How a failed economic theory still rules the digital music marketplace


Unless you spent a lot of time listening to early ’00s techno-utopian babble, the Theory of the Long Tail probably means nothing to you. Yet if you live in the US or Europe and you run a digital music label, you’re living it – or the fallout from it – almost every day.

In 2004, Wired magazine editor Chris Anderson proposed The Long Tail, an economic theory blown up by futurist steroids. It theorized that with the introduction of the internet, blockbusters would matter less and everyone would sell “less of more.” The Long Tail prophesied “How Endless Choice Is Creating Unlimited Demand,” according to the subtitle of Anderson’s later book, which if true would turn the field of economics on its head.

For a practical example of what this all means, compare a brick-and-mortar record store like the old Tower Records vs. an online retailer like Traxsource. Your local Tower Records had to limit its inventory to take into account a finite shelf space. Their stock might have consisted of a couple hundred records. And each record didn’t get equal shelf space: your hippie boomer parents were going to buy more copies of Beatles records than all your Belgian techno records, so the store would stock and give more attention to the former. This “artificial” scarcity of physical products taking up physical space and depriving it from other products had bent consumer behavior out of shape for basically all of history.

. . . .

With the internet and the creation of intangible digital products, this was supposed to change. Traxsource and other digital retailers are limited not by shelf space but by the size of their server hard drive array. And buying more server space is cheaper than building a new store.

According to Anderson, sales would in the future would represent a classic “Pareto” or “power law” demand curve: 20% of sales would be by “star” artists selling millions of copies each in our record store analogy, while 80% would consist of many thousands, tens of thousands or even millions of artists selling relatively few copies of each of their albums as the store’s near-infinite inventory meant people could metaphorically “wander about” and choose from millions of options.

This was the “Long Tail” in a nutshell, represented on a chart stretching to the right into infinity: in the future, music retailers would sell “less” copies from “more” artists. Many more.

. . . .

As early as 2008 – five years after iTunes was founded and we began to get actual data of how this whole thing was working – keen observers began chopping the Long Tail down to size. Economist Will Page working with Andrew Bud and Gary Eggleton was able to obtain somewhat anonymized transactions from a “large digital music provider” rumored to be either Rhapsody or iTunes itself. They had so much data, in fact, that an ordinary Excel spreadsheet choked on it.

It was a gigantic sample of… nothing.

80% of the songs had no transaction data: they had sold no copies at all.

There wasn’t any volume in the “Long Tail” and nothing had really changed – except for the worst. The actual sales data showed an even greater concentration of sales in the “Fat Head.” Page later spoke about their findings:

“We found that only 20% of tracks in our sample were ‘active,’ that is to say they sold at least one copy, and hence, 80% of the tracks sold nothing at all. Moreover, approximately 80% of sales revenue came from around 3% of the active tracks. Factor in the dormant tail and you’re looking at an ’80/0.38% rule’ for all the inventory on the digital shelf.

“Finally, only 40 tracks sold more than 100,000 copies, accounting for 8% of the business. Think about that – back in the physical world, forty tracks could be just 4 albums, or the top slice of the best-selling ‘Now That’s What I Call Music, Volume 70’ which bundles up 43 ‘hits’ into one perennially popular customer offering!”

When the new owners of Rolling Stone recently announced they would challenge Billboard’s dominance of the pop charts, what was left unsaid is how pointless a “top 100” of ANYTHING has become. As far as big-time music industry relevance, a “top 100” could probably be cut down to a “top 8” or “top 11.” Sales are so heavily concentrated at the top that you’d expect artists to start their own campaign for industry income equality.

Link to the rest at

PG suggests the same thing may happen on Amazon with books.

But he could be wrong.

For one thing, it is possible that the market for a wide array of books is different than the market for popular music.

Or not.

PG will be interested to hear the opinions of one and al on this topic.

