use of only the risk free rate in the discounting.

However, since not everyone learns this “the risk is in the model” approach described above, I have applied a discount rate anyway. That discount rate is, frankly, arbitrary: 10%. Discuss.

With this discount rate in place, the model really boils down to two versions, each of which simply solves for the mix of payouts and plays that leads to a $0.70 present value: (1) a stream of payments at the pre-determined payouts per stream mentioned, and (2) a stream of plays per year at the pre-determined counts listed above.

The two 6×6 tables follow:

Table One, solves for number of plays given a set range of payouts per stream. On the left is a preset range of values for payouts per stream. Across the top are the streams per year that result in a set of payout-play pairings that would hit $0.70 in present value.

Table Two, solves for payouts per stream given a set range of plays per year. Across the top is a present range of streams per year. On the left are the payouts per stream that result in a set of payout-play pairings that would hit $0.70 in present value.

Rock on, right up to and through the New Year.

The Songwriter Equity Act — peeling away the price of eggs from the price of music

The Songwriter Equity Act is now making its way through the aisles and back alleys of Washington, and this legislative wandering is now accompanied by the usual supply of sound bytes in the hopes of surfacing support for the Bill. This time, eggs are involved.

One important premise of this Act would be that some version of “willing buyer, willing seller” transactions be used for inputs into the decisions made by the Copyright Royalty Board when it sets rates for use of musical works (aka, songs not recordings). The proposed language:

The Copyright Royalty Judges shall establish rates and terms that most clearly represent the rates and terms that would have been negotiated in the marketplace between a willing buyer and a willing seller. In establishing such rates and terms, the Copyright Royalty Judges shall base their decision on marketplace, economic, and use information presented by the participants. In establishing such rates and terms, the Copyright Royalty Judges may consider the rates and terms for comparable uses and comparable circumstances under voluntary license agreements.

Now, using market transactions to inform music pricing seems rather reasonable. And, I definitely believe that songwriters deserve solid compensation for their creative work.

That said, the sound bytes being used to support the Act are unnecessarily hyperbolic, and may actually defeat the arguments being made if people start crunching the numbers embedded in the bytes.

Let’s first just crack an egg.

In support of the Act, David Israelite, CEO of the National Music Publishers Association (which, ironically, represents Publishers and not Songwriters), apparently pointed out—according to Billboard— that while a mechanical royalty rate of two cents a song was set in 1909, today that rate has only increased to 9.1 cents, while eggs that would have cost 14 cents in 1909 now cost $3 at the grocery today. As such “the standard rates of inflation seem to somehow not apply to songwriters,” Israelite said.

If we stick with the egg-based inflation model, we already have an issue: NMPA numbers don’t align with historical data for egg prices.

The National Bureau of Economic Research has a dozen eggs at 31.5 cents in 1909. The US Bureau of Labor Statistics has collected pricing data for more than a century. They have egg prices at 37 cents in 1913 (a bit higher than the NMPA’s 14 cents), about 52.7 cents in 1965, and at $1.93 in 2013. Not quite the $3 used in support of the Act—perhaps the NMPA buys organic eggs.

As such, egg prices have risen 6.1x since 1909, and 3.7x since 1965. So, perhaps mechanical rates should beat least 12.2 cents.

That said… Music is not Eggs. And, therein lies the rub.

The appropriate price comparison here would be inflation in the price of music, not that of eggs. Unless songwriters are printing lyrics on eggshells. Unfortunately, historic inflation in music prices is not on the side of the inflation-adjusted argument. Historically, musical recordings may have experienced something truly odd: not very impressive increases in price.

In the 1920’s, albums had two sides and a track on each side. Prices were $1-1.50 per album, later falling to 75 cents and in some cases as low as 35 cents for the real discount labels. Let’s just put the average price at $0.75, or 37.5 cents per track. If anyone has better pricing, please send it to me. If you prefer we use the higher price—$1.50— trust me, the argument gets worse for songwriters not better.

In the 1960’s, singles on 45’s tended to be between 55 and 99 cents, in the US. That’s 27.5 to 49.5 cents per track, given two singles on a 45. By the early 1990’s, CD “singles” had risen to $4.99 in wholesale price, but then fell to $3.99 and then $3.49 (here’s an old Billboard article from December, 1995). CD singles, however, tended to have between 2 and 4 tracks on the CD (in the US).

Today, a single on iTunes is between $0.69 and $1.29. According to RIAA figures, the average dollar value of tracks were 99 cents in 2008 and $1.18 in 2013.

Albums appear to have been priced between $3 and $6 by the early 1970’s,  In 1991 the average dollar value of a CD album (using RIAA figures) was around $13.01 (after falling from upwards of $20). The RIAA even argues that CD prices have fallen in the decade from 2000 to 2010, from an average price of $14.04 (2000) to $13.02 (2010). That price may have fallen to as low as $12.33 using recent RIAA figures.

In both cases, a back of a napkin suggests music prices have gone up around 3x (=/-0.5x) over the last fifty, if not 100 years (give or take little room for error). That multiple would put the “inflation-adjusted” price for a song on a record at between 5 cents and 7 cents.

