I think prof. Elberse's analysis is really interesting and insightful, but I don't think it's conclusive proof that the long tail theory is dead.

Clearly, more specialist customers are likely to buy from the long tail. But this is the kind of niche customer that you want to attract to the site, anyway, since he/she has no other place to satisfy his/her interests - and hence is likely to remain more loyal.

Furthermore, I believe there is also a [yet unquantified] network effect involved: people with an in-depth interest may come to the site, repeatedly, precisely because there is so much potential choice. This does not mean specialist customers always tend to choose obscure titles (they often pick popular titles, as most customers). But if they feel there is less choice, they can also easily just go elsewhere (i.e. the shop "round the corner").

These people drive much of the sales, they are most likely to explore new titles and they will also spark and drive the "network effects" of popularizing new or obscure titles. In the long term, the people that [also] buy from the "long tail" may be precisely the customers a seller is most interested to keep.

I found Prof. Elberse's article misleading, and I strongly disagree with the conclusions she implies.

Her own research found that 68% of Rhapsody's sales came from titles not found in a Wal-mart store. This data *supports* the Long Tail idea! That a Harvard professor concluded otherwise is an embarrassment.

Mr. Anderson's response is too kind, citing differences in how one defines "head" and "tail." He's covering up a fundamental problem with Elberse's methodology: by using percentages (or deciles) of units to measure "concentration," Elberse essentially created a self-fulfilling prophecy.

Consider what happens if Rhapsody's catalog grows from 1 million to 10 million. Suddenly, the top 10,000 titles that generated 32% of sales represent not 1% but 0.1% of all titles. The top 100,000 titles that generated 78% of sales now represent 1% of titles. Suppose that total sales from this expanded catalog are double the sales from the original catalog. Elberse's methods would find that the top 1% now represents 78/2, or 39%, of all sales. Elberse would call this an increase in "concentration" -- implying that the catalog expansion was unprofitable and that the Long Tail theory has been debunked.

The whole point of Long Tail theory is that adding 9 million obscure titles to a digital catalog has negligible marginal costs, so *any* incremental sales would be profitable. Elberse's ill-conceived analysis completely missed this point.

Rhapsody, by taking the Wal-mart catalog and expanding it 100x, roughly tripled sales. If Rhapsody were storing this catalog in a physical warehouse, that might seem like a bad result. But the whole point of the Long Tail theory is that when you aren't held back by physical costs, a 3x growth in sales from a 100x growth in catalog can be profitable. But from Elberse's perverse view, any lengthening of the tail, almost by definition, would be seen as "concentrating" sales in the top 1%, unless it also delivers an arbitrarily high bar of incremental revenue.

Her statement, "customers give lower ratings to obscure titles," is similarly flawed. Elberse has taken large groupings of obscure products, averaged the customer ratings across the entire group, and found that products at the obscure end of the tail get lower ratings (on average). She implies that the "gems" found in the long tail don't compensate for the sea of mediocrity that they're surrounded by. Again, she seems to be holding the Long Tail theory to an arbitrarily high bar. So what if there are 100,000 mediocre titles in the lowest decile of a 1-million-song catalog? Given the costs of digital media, almost *any* incremental sales from those titles is incremental profit.

(Elberse also neglects to recognize that digital media helps quality rise to the top more quickly. On the internet, the best-rated "gems" can rapidly get more exposure, taking them out of obscurity and out of the tail. This is one of the great benefits of investing in the long tail! But a side-effect of this benefit is that as highly-rated items get elevated towards the head, then what remains in the tail will necessarily be lower-rated. Elberse would have us conclude that the tail is therefore less deserving of investment.)

There was one aspect of Anderson's original theory that I always disagreed with: the idea that a certain type of consumers have a preference for obscure items over popular items. I was glad to see Prof. Elberse's statement to the contrary: "there is no segment with a particular taste for the obscure; rather, customers with a large capacity for content venture into the tail." This reflects what I intuitively thought, and I'm glad to see analysis backing it up.

Anderson's original Long Tail theory deserved the popularity it gained, but unfortunately with that popularity came a level of hype -- people applying the concept with excessive fervor and expecting more from the long tail than it could deliver. (If you double your catalog expecting to see your sales double, then you are bound to be disappointed.) Elberse's findings may debunk some of that hype; but they do little to debunk the original theory.

Some quick thoughts on Anita Elberse's article.