10 thoughts on “How a failed economic theory still rules the digital music marketplace”

  1. It’s buried in the math. So-called “long tail” distributions only “work” when sales increments are not rigidly quantized, or the incremental quanta are not significant fractions of each other.

    Here’s an example: Consider the effect on an indie author when monthly sales rise from “2 units” to “3 units” at the same price. That’s a 50% increment, but it’s also the lowest possible increment. And, of course, it goes the other direction, too…

    But when the quantum remains a single unit and baseline is 200, a single-quanta change is not a significant fraction of the baseline. That means both that “noise” doesn’t mask trends, and that a 50% sales bump is itself meaningful and not part of the noise.

    This is basic signal-processing theory and basic statistical mechanics. It’s stuff I learned waaaaaaaaay too many years ago as an undergrad in math-based science classes and laboratories. It’s obviously outside the experience of Mr Anderson (no, not the one from The Matrix) and Wired… and the irony of that leaves too much to be desired.

  2. I can’t argue about the math, but it’s possible that living authors have more ways of promoting themselves and their books than musicians. We certainly see more writers earning more money through indie publishing than trad-pub writers.

    Meanwhile, a lot of the musicians at the top are denizens of the traditional music industry, which dominates the media. The rest are scrabbling for bits, and it’s difficult for their fans (and potential fans) to find them.

  3. My suspicion is the data would look different from a free or “buffet” style subscription vs. having to pay money to listen to a particular song. You’d have to be VERY sure you wanted to listen, so of course very top popular songs would be chosen. If it is free or with no *added* cost to try, you might see more curiosity-driven listening. I extrapolate from my own behavior, rummaging around the “free” (with Prime) music Amazon shares to find stuff I like, which I only buy when that music is taken out of the free rotation. I also scour Bandcamp and make full use of the free samples.

    Isn’t it curious we are seeing an article that tells the Big Publishers what they are doing is Completely Reasonable and there is no reason to worry about midlist authors, only blockbusters? 😀

  4. I dunno, but I thought “long tail” was about total lifecycle sales for specific titles, not about market sales distribution at a specific point in time.

  5. It looks like the “long tail” data is being buried in the “tsunami of crap” data. Not everything is going to develop a long tail: and some works find their popularity generations later. I’m sure HP Lovecraft has more readers now than he ever did in his lifetime.

  6. Does the long-tail theory fail everywhere, though? It seems to me that it might make more sense applies to widgets, where the subjective factor of it (call it “perceived or actual quality” or “appeal”) is more uniformly distributed and price comes more into play. But with music and books, that quality/appeal is not nearly so evenly distributed. If I’m a rock fan, for instance, Springsteen, Dylan, The Stones, The Beatles, etc., carry a lot more weight with me than Shecky’s Roadhouse Accordion Players. Same with Stephen King, the Harry Potter books, etc. There’s something about that stuff that reels more readers in, so it’s tougher to quantify across a wide range of offerings.

    • Actually, those books *are* benefiting from the long tail: they are perennials, eternally in print so the continue to accrue sales long after the launch window and initial print run. In fact, the Potter books are perfect examples of the long tail, as the bulk of their installed base has come long after launch and long after the series was “finished”.

      The long tail is about the spread of sales across time.
      It is counter to the “fresh produce” approach of the big publishers that limits availability of a (midlist) title to just part of the first print run and pulps most of what doesn’t sell during the launch window.

      So, yes, under that regime, the long tail doesn’t apply.
      Because they make sure it doesn’t. They really don’t want Jane midlister’s five year old book “cluttering” warehouses to limit sales of Mary Bigname’s new release.

      That is one of the “problems” ebooks present for the BPHs, especially in fields like SF&F where the deep backlist is heavilly perennial. You can be sure that the bulk of sales for Heinlein, Asimov, Niven, etc came from reissues decade after decade, not from the 5-10,000 copy print runs.

      They’re misapplying long tail for reasons of their own.
      I think we can guess at a few reasons why they would want to discourage authors from thinking of their books lifecycle sales.

Comments are closed.