Is a stream on Spotify (or any music service) really worth less than an iTunes sale?

On numerous occasions over the past few months we have heard news of labels and artists disappointed with the monies they are receiving from Spotify.

Since Spotify is really just one of the music services alive and kicking on the planet these days (alongside Rhapsody, MOG, Rdio, Deezer, and others), its worth investigating this matter of the value of payments from streaming services versus those received from record sales.  So let’s get a napkin out and flip to the backside for some mathy stuff.


When you purchase a download you are paying in advance for all of your subsequent listens of that track.

When you listen to a track through a streaming music service, you are paying as you go for each and subsequent listens of that track.

First off, its important that we compare apples to apples and not apple to oranges. Comparing the value of a stream to the value of a sale is without question an Apples to Oranges comparison.  And so, let’s convert the sale to an estimate of payment per stream so we can compare apples to apples (or oranges to oranges if you prefer).

Second, let’s make a simple assumption: A single track purchased from iTunes will be listened to 250 times, on average, over the lifetime of the person who acquired the single.  This assumption may not seem that crazy—over twenty five years will you listen to the “average” purchased song in your library about 10 times per year?  Or twenty five times per year over the first ten years?

Dividing the amount paid for a single by the number of times you listen to that single over the lifetime will give us the effective rate you have paid for each listen of the track.  This payment per listen is the closest equivalent to the payment per stream courtesy a music streaming service.

Third, now we can just divide the amount music rightsholders would earn from a sale on iTunes to estimate a “per listen” fee—the effective rate paid over the lifetime of the track’s enjoyment for each listen of the song.

$0.70 -> the amount a label receives for the sale of an iTunes single (US)
250    -> the number of assumed listens to the track over ownership life
$0.70 / 250 =  $0.0028 per listen

OK.  So some quick math has led us to the conclusion that our “per listen” payment for iTunes singles over the lifetime of our enjoyment of these tracks is just short of 1/3 of a penny.

From the perspective of songwriters and publishers, in the US, some of this $0.70 above should be flowing back to you—about 9.5 cents (give or take for those various alterations like controlled compositions). And so, the per listen portion paid to the songwriter/publisher would be around $0.00038, leaving about $0.00242 for the sound recording owner(s) and performing/featured artist(s).

To repeat, if you convert purchases of downloadable singles to some value “per listen,” we get the following:

$0.00242  -> for the label(s) and performing artist(s)
$0.00038 -> for the songwriter(s) and publisher(s)
$0.0028   -> total amount paid to rightsholders per listen from a purchased track

Now we can ask, Is $0.0028 greater than or less than what rightsholders are receiving per stream from streaming music services?

Since disclosing the licensed rate at which these music services pay for music per stream would be controversial behavior, let’s just use the going rate as suggested by those who have chosen to disclose (or claimed to disclose) their payments from Spotify.  That rate is apparently about $0.0033 per stream paid to the label, some portion of which may or may not ever make its way back to the artist(s).

$0.0033 -> the amount believed to be paid per stream via Spotify (for SR only)
$0.0028 -> the effective “per listen” payment for a purchased download

$0.0033 > $0.0028

Could it be true?

Is the per stream rate paid by music services actually greater than the effective per listen rate we pay for a purchased download?

Next up, we will use some not-so-fancy financial tools (NPV) to compare the present value of these two payments schemes: up-front versus over-time.



    1. Thanks for your thoughts. NARM would likely have survey data just like most others. I will further investigate with actual listening data.

      That said, I am pretty skeptical that the 250 listens is off by a factor of 10. We are talking about the listening patterns for general consumers over decades not a few years. And the average of listens among the tracks in their library.

      These averages may be very different for music purchased than for that acquired otherwise. And for music “rented” as compared to music purchased.

      But better inputs would be better.

      Regardless, the purpose of the estimate was to trigger the comparison of a per-listen rate baked into the one-time payment versus a more explicit rate being paid by music services.



  1. Great write up, this is the exact thing that has been going through my head lately. It’s short sighted to want the larger sum up front, when artists seem to clearly be better off in the long run to get paid per listen forever.



  2. Fascinating analysis of the “rent v own” facet of “streaming v. download”. I’d think you could use scrobble data for a sample of people to get a still-anecdotal but better assessment of your plays per song theory.

    I bought a bicycle lately, and am trying to see its value in “cost per ride’, a decreasing figure that reaches palatable levels only if I use my bike properly :).