When looking at consumer research, what I find interesting are non-intuitive, actionable insights. Elberse's discovery that, as she modestly states, "may be of intellectual interest to readers," that there's money to be made in blockbusters is neither. A record executive that states "if we just had more platinum records, we'd make more money," would not be applauded for strategic vision.

I found the tone of the piece somewhat arch as if it was penned to provide a necessary tonic to the success of Anderson's book. There's a hint of straw man building when she suggests "...that it would be imprudent for companies to upend traditional practice and focus on the demand for obscure products." As if, a Blockbuster executive might put forward the shifting the product off the new release wall and replacing it with the French New Wave.

To look at the Rhapsody data and to come away with the insight that really popular music is still popular, even at Rhapsody, is not that interesting. That, as Anderson puts in his rather polite response, Rhapsody gets 68% of its demand from titles that are not available in traditional retail stores, is the type of insight that really could upend the entertainment media value chain.

Even limited just to the arena of producing more blockbusters, if companies can monetize inventory they had previously written off and find more ways make profits on titles that fail to appeal to a larger audience, they can afford to make more bets and increase their odds of having more blockbusters in their portfolio.

Her longitudinal data from Nielsen VideoScan on topsellers seems irrelevant as a counterpunch to The Long Tail.

The notion that entertainment co's might be able rethink their portfolio strategy seems much more useful than the article's advice that "When producing niche goods for the tail end of the distribution, keep costs as low as possible."

One of the key tenets of the Long Tail was that new modes of distribution were causing a regression towards zero in inventory carrying and transactional costs. This is why a company like Netflix (btw: NFLX trades at 3x BBI) can have an inventory position vastly superior to a regular video store. Elberse's advice that retailers need to "Strictly manage the costs of offering products that will rarely sell," especially given to retailers who traditionally have operating margins close to 3% and are already very cost focused just doesn't seem that useful.

In a final dig, noting the delicious irony that Anderson's ideas were contained in a bestseller, she again (just as her Rhapsody and Quickflix data seem to reinforce Anderson's thesis) states the obvious and inadvertently makes Anderson's point. Publishers were willing to make a surprisingly large bet on a book with a few too many mentions of Power Laws to generally attract a large audience. In making a large bet, Hyperion signaled their awareness that bestsellers can make money and in Anderson's case, maybe they took to heart that with new potential audiences for niche content, they had a hedge that allowed them to make the "jaw dropping" offer and subsequently win the bidding war and make a decent return.

Bill

My rebuff of Gomes' reprove of the Long Tail

In an article by Lee Gomes of The Wall Street Journal out July 2, Gomes uses a recent article (online at HBR) about a study that suggests "The Long Tail" theory from a 2006 book of the same name is misguided at best -- or incorrect at worst.

Reading this idea inspired me to respond with this published rant:
Response to Lee Gomes (Lee.Gomes@wsj.com) of the WSJ

With reference to your July 2 Portals article, "Study Refutes..." you obviously do not dwell in the "Long Tail".

As for the article you use for your argument, if one studies the mainstream, one sees the mainstream results. Maybe the bulk of people have yet to learn (or live) the lesson of "The Long Tail" because mainstream industries, that greatly profit from the super-hits, spend a lot time, effort and money to ensure people continue with their buying and shopping habits.

Corporations -- especially multinationals -- aren't known for embracing change that they did not introduce or profit from. Overall [Lee Gomes] your article reeks of protesting too much. Want to let me in on why you might do this...?

I say this is a miss: "Do we want infinite choice, or do we prefer to pick up on the likes and dislikes of others in forming our own tastes?" It is the wrong question all together. The "we" in it is undefinable and any conclusions draw from this are misleading generalizations.

The Harvard professor's suggestion -- as summarized in the sidebar, counterpoint heading -- that the Web is "cementing the prominence of a small number of cultural favorites" is false for if no other reason than the Web is a tool. It is the acted upon, not the actor.

Thus, the actor cementing the prominence of the super-hits is the people who profit from their continuation. Again, I suggest your argument is biased because you [Lee Gomes] work for the WSJ and are a part of the industry cranking out the mega attractions.

As the Long Tail grows (as time passes), the amount of choice within the Long Tail grows exponentially faster than the number of people looking to make the choice. Prof. Elberse correctly points out, early in the article, that people will first - and increasingly - go to the 'best' choice. While there is some theoretical truth in the Long Tail concept, any particular component of The Long Tail becomes increasingly valueLESS with time.

