Objections

    A test of any theory is how many impassioned objections it generates. By that measure, the Long Tail could be doing better, since nobody's really ripped it or me to shreds yet (bring it on!). Indeed, all the best criticism has come in the comments here, where all sorts of sloppy thinking and phrasing on my part have been caught and corrected with remarkable wisdom, insight and civility. Thanks, but where's the fun in that?

    Today, however, brought some hope with one of the more straightforward attempts to refute the theory. It comes from Shelly Palmer, a composer and media consultant (and blogger), Writing in MediaPost (free registration req'd) he calls the Long Tail "a myth", mostly because I've underestimated the costs of making archive material available. Fighting economics with economics is my favorite battle, so let's dive in with an extensive quotation from Palmer's piece:

There is a very real point where marginal cost exceeds marginal gain and it happens very close to the borderline of the top 20 percent of the selections available on the Long Tail. So, in essence, the Big Hump [his phrase, unfortunately based on a statistical misunderstanding, for Pareto's 80-20 rule] is still the appropriate model for investors and content owners to follow.

Take a movie that already exists. Retain counsel (internal or external) to get all of the appropriate clearances for reduction of the work to the final form factor - ones and zeros. Rent a film chain and color correction suite for the day. Make a 4K file of the entire movie. Down-convert it to every usable format for current distribution and then create files in every resolution required from 56k up.

Now the fun starts. Create meta-tags and descriptions that are meaningful to search engines so people can find the work. Add a DRM (digital rights management) wrapper, so you can get paid. Come on, take a guess... how much have we spent on this single title? Call it $15,000 sitting on a server ready to go. Now remember, we don't know who wants to watch this movie, we are just sure that in time, every movie will be watched according to the Zipfian distribution. Add the cost of money to this equation and multiply by every title in your library. Still sound like a good idea?

I didn't think so. You probably want to do an analysis to determine which movies are the best investments: Which will bring the best return in the shortest time frame? In other words, you will apply the Big Hump to the Long Tail to determine if your content is worth encoding at all.

But isn't the Long Tail an accurate way to look at content usage? Yes, it is. It's just not a profitable way to look at a back catalog that was not created with the Long Tail in mind. It is never less expensive than in the original edit suite. But after the fact, there are very real costs associated with encoding and storage. When you calculate those costs and the cost of capital, the value chain (or lack thereof) is clear.

So, as it has been for time immemorial, the Big Hump is the clear winner for modeling return on investment in the information age or any other. No disrespect intended to those who have been spinning the tale of the Long Tail.

    Sigh. Not as bruising as I'd hoped. There are some good points here, but unfortunately they're in the minority, largely because Palmer sees the Long Tail from too narrow a perspective.

    First, it's a trivial point but let's correct the math. A Pareto distribution is not a bell curve, but is instead another long tail curve that looks pretty much like a Zipf or any other powerlaw.

Pareto distribution

Bell curve (aka "normal distribution")

    Second, the Long Tail is about more than just archives and back catalogs. As I mentioned in the previous post, less than 1% of the CDs currently available are sold at Wal-Mart, which also carries less than 3% of the new music released each year. That's not stuff moldering away in some archive; that's music still on the market, even though only a tiny fraction of it is available in traditional retail. iTunes, Amazon, Rhapsody and their ilk didn't have to look very far to add a few orders of magnitude to the average consumer's range of choice.

    And that's just entertainment media. There's also the Long Tail of advertising, physical goods, software, people and all the new independent and peer-production content that's now being made because there are ways for it to reach a market or at least an audience. The notion of an archive isn't even meaningful for most of those. Lesson: the Long Tail isn't just about Hollywood.

    Third, the example Palmer gives of wrapping old movies in DRM and making them available online is indeed a pretty poor Long Tail economic model, largely for the reasons he gives. But that's not the only way to tap the potential of a movie archive. A much better one is to distribute DVDs via a Long Tail retailer like Netflix.

    Take Sony's $5 billion deal to buy MGM last year. Why did they pay so much? According to the video trade press, "Sony is buying MGM chiefly for its rich catalog of classic movies and plans to exploit that catalog however it can." DVD sales are the fastest growing part of the movie business, and lend themselves to any number of Long Tail distribution models. Just ask the guy who built a huge business in $1 DVDs of old movies, somehow avoiding most of the costs Palmer describes. Maybe Palmer doesn't see DVDs as part of Long Tail distribution. If so, he has either misunderstood the concept or has some good reason for discounting DVDs that he unfortunately hasn't described.

    Finally, most of the costs Palmer enumerates don't apply to other media, such as music or books or even most modern TV and other video. All of them have far more friendly Long Tail economics, either because they're already digital or can easily be converted. Perhaps Palmer has a reason for discounting them, too, but again he hasn't given it. If anyone takes a better swing at the theory, I'll be sure to share all the gory details here. Hope springs eternal.

Comments

[aplumb]

Apologies for slipping into math-geek mode. A Rayleigh Distribution may be a better match than a bell curve, as it has a zero-point.

[Tim]

Actually, he also talks about conversion of film to many formats, but if Internet-based distribution is used, formats might not matter so much. Most of us are still watching content on TVs which are at _best_ VGA-quality (640 x 480) . And for those of us looking for "niche" content, just _finding_ the content available is amazing enough!

[Simit Patel]

Another point to consider is that the long tail allows for a single venue to distribute all the content but does NOT require that same venue to bear all the costs of distribution. Example: I can buy millions of books through Amazon, but not all of them are stored or shipped by Amazon; they offer many books through their network of book sellers. Hence Amazon can expand its catalog without the vast majority of associated distribution/inventory costs (which in turn naturally incentivizes them to have as large of an inventory as possible).