2 cents x 2.5 = 5 cents, or
2 cents x 3.5 = 7 cents.

There may be some better ways to argue for Songwriter Equity.

 

Is a Spotify Free user worth $1.50 per month? Or more…

An ongoing controversy in the land of music licensing relates to the role and value of free music within the complicated financial plumbing of the music industry. Importantly, music is almost always worth something, if not at times a great amount. As a result, free music—like free beer—is rarely, truly worthless.

For example, a simple quesstimate—using recently revealed numbers from Spotify—suggests that Spotify Free users may be worth, on average, $1.50 of revenue each month ($18/year).

[The range for the estimate would be between $1.22/month ($14.64/year) and $1.77/month ($21.24/year), conditional upon whether the true proportion of Paid users within the pool of Total Active Users is 20% or 25%.]

IMPORTANT: If what Spotify really meant to say when it reported the Paid-to-Free user ratio was that the number of Paid Subscribers were 20% of the number of Active Free Users, then the revenue quesstimate above rises to as high as $2.10/month ($25.20/year). Furthermore, if Spotify were over-reporting the proportion of users that are Paid versus those that are Free (for whatever reason), then the figures above would under-estimate the monthly revenue value of Free users.

How did I get to the guesstimate above? The rest of this post describes the algebra on the back of the napkin.

The Back of the Napkin

Recently, the folks at Spotify were kind enough to post “Spotify Explained,” within which the firm uses various data to both describe and justify the underlying business model and impact of the service. Artists may debate whether they like or don’t like the Spotify business model. Regardless, the release of these numbers helps people like me do what we do—attempt to convert little breadcrumbs of data into at least a bit more transparency within this music industry.

Within this report, the firm released one very useful byte of information:

By 2013, the amount of money we earn per user (average revenue per user) has grown to $41 per year. This is an average between our premium users who spend $120 per year and our free users who pay for their consumption by viewing and listening to advertising.

Given the above statement, and the ratio between Premium and Free users (another metric the firm reveals from time to time), we can guesstimate the value of Free users within revenue portfolio. Caveat: Spotify makes reference only to the $120/year ($9.99/month in the US) subscribers in this report, with little to no reference made to the $60/year ($4.99/month) subscriber pool. Therefore, I will work with this simplification of the user pool.

In February 2013, Spotify claimed the service had 6 million Paid subscribers and 24 million total Active Users (Active User =  the pool of both Paid and Free users who had logged into the service during the last 30 days). Spotify also suggested the ratio of Paid subscribers to Active Free Active was “greater than 20%.”

For this analysis, the ratio between Paid and Free users is what matters, rather than the total number of users in either of these bins. Meaning, we would get the same guesstimate regardless of whether there are 6 million, 8 million or 15 million paid subscribers as long as the ratio between Paid and Free Users remains the same. That said, Spotify reveals these numbers in a somewhat confusing way.*

As such, I made use of two bounds for the quesstimate:

(1) The ratio of Free Active Users to Paid Subscribers is 4 to 1;
(2) 25% of Active Users are Paid users, 75% are Free Users.

Note that point (1) above would mean that the number of Paid Users is 20% of the number of Free Active Users + Paid Users; given the 4:1 ratio converts to 1/5 = 20%.

From here, all we have is a question of solving for an unknown within a weighted average that leads to $41 per year in weighted average revenue.

(1) (20% * $120) + (80% * X) = $41

$24 + .8X = $41
.8X = $17
X = $21.25 per year in Free User Revenue
X/12 = $1.77 per month in Free User Revenue

(2) (25% * $120) + (80% * Y) = $41

$30 + .75Y = $41
.75Y = $11
Y = $14.67 per year in Free User Revenue
Y/12 = $1.22 per month in Free User Revenue

NOTE: if what Spotify really meant to say was that the number of Paid Subscribers were 20% of the number of Active Free Users, then the quesstimate rises to $2.10 per month, or $25.20 per year.

ASIDE: Anyone paying close attention to licensing rates and revenues should be able to place that $14.67 – $21.25 per year above in the context of recent iTunes Radio licensing parameters.

* Unfortunately, the user figures from Spotify present the ratio in at least two ways. As such, I chose to quessimate things based upon my understanding of what appeared to be two of these ways for communicating the figures.

On the one hand, the firm refers to 6 million Paid Subscribers amidst a pool of 24 million Active Users. This presentation suggests (reasonably) that some proportion of Paid Subscribers are not Active (i.e., did not login to use the service) in any month. 6 million is 25% of 24 million users.

On the other hand, the firms also suggests that ratio of Premium subscribers (who may be non-active in a month) to Active Free Users is 20%. I find a ratio communicated as a percentage to be a bit confusing. And so, I simply assumed they meant: Paid Subcsribers / (Paid Subscribers + Active Free Users ) = 20%. Or, the ratio of Paid Subscribers to Active Free Users is 1:4.