    I use the streaming service MOG, and I am much more willing to stream unfamiliar material than I am willing to buy a CD of unfamiliar material. The monthly pricing (I use the pay plan rather than the free plan) allows me to stream lots of music I did not know and would not have bought. The right to ‘rent’ this material by the month is certainly a less desirable option than owning the mp3, or we’d see the streaming services taking over from the download market.

    Also, I’d think, on the face of it, that lots of folks would go listen to a song on spotify or rdio due to a tweet, but far fewer would buy a CD due to a tweet. Streaming, in this sense, is more social-media friendly, and perhaps a pragmatic argument that folks ought not get so worried about the current per-stream royalty is that it preserves a marketing advantage for streaming in social media.

    On the other hand, the differential per-stream royalty paid to the majors is a travesty even despite the good points you make in this post.



    1. I reckon another known unknown here is any uptick in listens to a “free” streaming tracks versus those to a purchased download.

      If five times as many people listen to the free track versus the paid track, the ratio of lifetime listens for pay as you go gets a boost.



  3. There will never be a true comparison in that a purchase is a completely different business model than a subscription. I purchase a track not to “save” money (theoretically true if I listen to a track many times more than 250), nor do I subscribe to spotify to stick it to record labels (theoretically true if I listen once and move on). I suspect Spotify pays record labels the way radio does, by a fee based on listeners. Labels are then supposed to disperse funds to artists. I bet the trouble is not with how much Spotify pays but with how much labels keep.



    1. what about the tens of thousands of artists on there who aren’t signed to labels, or who run their own labels just for their own recordings?



  4. Interesting analysis. However, you are not factoring in the net present value of the money – you need to discount dramatically the money that is assumed to be earned in the ‘long run’. On a NPV valuation, it would always be better to get your money upfront from a download sale, than to see it come through to you over 10 years. In fact, the p/stream rate of a listen that happens in 10 years time would be of significantly less value than the current valuation of $0.0033 (in terms of the current value of $).

    I also agree with others that your 250 listens p/track is incredibly arbitrary – you need to have some research to support this for valuations to have any weight. I’d suggest that the lifetime “listening rate” p/track would be significantly less when averaged across the typical iTunes users catalogue. I bet there is research somewhere to deliver this crucial number.



    1. Thanks Jon.

      As noted in the blog burst, I will be trying to address the issue of NPV in a subsequent post.

      And has been alluded to elsewhere, the “discount rate” will matter a great deal.



  5. 3165 in my collection on this laptop.

    most listened-to track played 110 times.
    that track added to my library in february this year.

    2nd most listened-to track played 54 times
    that track added to my library in march this year.

    Age of the entire library is 26 months.
    Average plays across the rest of the library is 4.

    250 plays at current average listen rate (assuming the library doesn’t grow and spread the listens even further) would take 62 years.

    If I wrote music I’d take a one-off payment for a track over streaming any day. the maths makes no sense.
    how are people supposed to fund their music-making while they’re waiting for these POSSIBLE revenues to appear?

    comparing unit payments is irrelevant. the important question is how what you pay the bill that’s due tomorrow, not the bill that’s due in 3 years time.



    1. Thanks for sharing you listening data. And I reckon you are correct in being concerned.

      Regardless of whether the per listen rates are identical, one set of payments arrives immediately while the other has to be deferred.

      The need to be paid now, rather than later is a need that labels and publishers could exploit in the past. The artist upfront payments were, in general, less than the value of future payments. That condition changed in the last decade.

      My sense is that this shift from upfront to as you go is unavoidable.



      1. then we will see a lot of indie labels die. they’re small businesses relying on each release to fund the next one. cash flow is a massive issue, not some abstract thing.

        this factor of timing is a crucial missing piece in the discussion of indies and streaming.


      2. I agree. In the next update to this post I am going to look at the “time value” of this difference in payments. The implications are important.

        However, i do believe we really shouldn’t be surprised if an access model payment is smaller than a purchase model payment. The same is true for TV and Film—the DVD sold is worth more (per view per viewer, in particular) than a “stream” on TV or online.

        And, ultimately, the bigger question is whether access models expand the market. The instinctual response at this point appears to be straight up substitution.


  6. Spotify does not claim to pay per stream. They have stated this many times. Here is one of their statements:

    “Spotify does not sell streams, but access to music. Users pay for this access either via a subscription fee or with their ear time via the ad-supported service [just like commercial radio] – they do not pay per stream. In other words, Spotify is not a unit based business and it does not make sense to look at revenues from Spotify from a per stream or other music unit-based point of view.” -source: hypebot

    What their payments are based on they never say, but it is not per stream.