It seems as though the numbers of digital downloads cited for the U.S. in 2007 is too low.

According to the International Federation of the Phonographic Industry (IFPI) 800 million tracks were downloaded worldwide in 2006 alone. That number is also too low since iTunes sold about 1.1 billion tracks in 2006 and 2 billion more over the course of the last 12 months (7/31/07 - 06/19/08).

Finally a rigorous study to burst the 'Long Tail' bubble. I welcome Anita Elberse's analysis. It is right on the mark!

Professor Elberse's article is an interesting one but I believe her logic is fundamentally flawed when it comes to "profiting from the long tail". She states:

"my findings suggest it would be imprudent for companies to upend traditional practice and focus the demand for obscure products. The data shows how difficult it is to profit from the tail."

In my opinion her approach is monopolistic in nature. Yes, it could be difficult for one player to extract ALL profit from the long tail assuming that tail becomes longer and longer. Despite something close to marginal cost of zero, marginal revenue of zero still equates to zero profit.

However, for me the long tail is not about one player trying to extract ALL the value from the tail. Instead it is about multiple players who can contribute/extract value to/from different points in the ever lengthening tail.

Furthermore I disagree with Professor Elberse's implication that "there is no segment with a particular taste for the obscure".

Despite her interesting research supporting the double jeopardy theory, just because a "connoisseur" may rate an obscure title lower than a hit, one cannot conclude that person doesn't receive sufficient value from the obscure title to make it worth his while.

The author in her belief that the long tail will remain extremely flat makes what I think are two false assumptions. The first is that search and filtering technology will remain as they are today. I believe these technologies will improve and that as they do, people will be able to more easily wade into the tail. The second is that the knowledge and skill levels of the consumer will remain as they are today. As time goes on, the Internet generation and those behind it will make up an increasingly larger share of the consumer base. As a result, the consumer base will be one that is more aware of the long tail and more adept in using the resources available to tap into it.

As the person who introduced Anita Elberse to Chris Anderson (and vice versa), I feel a duty to weigh in on the debate.

Anita, as she points out in her rebuttal of Chris Anderson's rebuttal, does mention in her article that the data indicates a large portion of total sales are now coming from titles too obscure for physical brick and mortar retailers to carry. However, this point is made almost parenthetically to her larger conclusion that the long tail isn't all that relevant - based on percentile data which, as many here have pointed out on, doesn't offer a fair comparison if the catalog sizes are not equivalent.

For me, the strongest point Anita makes is in the Neilson data indicating that 91% of the transacted digital music catalog is downloaded fewer than 100 times. While in aggregate, these 'long tail' titles may represent a large part of overall sales, on the individual product level the media business model makes such low volume unprofitable for producers. How can the investment (fixed costs) in production and marketing of one of these products ever be covered by such low volumes? Yes, the variable per unit cost is low - especially for digital - and this makes the tail profitable for large catalog retailers. How producers should adjust their strategy in an era of decreasing sales of hits (in absolute terms) and low volume in the tail remains a difficult question. Clearly those already in possession of large catalogs stand to profit - if they can surmount the legal hurdles to distributing these digitally.

Finally, a point often overlooked is the changing shape of physical retail - where a large portion of music sales still occur. For music, the large deep catalog music retailers, Tower, Virgin, etc, have been disappearing and the brick and mortar business has largely transferred to mass-merchants with very shallow catalogs - ie Walmart and Target. This is creating an effect inverse to the long tail in the physical market. The result is a greater concentration of hits (The head) being sold by brick and mortar. Less obvious but also important is that brick and mortar stores play an important role in music discovery and promotion - so less retail space, fewer stores and smaller catalogs mean that it has become much more difficult for more niche/obscure/independent titles to gain visibility in this channel. The blockage of non-mainstream titles from this important not only sales, but marketing arena spills over into the online and digital channels - hurting sales of the non-mainstream there as well. What's clear despite all this is that the shape of the sales curve is changing and that the long tale effect is real. Again, what is less clear is how producers of new products - either mainstream or niche - can adapt and benefit.

perhaps the data is not always the gospel. if you recall the data prior to medicine discovering germ theory, they locked up and revoked the doctors license who suggested germs existed. Meanwhile, going off 'the data' the entire medical profession provided a 50 year platform for the death of thousands of innocent people because "they had the data.'

i can personally attest to the fact the longtail has provided tens of millions of dollars in site revenue for an ecommerce client and will continue perusing the longtail regardless of your findings related to 'data.'

for those who follow the advice of this Harvard Report, good luck.