Songwriters Under Attack? Is Fairness really that simple? An investigation of 8 cents per 1,000 plays

Recently, ASCAP launched its “Songwriters Under Attack!” campaign, coinciding with a court case that involved ASCAP and Pandora. The rather sophisticated decision in this case—which hinged on issues of the consent decrees, anti-trust, and the nature of performance rights—did not go in favor of publishers or ASCAP, so expect ongoing disagreement.

In this post, I am going to dig into a very specific claim that ASCAP makes within this campaign: “Right now, every 1,000 plays of a song on Pandora is only worth about 8 cents to songwriters and composers.” The implication of such a small number would be that the amount is unfair. Unfortunately for both myself and artists, the Fairness math may not be that simple.

I cannot help any reader who wants a clear conclusion as to whether 8 cents per 1,000 plays is fair or unfair to songwriters and composers. Undoubtedly, $0.00008 is a small number, but the math of the music industry is often a function of many small numbers added together.

What I can do, however, is try to put such small numbers in context.

As far as I can tell, this 8 cents per 1,000 plays is amazingly similar to the value of a performance to 1000 listeners on terrestrial radio. Meaning, if you divide up the value of a Spin on radio by the number of people listening to that Spin (which is how Streams are counted and accounted for royalties), we end up with a payout to songwriters and composers that is nearly equivalent to 8 cents per 1,000. For the record, when math works out like that, I kinda freak out.

[IMPORTANT UPDATE: Furthermore, I want to make it clear that I have no idea if ASCAP did its math and method correctly to arrive at 8 cents per 1000 as the payout to songwriters from Pandora. I am simply working with the numbers ASCAP chose to  release to the public as part of the campaign. ASCAP’s own numbers may be incorrect, and if those numbers are incorrect then so is the conclusion of the comparison made in this post.]

The equivalence in payments between Big Radio and Pandora could be the result of at least four possibilities:

  1. Pandora and Big Radio pay roughly similar amounts for their performance of musical works, on a per performance per listener basis;
  2. Pandora and Big Radio pay very different amounts for their performance of musical works, on a per performance per listener basis. However, ASCAP throws all “Radio” monies, whether from over-the-Air or over-the-Internet transmissions, together into a big pool and distributes both Webcasting and Terrestrial royalties out of that combined pool. As a result, the differences in payments from these sources wash out;
  3. ASCAP did their math wrong;
  4. I did my math wrong. But at least I will show you my math.

It would be great if we could have a very frank, and open discussion about all of the possibilities listed above, whether #1, #2, #3, or #4.

In my opinion, arriving at a conclusion regarding the Fairness of Artist Royalties requires a far more nuanced (if not also nerdy) conversation than that occurring in the Fairness Debate around these royalties. In this Fairness Debate, people are speaking at various times about quite different things:

  • The fairness of a seemingly small number;
  • The fairness of a less-than-straightforward process;
  • The difference or ratio between two things (e.g. the monies paid to sound recording versus musical work stakeholders, or the percentage of revenue paid for music licenses by Terrestrial Radio versus Satellite Radio versus Webcasters versus On-demand services);
  • “Fairness” as an influential and provocative issue.

As such, I would encourage artists to dig in and really get to know these numbers—the inputs, the outputs, and the assumptions—in order to draw your own conclusions about not only the Fairness in Artist Royalties, but also the fairness in the Fairness Debate.

Digging into the numbers:

As noted earlier, ASCAP has suggested that according to its own internal analysis a play (aka, a stream, a performance to a single listener or device) on Pandora is worth 8 cents per 1,000 plays for the songwriters and composers involved.

Unfortunately, the firm provides none of the inputs to this estimate, other than citing the estimate as the result of “ASCAP internal calculations.” These inputs would really help us. In the absence of these inputs, there are a few things we can do:

Because of how ASCAP distributes royalties—50% for the songwriter/composer, 50% for the publisher/administrator—if $0.00008 is the songwriter/composer share, then an additional $0.00008 is the publisher/administrator share.

And, we need to adjust this  $0.00016 total even higher to account for ASCAP’s operating expense ratio (the share of revenues that gets removed before royalties are paid), which was 11.6% in 2012 (the most recent reporting year, according to ASCAP’s Annual Report).

Alternatively stated, since the $0.00016 for the music stakeholders is paid after ASCAP accounts for it’s costs in collecting, distributing, and even litigating on the way to those payments, we need to add the value of these collecting/distributing/litigating expenses to the royalty payouts, in order to hit the total amount paid by a music service (e.g., Pandora) to a PRO like ASCAP. In simple math: (Total Monies Paid to PRO) – (PRO expenses) = Royalty Distribution Pool

$0.00008 + $0.00008 + (X * 11.6%) = X, where X equals to the total amount paid to the PRO
$0.00008 + $0.00008  / (100%-11.6%) = 0.00018, or 18 cents per 1000 plays.

At least on the back of this napkin, ASCAP ultimately collects $0.00018, or 18 cents per 1,000 plays, from Pandora. For every 1,000 plays, roughly 2 cents of the monies paid by a Pandora go to ASCAP expenses, 8 cents cents publishers/administrators, and 8 cents to songwriters/composers.