    Indie label Projek pulled their catalog stating that when they divided the number of streams on Spotify by the payments the payments per track came out to $.0013, not your figure of $0.0033. Other artists and labels have also stated being paid far less.

    As for sticking it to the record labels, the major labels all have invested in Spotify and are thus making money from the service from ad revenue and other income streams, not per pay. (Revenue that no artist gets paid from.) So in the end, it is the artist, both signed to a label and independents who, once again, are getting screwed.

    Like others who have commented, my iTunes library, listed by number of plays shows no songs being played 250. One track lists 221 plays, another 207, but then it drops to 83. My iPod shows higher numbers, but only the first dozen songs shows plays at >250. The rest are exponentially lower.



    1. Thanks for your comments.

      Spotify may not sell streams to end users, but their deals with labels have both per user and per stream minima. In some cases, they have blanket licenses for authorial and publishers performance rights.

      I do not think spotify has denied the above. They simply remark that they don’t sell streams to end users.

      And please note that I was looking at the number of plays over a line period of time–a few decades.

      As such, would you mind sharing the time period for the number of plays you described?

      Beyond that, once back from holiday I will source the data I used to make the $0.0033 estimate.

      Regardless, my opinion is this value debate is a good debate to be having.



      1. Expecting plays to occur over decades is probably unrealistic. People’s listening habits simply aren’t so linear.

        I don’t listen to the same music I was listening to 5 years ago, nor 10 years ago, nor 20 years ago.

        There may be *possibly* one or two tracks from my youth that still get a play from time to time, but they’re very much the exception rather than the rule.
        You also have to consider that a person’s knowledge and tastes broaden hugely over the years which also reduces the likelihood of the same tracks or artists getting played over and over.

        When I was 3 years old I’d only heard a handful of songs, and I wanted to hear those songs over and over. Now I’m in my 40s and I’ve heard hundreds of thousands of songs and I rarely want to hear the same thing twice.


      2. OK. If it is the case that you or many other people would rather not listen to the same track more than a handful (or small bucket) of times…

        Then why would you ever buy an album or single in the first place? Have you really been spending significant money on music only to never listen to the same track twice?

        At some point, it seems like access models and per stream payments are the models that match your listening habits.

        And so, from an artists perspective, comparing the two payments is beside the point—one model fits the needs of one sort of customer, the other model fits the needs of a different sort of customer.


  7. It’s doubtful whether Spotify will exist in 5 year’s time, so counting ‘streams’ that far in the future is dubious. As well as a conventional discount factor, you would need to discount future streams by the (im)probability that they will be available at all.



    1. Perhaps true, regarding Spotify’s prospects. But the larger issue here is purchase versus access models and their implications for creators.

      Spotify is just an example of an access model, as would be other music services. Rhapsody, for example, has been around for 7-9 years already.



  8. One thing that hasn’t been mentioned is the issue of depth of catalogue, which I think has a significant impact on total track listens. Put simply, the greater your choice of listening, the smaller the chance that any particular track will be played over and over.

    A private music library will have a limited amount of songs you can listen to. Assuming that you find individual songs satisfying to different levels and that the satisfaction derived changes over time (you get bored after a while), you will probably find similar listening patterns as described by commenters above: a small number of standout songs, in terms of number of plays (those that provided most satisfaction for the longest period of time), and a lot of songs that were listened to only a couple of times.

    Spotify presents a different scenario: here the volume of choice exceeds one’s capacity to process that choice. Spotify itself claims that “it’d take you more than 80 years of non-stop listening just to get through [our library]”, which means that at a more reasonable rate of two-and-a-half hours of daily listening, your chances of sampling even the greater part of their library are pretty much zero. However, you’re never going to be short of new material to listen to and that suggests that the “moral fatigue” of tracks on Spotify might be faster than in the case of privately-owned libraries (“Ooh, now I want to listen to this and now this, oh and that one was pretty cool too…”)

    In summary, I think that tracks that get played 250 times by a single user over the lifetime of their subscription will be a tiny minority. One of the reasons to pay a subscription fee is to have access to whatever you want, whenever you want it. Music fans who want to play a single track over and over would be better off buying the download.

    Having said that, based on my most recent statements from Spotify, the payout per play has recently risen to a much more respectable 0.85 cents or so, setting the stream-to-download ratio at around 80:1. Even so, I don’t think that the majority of songs on Spotify will generate that many plays per average listener.



    1. Thanks Faza,

      I reckon your points about the listening context–“scarce” albums versus the scarceless service of 15 million tracks–is important.

      Essentially, while broad generalizations are being made about music fans today versus those yesterday, it’s worth considering whether regardless of generation whether certain musical products/services match best to music demand.