[Gary]

I question the $15,000 figure he uses. Economies of scale will come into play when content owners convert their large back-catalogs. Furthermore, in the future, the mechanisms (and associated costs) necessary to prepare new content for the Long Tail will be built-in and streamlined when the content is created. The extra step, necessary now to prepare old content, will disappear.

[David Palmer]

I agree, Palmer (no relation to me) takes a kind of extreme example in terms of costs involved, and uses that as the basis of his argument. As Simit mentions above, Amazon is a great illustration of a business freed from the constraints of shelf space. Much as I love browsing at the local book or record store, if I'm looking for a specific book or CD, I consider myself lucky if I actually find it. On the other hand, I know I can always get it on Amazon. Nothing special has to be done to those products to make them available online, beyond just creating an entry in the database and having access to the product if it's ordered.

[Zon]

Costs will drop ... benefits of the long tail will be seen ... and one day Wal-Mart will have a kiosk where you can download any piece of music ever recorded. Is that a good thing, or a bad thing?

[kevin Kelly]

IP owners don't even have to do the digitazation themselves. The irony of it all, is that the FANS will digitize all material themselves, on their own nickles. All a studio has to do is publish their tech requirments. (And these can evolve to higher standards over time). The fans will do the rest if you give them permission. And they will probably do it as well as the pros, if not the first time, than eventually.

[Matthew Gertner]

The issue of marginal cost is something that I touched on in this blog post immediately after reading the original "Long Tail" article. Palmer sets up a straw man (and promptly knocks him down) by using the example of existing movies to illustrate a supposed weakness in the long tail theory. This is because the new focus on niches is gaining traction as a result not just of digital discovery and distribution of content, but also digital production. A more appropriate example would be a talented film student with a Mac and a mid-range digital video camera, who could doubtless produce a film for a few hundred bucks (and most of that invested in lattes from Starbucks). Would it have the production values of a Hollywood blockbuster? Of course not. Might it have enough inherent value to make money from a specific niche of specialist film enthusiasts? I can't see why not.

[Karmakin]

15,000? To do all that processing? hahahahah. That's funny. There are consumer gadgets that will convert from any format you like into digital. (Look in the back of any computer mag. You'll be shocked what you find). Color Correction HAS ALREADY BEEN DONE. You don't correct again for a digital environemtn, unless you want to. There are GPL re-encoders to take care of that. DRM might be a little expensive, but I'm sure there are scalable plans where you pay a fraction of your sales.

In any case, satisfying the Long Tail can be the result of as little, or much investment as you like.

[Jake]

Whil it probably will get worked out eventually, I think there may be something to the issues raised. Throwing up an mp3 on the web with no DRM or anything is cheap.

But making paid digital downloads work really well is going to be more expensive. When people are getting something for free, they accept little problems like faulty ID3 tags, and stuff like that. When people start paying for it, they are less tolerant of this stuff. They also may want more of it.

Even for a small record label, encoding product with DRM and other relevant metadata is going to cost something. It may be cheaper per unit than making CDs, but I suspect it will turn out to be more expensive than some of the assumptions people are making here. And and over the long run, this may conspire to make the tail shorter than some are assuming.

I guess we'll just have to see.

JL

[chris]

The fact that the Bell curve and the Pareto curve can look very different is a bit of a red herring. A lognormal distribution can be drawn as a bell curve in lognormal space, but give you a pareto curve in normal space. And the two graphs you show can be returned by the same function, with a different input. See http://en.wikipedia.org/wiki/Log-normal_distribution.

[Darrell]

Although the term "Long Tail" has a certain marketing cachet, I question the newness of your epiphany to describe market behavior. You have not unearthed anything new that hasn't already been covered by the basics of Revenue Management (aka Yield Management).

Using the Rhapsody example from the Wired article, it was observed that different price points elicited different purchase rates. Um...yeah....that is what is supposed to happen. There is a recognizable demand curve (the Long Tail). Raising or lowering the price on the demand curve creates a fluctuation in consumption.

A couple of decades ago, a fellow by the name of Bob Crandall used to lead American Airlines. After hanging out in his Operations Research department, he discovered that altering the price to take advantage in fluctuations in demand which yield higher overall revenue. "Revenue Management" is now the reason you paid more (or less) for your ticket than the traveler sitting next to you. Southwest and AirTran also demonstrated that reducing your fixed costs and selling to the masses enables more Long Tail.

What is somewhat interesting is that we have now entered an age where we can see the effects of global consumption of product categories with low costs of production/distribution (music, movies, & books). However...would NetFlix' Long Tail grow longer or shorter if their cost of distribution increase (the USPS is rattling its saber about postage increases)?

Another subtle nuance that differentiates these Long Tail product categories is that they don't have the perishability factor that is found in time-sensitive categories like air tickets, cruises, or hotel rooms (e.g. you forever lose the right to sell the July 15th flight from DFW to ATL once July 15th has passed).

Aside from the small differences in the product categories, there is nothing new that you present that hasn't already been covered by very smart people running around with PhDs in Ops Research and Decision Sciences.

[Tony]

Very cognitive.

[Mark Vane]

Hi, Found a cool news widget for our blogs at www.widgetmate.com. Now I can show the latest news on my blog. Worked like a breeze.

[foreign pharmacy]

Canadian medications offers a convenient, private, affordable alternative to conventional pharmacies and other costly online services. We enable those patients with transportation issues,limited access to quality medical care and privacy needs to have their medications safely delivered directly to their door, without the hassle and time constraint of visiting a doctor's office.