[To be clear, the estimate above is based upon my interpretation of the numbers released by ASCAP, not upon my interpretation and analysis of Pandora’s public financials and metrics.]

So how does this $0.00008, or 8 cents per 1,000 plays compare the the value of performances on terrestrial radio? We’re going to need the back of a bigger napkin. That said, I’ve picked up a few tricks over the years, in an effort to simplify this problem.

The simple solution to the problem is to realize that the various PROs collect a total dollar amount in exchange for all the music performed by Terrestrial radio. Even simpler, the monies collected by the PROs in any year for every active radio listener, are in exchange for all of the music that listener hears during that year. The numbers below are rounded for the sake of simplicity.

$385,000,000
Estimate of Total dollars collected from US Radio by ASCAP, BMI, and SESAC (2012)*

240,000,000
Total number of monthly, active radio listeners in the US (Arbitron, 2012)**

$385,000,000 / 240,000,000 = $1.60 collected per active listener.

Now we just have to ask ourselves, how many songs on the radio did that single, average listener experience in exchange for that $1.60?

14.46
Average weekly radio listener hours (RAB, 2012)

52
Weeks in a year

12
Songs experienced per hour (Radio tends to use more time each hour for Ads than Webcasting)

14.46 * 52 * 12 = 9023.4
Songs experienced by the average listener in a year (which can include listening to the same song more than once).

Now, we just divide the dollars collected per listener in 2012 by the number of songs the average listener experienced during that year.

$1.60 / 9023.4 = $0.00018, or 18 cents per 1000 plays.***

According to the back of this other napkin, PROs collects around $0.00018, or 18 cents per 1,000 plays, from Terrestrial Radio. Which would mean that for every 1000 plays roughly 2 of these cents would go towards the expenses of the PROs, leaving 8 cents for the publishers/administrators and 8 cents for the songwriters/composers.

For the record, and as noted earlier, when math works out like that, I kinda freak out.

For those who care, what follows are the Radio estimates if I use 15, 12, or 10 songs per hour.

15 songs per hour -> 14 cents per 1000 plays
12 songs per hour -> 18 cents per 1000 plays
10 songs per hour -> 21 cents per 1000 plays

In the end, and to repeat, I cannot answer the question of whether 8 cents per 1000 plays for the songwriters and composers is the truly “fair” amount for the use of these songs on Radio, whether over-the-Air or over-the-Internet.

In my opinion, and as I noted at the beginning of this post, arriving at a conclusion regarding the Fairness of Artist Royalties requires a far more nuanced (if not also *incredibly* nerdy) conversation than that occurring in the Fairness Debate around these royalties.

 

*This amount does not include the additional pool of monies paid by some Terrestrial Radio networks, stations, or programs to publishers, administrators, songwriters, and composers by way of direct license arrangements. As such the total amount paid by so-called Big Radio for the performance of musical works likely exceeds the amount paid to PROs. As a result, I may underestimate the value of a performance.

** I did not adjust this audience figure for that proportion which is Talk Radio since I do not know what portion of this audience, or weekly listening, is exclusively Talk, and what proportion is both Talk and Music. If I assume the so-called Talk Radio audience at any moment is exclusively Talk throughout a month or year, then the final royalty estimates would increase by up to 15% (the rough estimate of the Talk Radio audience). As a result, I may overestimate the value of a performance.

The combination of * (leading to a potential underestimate) and ** (leading to a potential overestimate) above might easily be a wash.

*** It would be mathematically implausible for me to reckon which proportion of songs on the radio are “ASCAP” songs, and in turn, re-calculate this figure based upon only that % and ASCAP’s revenue alone. But, it seems plausible that ASCAP’s proportion of total PRO collections, or 45-46%, is likely similar to that proportion of music on the radio that might be ascribed to ASCAP as the PRO. In the end, I find this issue a wash.

Three things a Copyright nerd immediately notices when using iTunes Radio

If you are not a copyright nerd, you won’t really be motived by the following analysis of iTunes Radio. However, if you are such a nerd, then you will likely and quickly realize what Apple obtained through its licenses.

  1. I could (at times) seed a new station with my own song choice. Right there, the license steps outside the limitations of the Law. That said, this ability was not always possible, so there may be some “random” selection of tracks on an album, which means that only at random can I select the specific song that seeds a station.
  2. I was afforded six (if not seven) skips for each new personalized station I launched. Meaning, my listening was not limited to six skips per hour, but rather my listening to each station—no matter how many stations I created—was limited to so many skips. The limit on skipped tracks applied whether I skipped before or after 20 seconds.
  3. While not exclusively the interest of copyright nerds, I found that my first effort to skip a track led to a streaming video ad filling the iTunes app. In fact, I have yet to see or hear an ad other than after a skipped track. Given Apple does not owe any royalties for skipped tracks, I find it interesting that my experience with monetization immediately follows tracks that have no royalty obligation.

That’s it for the nerds today.

$0.0001: The value of a Radio performance, given rumored ClearChannel/Warner deal terms

Earlier today, the rumor mill milled out a rumor that Clear Channel and Warner Music inked a deal covering royalties related to the performance of sound recordings on Clear Channel’s radio stations.