      Pete Peterson, a sociologist, made the comparison years ago between a sort of monoculture and cultural omnivores (those with a broad range of genre tastes).



  9. Hi. This was interesting, but unfortunately I feel the math is all wrong.

    As themusicaldisconnect pointed out above: Projekt receives .0013 per stream at Spotify. You state the rate is .0033. Now maybe that’s the back-door deal with the majors. It is NOT the deal for the indies.

    Furthermore, I also doubt that a purchased songs gets 250 plays over a lifetime. Some music gets lots of plays; some gets played once or twice. I read something today that said an average album is played 8 times…… so this really skews your math.

    I would much rather get an upfront payment from iTunes, then the non-transparent way payments are being made for streams.

    This is entertaining, sure. But not accurate.




    1. Sam,

      Thanks for your comments.

      First, let me assure you that my objective here was not to be interesting. In fact, in the absence of enough data to be truly accurate, I was trying to (a) tease out a rough estimate and (b) at least get people thinking about the inputs.

      I will follow up with how I got the $0.0033 rate as I know your own experience has been lower.



  10. here’s my lastFM stats

    27881 plays since Jul 2005

    Most listened-to track in 6+ years clocks in at 148 plays. £0.44 of revenue at streaming payment rates.

    I think it’s safe to say a one-off payment of £0.99 from a digital store would be preferable.



    1. Thanks for sharing these data.

      Over six years, yes, a purchased track would be preferred to a per stream payment. I don’t deny that, in general.

      But the life of the copyright is life of the author plus however many years (recordings are different from works. And these purchases will be yours for listening for multiple decades.

      Your least listened to tracks we’re heard roughy 3.5 times per year, while your most listened to were clocking in at 20+ times per year. The average here may well be in the range of 8-12 listens per year. Apologies, as I don’t have the time to calculate at the moment.

      My sense is that the present value of steaming payments is indeed less than the value of the download purchase.

      And so, of course your prefer to be paid upfront.

      The question becomes, how does the consumer prefers to pay.

      Because in the end, you are trying to sell music in some way or another. And how the consumer prefers to pay will matter, won’t it?



  11. Hi David,

    I have had the “good fortune” to have written a song or two that has done well and I get statements with Spotify income. Here’s my experience.

    The average per stream accounting to me on a song I wrote 100% of is $.00067. That’s 1 more rather important “0” in the equation. Also, I am referring to my European streams where Spotify is much more established.

    Given that rate, it will take me something like 30 million streams to get close to the US poverty level (around $20,000), and that’s 30 million streams a year for the rest of my life. The vast majority of music creators won’t get 30 million streams in their lifetime.

    And yes, the labels do own a piece of Spotify and do keep all the real revenues that ownership entitles them to. Lovely for them and goes a long way to explaining why Spotify exists at all.

    You might argue in fact that unpaid music file sharing (P2P) is better for creators. It pays almost the same (the difference between 0 and a number almost indistinguishable from 0), and it doesn’t pretend to do right by folks like me.



    1. Eddie,

      Its great to hear from you on this. As you know, with a songwriter in the family I am equally concerned about the money on the table in this situation. I am not denying that the mathematics of survival are changing and may even be depressing, I am simply trying to surface (a) how they are changing and (b) at what levels they get better rather than worser.

      And so, if $20,000 is our hurdle income from a single track, what I am digging into is the 222,222 downloads that would have to be sold or the 30,000,000(ish) streams that might have to occur—given the rates at which Spotify is presently paying some people. So I am looking at the ratio, which in this case would be 135 streams to download.

      Let me ask a few questions by way of some math:

      Since you referred to songs you have written (not those upon which you have performed), I am going to assume—as dangerous as that can be—that you are referring to songwriter royalties and not payments for being the owner of the master or the performing artist.

      If I take our statutory US mechanical rate (which you would receive from a CD sale or digital download) of $0.095(ish) and divide it by the effective per stream payments you have received from Spotify of $0.00067 I get a ratio of 134 (streams to download). Unless I either misunderstood and/or did my math wrong.

      If I take the classic “controlled composition” version of the statutory rate, $0.0675 or 3/4 of $0.09 per CD sale or digital download, I would find a ratio rather close to 100 (streams to download).

      Note, that in both cases above I used the most recent(ish) rates for mechanicals and not the rates at which your original works may have been licensed.

      In essence, there is a built in ratio that asserts (wrongly or rightly) that 100-134 listens over the life of the song on a per stream basis may be “equivalent” to the number of listens to a sale.

      We don’t usually think of sales in terms of listens of the lifetime of the CD in our collection. I am simply thinking that way.