As some might expect, the rumored deal terms sent me directly to a calculator, in and effort to value a performance, per listener, on Clear Channel US terrestrial radio given the rumored deal terms.

To be clear, the following is as estimate for the performance of sound recordings, on a per listener basis. Akin to the way in which Webcasters are asked to pay for their use of recordings on a per stream (to each listener) basis.

The back of the napkin estimate:

between $0.00009375 and $0.000107 per performance, per listener.

Simplified as: $0.0001 per performance, per listener. Or, less than 1/10 the Pureplay rate, and about 1/22 the General Webcaster rate.

The inputs:

$15 billion
the rough estimate of radio revenues.

$150million,
1% of radio revenues. This is the rate rumored to be defined in the deal. A rate that looks mysteriously just like that proposed by the RIAA to the NAB only a year ago.

1.4trillion (low estimate)
1.6trillion (high estimate)
the number of unique performance, per listener across US radio

Which is based on…

14.46, weekly radio listening hours (RAB, 2012)
52 weeks in a year
10 songs and hour for the low estimate.
9 songs an hour for the high estimate.

The napkin math:

$150,000,0000 / 1,600,000,000,000 = 0.00009375

$150,000,000 / 1,400,000,000,000 = 0.00010714

Comparing Apples to Webcasters: Making sense of the differences in royalty rates and terms

Now that some version of Apple’s iRadio contract terms are available online, we can begin to do some comparisons between the effective rates contained in these terms and those statutory rates paid by Webcasters.

In this post, I will (eventually) present some simple tables that adjust Webcaster rates to account for certain affordances in the Apple contract — specifically those terms through which a stream on iRadio would not trigger a royalty obligation.

The purpose here is to try to explain why the question of whether or not the effective rate paid by Apple under the iRadio terms will prove to be greater than or less than those effective rates paid by other webcasters ultimately relies upon the nature of user behavior (e.g., skipped tracks) and the composition of the user base (i.e., free versus paid users). I fear the “straight up” comparison between the raw iRadio and Pureplay rates being offered in most outlets overlooks these important nuances.

In the following analysis I will speak in “RPM” (Royalties per “Mille,” or Thousand streams) terms. An RPM is simply the Royalty obligation Per 1000 (Mille) performances/streams. For example, the Pureplay rate of $0.0012 would convert to an RPM of $1.20. The iRadio contract uses listener hours for its rpm, which I think is a bit confusing.

Furthermore, I am neither going to take a stand on whether payments per stream are relatively high or low, nor present some conclusion as to whether iRadio rates are greater or lesser than those rates paid by Pureplay or General Webcasters. I am simply going to present tables based upon rather straightforward math, hopefully enabling the reader to come to their own conclusion based on their own understanding of, or assumptions about, online radio listening behavior.

The short story:

Webcasters operating under the statutory rate/terms pay different rates for ad-supported streams as compared to those rates paid for streams supported by a paid subscription. Furthermore, these webcasters pay a full stream royalty for partial streams of a song, even streams that are skipped in the first few seconds. As a result of the mix of the user base and the cost of skipped tracks, the effective rates paid by a webcaster differ from the “raw” rates established by the CRB.

For example, after accounting for the mixture of Free and Paid users in the Pandora user base — the webcaster to which iRadio is most often compared and for which the most public data exists — and the different royalty obligations for these different users ($1.20 RPM for free users, $2.20 RPM for paid users), I’d estimate that the initial (i.e., pre-cost of skipped tracks) effective rate being paid by the biggest US webcaster is about $1.24 RPM.

This value is greater than the raw Pureplay rate simply because somewhere around 3.6% of Pandora’s users are paid (upgraded to Pandora One) users, who trigger a higher royalty rate owed by the Webcaster than that owed for free (ad-supported) users. Conversely, and unless I am mistaken, according to the iRadio contract Apple will not owe royalties for streams to users who have upgraded to iTunes Match — even if the track being played is not in the user’s collection.

Considering the issue of skipped tracks per hour (up to six) for which webcasters do owe royalties, but iRadio would not, at an average of 15 fully played tracks each hour, you end up with a range of effective rates that increase quickly:

0 tracks skipped -> $1.24 RPM
1 tracks skipped -> $1.32 RPM
2 tracks skipped -> $1.40 RPM
3 tracks skipped -> $1.48 RPM
4 tracks skipped -> $1.57 RPM
5 tracks skipped -> $1.65 RPM
6 tracks skipped -> $1.73 RPM

The above simply means that given 15 fully played tracks and 3 skipped tracks in an hour, a ‘caster like Pandora would pay an effective, overall rate of $1.48 RPM. For further background, an exclusively Pureplay webcaster would owe $1.44 RPM, while an exclusively General Webcaster would owe $2.64.

With iRadio’s minimum royalty obligation being around a $1.42 RPM ($21.25 for 1000 listener hours at 15 tracks per hour), the royalty rate debate gets a bit more complicated. Comparing Apples to Webcaster requires coming to an understanding of where on this scale of skipped tracks and within this mixture of free versus paid users any webcaster resides.

The Long Story:

A wise lawyer once said, “Always read the [expletive] contract.” And so, we are going to work our way through the most royalty-relevant terms in Apple’s iRadio contract.

The Terms of Interest

Two particular issues come into play when comparing iRadio rates to those paid by other webcasters: (A) the performances for which Apple will not owe royalties, but statutory webcasters would and (B) the rates paid by webcasters based upon the classification of their service(s).

In the case of issue A, and according to the publicly available contract, Apple will not owe a royalty for:

  1. Up to six song streams in any hour that are skipped within the first twenty seconds; and
  2. Up to two streams each hour that come from Listener Matched Content (a track that is in your collection), a Complete-My-Album Play (a track you don’t own, but which comes from an album from which you do own one or more tracks), or a Heat-Seeker Play (aka, Promotional play); and
  3. Any streams to iTunes Match subscribers, labeled “cloud services” in the contract.

All three of these points are important simply because Webcasters operating under the statutory terms, whether classified as General or Pureplay webcasters, owe a royalty for each skipped track — whenever any portion of any sound recording is streamed/performed — and regardless of whether the track is in your collection, from an album in your collection completed with this track, or a promotional play.

In the case of issue B, webcasters pay different rates based upon the nature of their service. For example, the Pureplay rates, $0.0012 in 2013, to which most people compare iRadio rates would only come into play for Ad-supported services. Paid/subscription radio services (e.g., Slacker, or Pandora’s One tier) pay a different rate, $0.0022 in 2013. As such, you need to consider this blend of Free alongside Paid users to estimate the overall and effective rate paid by a Webcaster.

As such, the difference in obligations for Apple’s Upgrade tier (i.e., iTunes Match) and other Webcaster’s upgrade tier(s) matters a great deal. Unless I have misread the contract, Apple will not owe royalties for iRadio streams to iTunes Match subscribers — even if you don’t own the track being played. The population of Webcasters, however, owes a higher rate ($0.0022 per play rather than $0.0012) for these paid users.

Don’t Forget the Fine Print

In the Apple contract, a minimum royalty obligation exists. This minimum has two tactical dimensions:

  1. 45% of Net Advertising Revenue, or and
  2. $21.25 in royalties paid per 1000 listener hours (in the first year), and $22.25 per 1000 listener hours for the remainder of the deal.

These dimensions are tactical simply because they line up with past language in webcasting statutes, which tied rates to some percentage of revenue and/or some value per aggregate tuning hour.

Importantly, these terms for the minimum obligation are positioned “Notwithstanding” the foregoing terms (i.e., those that include the ups and downs of streams that do or don’t trigger a royalty). And so, this minimum is the minimum — any perquisites in the contract for skips, match content, etc. cannot lower the obligation below this minimum.

Given the fine print, Apple would seem to most likely owe the minimum RPM of approximately $1.42 (i.e., $0.00142 for each stream, $1.42 for 1000 streams), based upon the $21.25 for each 1000 listeners hours at a rough average of 15 songs per hour, regardless of skips, subscribers or other factors (a value that matches the calculations of Glenn Peoples over at Billboard). For the moment, I will use this minimum as the comparable since predicting Apple’s future ad revenue is tough, and few if any webcasters don’t already pay greater than 45% of their ad revenue in royalties.

Comparing Apples to Webcasters

In order to convert Webcaster rates to iRadio rates, all you really have to do is recalculate the effective per stream rate any webcaster might owe, after accounting for their royalty obligations given (a) the mixture of free and paid users and (b) the number of tracks a user might skip in any hour (up to six). Since Pandora is the webcaster against which iRadio is consistently compared, and the one with the most publicly available data, I’ll target this caster for the comparable.

In May of 2013, Pandora announced roughly 70.1 million active users, of which 2.5 million were paid subscribers. This free versus paid composition means that approximately 3.6% of Pandora’s users are paid listeners (at $2.20 RPM) and 96.4% of the users are ad-supported listeners (at a $1.20 RPM). A quick weighted average:

(96.4% x $1.20) + (3.6% x $2.20) = $1.24 RPM

So, after taking into account the different rates paid on account of free versus paid users, it seems Pandora operates under an overall RPM of $1.24.

Now, we simply have to tack on the added cost borne by statutory webcasters by way of folks like myself who either directly skip tracks or indirectly skip tracks via a thumbs down. Since both webcasting statutes and the iRadio contract limit the number of skips to six during an hour, we can calculate rate adjustments for zero to six tracks skips during any hour. Using a general thumb rule of 15 listened tracks per hour, if a user skips two tracks, a total of 17 tracks would play that hour.

0 tracks skipped -> $1.24 RPM
1 tracks skipped -> $1.32 RPM
2 tracks skipped -> $1.40 RPM
3 tracks skipped -> $1.48 RPM
4 tracks skipped -> $1.57 RPM
5 tracks skipped -> $1.65 RPM
6 tracks skipped -> $1.73 RPM

Translating a table like this is pretty simple. For example after accounting for (i.e., paying for) three skipped tracks in an hour of music, for which a webcaster would owe royalties, a Pureplay webcaster is paying an effective RPM of approximately $1.48. If the full monty of skips tends to accrue, the effective RPM grows to $1.73.