      1. Thanks for the reply David, and happy to be in the conversation. I do see your logic and here are some further thoughts.

        The statutory mechanical and the controlled comp rate are unfortunate places to start the comparison (not your fault of course), because it’s a little like starting with what Welsh coal miners were paid in 1830 and building an argument for reasonable compensation from there. In other words, the baseline is subterranean.

        I also wonder if this approach would hold up for any other industry. If you paid for a rake for example based on how many times you raked your backyard over the next 10 years, could rake manufacturers stay in business? If the return is pushed out to some indeterminate future, how do you survive in the meantime? After all, I can only wait so long for my 4.5 cents to accrue.

        Thanks again



      2. Two important subject for follow up:

        (1) What “should” these rates be if our goal is professional vitality in the culture industries?

        (2) What are the financial consequences of this model of micro-payments over time?


  12. Should we provide artists with mere money or direct goods, as would happen in a barter system? Are we all propogating a culture-wide collapse of creative investment through oblique payment schemes? Will the ancillary fallout of this massive and thoughtless consumption bring to bear a desolate landscape of wafer-thin products?



  13. music is not a rent vs own equation. the reason we buy music is to express our identity by collecting records that represent who we are and how we feel and how we want to represent ourselves to others. we have never really purchased music just for utility value. that’s the product that musicians have sold for decades : a signifier of lifestyle, an object that holds emotion and memories.



  14. Assuming 250 is correct, the satisfaction (utility) you get by listening 250 different songs compared to 250 times the same song is much higher. So you are expected to be willing to pay more.



  15. There may be another way to work out average plays for each item of owned media. Without this metric, we may struggle to find an objective comparator with streaming rates per listen (play).

    The average number of plays per product (owned) could be calculated as follows:

    (Tracks Listened to Per Year * Lifetime of Recorded Media in Collection) / Total Number of ACTIVE Tracks in Personal Music Catalogue

    Catalogue size is estimated elsewhere to be 7160 (for iTunes at least, but no harm in assuming this is the same as physical – about 500 albums)
    Not all of this is listened to. Some estimates out there state 26% is unplayed on average. That gives us an active catalogue size of 5,298

    We get tracks listened to per year from an estimate of minutes per day / average length of track
    Minutes per day ranges from 33 to 72 minutes from various online surveys – average is about 50 minutes
    Minutes per track is around 3:38 from what I’ve seen – or 3.63 in decimal
    Tracks per year = Average Minutes per Day/ Average length * 365
    This equates to 50 / 3.63 * 365 = 3,650 per year (13.77 per day)

    The next variable is “lifetime of product” – Steve Jobs is quoted as saying technology cycles every ten years. Let’s assume someone ports media from one platform to another (CD ripped to MP3 being the example) and thus the lifetime is 20 years – before the product is lost, discarded, replaced or archived forever.

    So, in 20 years, on average, we may get to play 20 * 3650 tracks = 73,000

    With the assumption that the above numbers are OK, we achieve this:

    (3650 * 20) / 5298
    73,000 / 5298 = 14 plays per year on average for 20 years = 280 plays in total lifetime for owned media

    Not far off the 250….

    It’s accepted that all of the variables above are up for challenge, but it’s another way of thinking about what makes an average number of plays in an average lifetime of play.

    We cannot use the life of copyright as the lifetime (70 years in UK), as we cannot reasonably expect any medium to be still functional and useable in that time period. Think DAT, cassette, eight track etc. At some point, we cannot use the medium anymore, unless it’s ported to a new one.

    Also, the actual listening habit is likely to be happening in the first year or two – from purchase data it has been seen that 99% of all purchases over a ten year period occur in the first year. We do have a small problem here – streaming is available 24/7 – media does not stay in shops indefinitely, so the supply is cut off at some point for purchasers (unless it can be found online or in specialist outlets). This may skew figures because I can stream much more easily than I can purchase, certainly after 10 years for example. This helps to explain why so much of the purchasing occurs in first year.



  16. The big part that people are missing in the streaming versus purchased debate is that with purchased music, you have a limited library of music. Perhaps it isn’t high to suggest that a song will be played 250 times over 25 years in a limited library (though I seriously doubt that number is typically achieved as listeners tend to listen quite a bit at first and then it drops off significantly over the years as new music replaces those songs – I would think that in the last 15 years of having that song it would only be played a handful of times), but in the streaming world where you have ALL of the music in the WORLD ever created at your fingertips, you’ll almost NEVER reach that play count. Streaming encourages music discovery and also music ADD.



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s