Imagine a Webcaster with a split of 90% Ad-supported and 10% Paid users and you would get the following table when considering the impacts of different rate teirs and skipped tracks:

0 tracks skipped -> $1.30 RPM
1 tracks skipped -> $1.39 RPM
2 tracks skipped -> $1.47 RPM
3 tracks skipped -> $1.56 RPM
4 tracks skipped -> $1.65 RPM
5 tracks skipped -> $1.73 RPM
6 tracks skipped -> $1.82 RPM

And so, whether or not the rates paid by Apple under the iRadio terms will prove to be greater than or less than than those paid by other Webcasters relies heavily upon the nuanced issues that are the nature of user behavior and the composition of the user base.

$42 Billion: what Apple could owe in royalties as a truly “Pureplay” Webcaster, if playing by the Letter of the Law

File this post in the “Letter of the Law” file…

The newswires are all a flutter with rumors of Apple’s iRadio deals with both Universal and Warner music. Under the Warner deal, the rumored rate (according to Billboard) appears to be $0.0016 per stream plus some portion or variation of ad revenue (even though the $0.0016 would be paid, at least in part, out of ad revenue).

Billboard and other sources have positioned these iRadio rates against those per stream rates paid by “Pureplay” webcasters such as Pandora. Unfortunately, these sources have, by and large, underestimated the true obligations that an Apple iRadio would face if it were truly to pay royalties as a Pureplay webcaster.

Enter the letter of the Law.

First, and to be clear, if Apple wanted to license as a Pureplay webcaster, it would have had to declare such ambitions by January 31, 2013, in order to partake of the settlement rate. Perhaps Cupertino already declared such ambitions, and is now simply negotiating directly for a better deal. Who knows.

Second, the royalty rates (in the US) against which Apple’s iRadio should be compared are as follows (from the Sound Exchange website):

2013 RATE (Syndicated/Bundled/Subscription Services): $0.0022 per performance (per play, per listener; definition on page 88). (Rate Summary, 2011-2015: 2011: $0.0017 (per performance); 2012: $0.0020; 2013: $0.0022; 2014: $0.0023; 2015: $0.0025)

2013 RATE (Other Services): The greater of 25% of Gross Revenues or $0.00120 per performance. (Rate Summary, 2011: $0.00102 (per performance); 2012: $0.00110; 2013: $0.00120; 2014: $0.00130; 2015: $0.00140)

In simple english, if Apple would classify itself as a “Syndicate/Bundled/Subscription” service it would pay $0.0022 per performance, per listener, under the Law. As such, the $0.0016 rate being rumored in the press is a 27% discount from the established rate for this class (notwithstanding whatever the ad revenue kicker looks like).

If, however, Apple would classify itself as an “Other” service, akin to how Pandora or other Pureplay webcasters tend to be classified — the services against which most of the press is comparing iRadio — then Apple would face the obligation of paying the greater of (a) 25% of Gross Revenues or (b) $0.0012 per performance, per listener.

That “greater of” language is important.

Just so we are clear, “Gross Revenues” in this case really means Gross (as in All) Revenues (and then some). You may think I am kidding, but here is the language of the Law (page 34797 of the linked PDF at the Federal Register):

‘‘Gross Revenues’’ means all revenue of any kind earned by the Commercial Webcaster or its Affiliates from all its operations, in accordance with U.S. Generally Accepted Accounting Principles, and includes –

(A) all cash or cash equivalents;

(B) the fair market value of goods, services, or other non–cash consideration (including real, personal, tangible, and intangible property);

(C) in–kind and cash donations and other gifts (but not capital contributions made in exchange for an equity interest in the recipient); and

(D) amounts earned by such person or entity but paid to an Affiliate of such person or entity in lieu of payment to such person or entity.

For the avoidance of doubt, Gross Revenues includes revenue from activities other than making Eligible Transmissions, including revenue from transmissions of sound recordings licensed directly from the relevant copyright owners. Commercial Webcasters with substantial revenue from activities other than making Eligible Transmissions under the statutory licenses in Sections 112(e) and 114 may wish not to elect to be subject to these Rates and Terms.

In other words, and quite seriously, if Apple’s iRadio were to pay for its use of sound recordings as if it were a Pureplay webcaster in the same category as services such as Pandora, Cupertino would likely have to pay $42 billion, or 25% of gross revenue for their use of music in the service. That’s a really big number.

To be frank, however, this back-of-the-napkin estimate of $42 Billion — or 25% of Apple’s Gross Revenues in 2012 — might actually be an underestimate since it does not take into account the value of cash sitting on Apple’s balance sheet (outside the US) — a big topic in other press circles these days.

Or, alternatively, Apple could pay $0.0022 per performance per listener — a rate the total value of which (in terms of streaming obligations) I can’t calculate at this time.

And so, and to re-iterate, the rates against which Apple’s iRadio should truly be compared given the Letter of the Law would be:

(1) $0.0022 per performance per listener; or

(2) The greater of (a) 25% of Gross Revenue or (b) $0.0012 per performance per listener.

Under (2), Apple’s iRadio would owe $42 Billion for its use of music, not $0.0012 per stream.

Churning Zombies on Ice, Part One: The link between the living dead and the new, new thing in music services.

Churn is a topic in its cooling off period at the moment, with Mulligan, Resnikoff, Peoples, Dredge and others contributing a short time ago to a discussion on this issue. The debate centers on whether the relative proportion of active versus inactive users (a.k.a., zombies) is a good or bad sign for new music services.

I think something more important than churn may be taking place. Or, stated differently, the market grew quickest during a period that likely also offered the highest rate of churn. The US market for music services offers a particularly useful space for digging into this issue.

First, a distinction I would like to make between two kinds of churn: (1) the movement of a subscriber from one service to another (i.e., your loss is my gain) and (2) the loss of a subscriber from not only one provider but from the market as a whole (i.e., your loss is our loss).

I am making this distinction for a very simple reason: When we only consider the user base characteristics for clearly freemium services (e.g., Spotify and Deezer), we likely over-estimate the zombie population. The over-estimation is the result of counting “your loss is my gain” subscribers as “your loss is our loss” subscribers.

As fas a I can tell, of the 3.4 million paying subscribers in the US to on-demand music services:

(A) 15-20% of the 2012 subscribers (650,000 [+/- 100,000] accounts) may have been “your loss is my gain” churn.

(B) 50-65% (2,000,000 [+/- 200,000] of this overall subscriber pool were not likely and directly acquired via a freemium model.

(C) Of the 1.6 million new subscribers in 2012, however, upwards of 60% (900,000 +/- 100,000) may have arrived by way of a freemium model. Said another way, of the 1.6 million new subscribers in in 2012, upwards of 40% may not have arrived by way of a freemium model.

(D) The total population of zombies in this pool, most of whom arrived in 2012, may likely account for as little as 38% (+/- 5%) of the registered user base — after taking into account “your loss is my gain” churn and new subscribers to the market who did not arrive by way of the freemium path. This percentage may be an underestimate as well, since its very difficult for me to capture and/or estimate (without internal data) registered and inactive users prior to 2010.

As always, the long story…

At the end of 2012 I guesstimated about 3.5 million paying subscribers in the US of A to so-called subscription music services (e.g., Rhapsody, Spotify, MOG, Rdio Slacker, Muve, etc.). When the RIAA released its figures for 2012, I was happy to see the guesstimation machine, while less active, was still operating — the RIAA claimed there were 3.4 million paying subscribers at the end of 2012. Hoorah.

Pegging the number of subscribers in 2012 was important. This number could provide the means for a deeper understanding of what does and what does not lead to subscriber adoption of, churn within (i.e., the loss of subscribers), and even cannibalization by music services.

Why? Because the US is one of the few countries wherein music services have been operating for longer than a decade. Rhapsody, neé Listen.com, neé TuneTo’s Aladdin, launched in December of 2001! MusicNet and PressPlay, now defunct are seriously changed in form, launched that same year. We have had subscription music services for nearly twelve years.

So what? Because the number of subscribers to new music services in the US had remained relatively constant for right around half of that decade, perhaps even contracting a bit during 2010. Most of the initial subscriber growth happened in 2004-5, only to hold largely steady through 2011. And then the number of suddenly bumped upwards during 2012.

For context, there were around 1.8 million paying subscribers in the US to music services in 2011. In 2005, there were around 1.8 million such subscribers. You don’t believe me? Here are some stories to back up the claim : CNN (Rhapsody, 1.3 million; Napster, 448,000); Rhapsody press release 2006, still 1.3 million subscribers; The RIAA placed the total number of subscribers at 1.5 million in 2010. By 2011, Rhapsody falls to 800,000, Napster holds 3-400,000, while Muve, Rdio, Slacker, and MOG capture the remaining for a total near 1.8 million.

So, so what? Because the number of paying subscribers expands in 2012 by 1.6 million accounts, or a 60% growth rate. Something likely happened during 2012. The question is what, or what combination of things.

Furthermore, subscribers who pay for non-interactive radio (e.g., Pandora and some proportion of Slacker paying subscribers) are not being counted as paid subscribers by the RIAA. These numbers may not be trivial. For example, by my guesstimate Pandora had an average of 1.4 million paying subscribers during 2012.

The play by play:

Muve music grew from zero users to 1.1 million from January 2011 to December 2012. No freemium model.

Spotify grew from slightly greater than zero users to 1 million between July 2011 and December 2012. Freemium.

Rhapsody falls on slightly from 1.2 million (via combination with Napster) to 1 million, by December 2012. Mainly no freemium model.

The remaining roughly 300,000 subscribers reside across MOG, Rdio, Slacker (full paid, not radio paid), and whoever else is out there. Most of whom were in place prior to 2012. Moderately freemium.

Back of the napkin math, taking into account users prior to 2012 and those at the end of 2012 led to the conclusions above the fold.