I felt that Prof.Elberse's article serves as a supplement as for how to apply "Long Tail" to profit making. Long tail is not for every online bussines and it is being over-used. How to transfer the concept into business model is a really difficult question.

I belive Rhapsody's business success reflects Long Tail rather then undermine it. The comparison with Wal-Mart explained its "Long tail" essence.

I have not read the original "long tail" book or article by Chris but through numerous indirect references to it in search engine optimisation contexts I have come to embrace the thought. I don't feel the need to read the book. I have seen it working and I see it in Prof Elberse's article. Hence I read her article twice and still missed the point.

I agree with Ali Partovi where he says "The whole point of Long Tail theory is that adding 9 million obscure titles to a digital catalog has negligible marginal costs, so *any* incremental sales would be profitable. Elberse's ill-conceived analysis completely missed this point." (although I would not say "ill-conceived").

We don't need to worry about the head - the accountants make sure that the business focuses on the 20% that brings in 80% of the revenue. As an SEO, I look for ways to ensure that long tail items in the inventory are as visible as the rest.

Conversation Starters
Debating the Long Tail
Chris
I agree with you totally. We have come to some stage that we see the glossy cover faster and do not even bother to open the book. We have created a saying, “Don’t judge the book by its cover”, but we are the very ones to buy the books hastily only to find the contents vary and deep reading tells us that the material is not in the format we want. There are many shortfalls and many copied. Many times when we open yet another book, it had the same materials.
I think this is to do with authors’ raw thinking and look for quick fix by copying and rewording these.
I thank you
Firozali A. Mulla MBA PhD
P.O.Box 6044
Dar-Es-Salaam
Tanzania
East Africa

The problem is not with either analysis of the data, it is with how Ms. Elberse's article has been promulgated in the press. She herself does not proclaim the long tail theory is dead or even problematic, she does argue convincingly that its application has varying degree's of appropriateness depending on the sector, business model and customer base.

In reply to Benjamin de Menil (whom I thank for stepping into this conversation, especially given your critical role in the genesis of it) and your question, "How can the investment (fixed costs) in production and marketing of one of these products ever be covered by such low volumes?"

They can't, which I know is your point. "They" being those with large fixed costs investments, especially on the physical production side, whether the product is manufactured in-house (rare) or sub-contracted to large-volume production houses (the norm). (It is, in fact, the inherently "leaky" nature of the latter model which has allowed so many "official" albums to leak into cyberspace in the weeks leading up to a formal release.) Those high-investment entities, of course, are the major record labels. Which in turn is why the bulk of the product released by the majors over the last ten years has increasingly been "safe bet" formulaic content which is more likely to generate large volumes.

For the emergence of more - call it what you will: non-formulaic, obscure, artistically ground-breaking - product that naturally tends toward long-tail-ish volumes, it will have to come from new methods of production that are lower in fixed costs (e.g. born digital and stays digital from recording to package) or have the fixed costs spread across a larger pool of artists. This is precisely what is driving the new (and label-sick) artists to define/adopt new means of production/distribution/marketing and define label-less business models.

Marketing (really, being "found" by the consumers) is particularly throny for emerging artists that lack an established "brand" necessary for sales out-of-the-gate that are sufficient to sustain their work, so aside from a corporate patronage model that covers their fixed costs (what many in the audience decry as "selling out" to brand sponsors), one solution may be to pool resources to cover their costs. Which begs the question: how might a pool of similarly-situated emerging artists boot-strap their "startup" (since $0+$0...=$0)?

Sure, online music communities provide a fourm for an emerging artist to "market their b(r)and" but precisely because it is so low-cost is why the amount of content there has exploded, making it so hard to find the "signal" you want among the "noise." As Gary points out (above, immediately preceding Benjamin), better search & filtering tools (and consumers' skill at applying them) are not static and evolving rapidly, and the minimization of this bottleneck will serve to un-flatten that tail and make this a more viable tool.

In the near-term, and (I hope) continuing in the long run, the more established artists should consider "paying forward" their success by serving as patrons themselves to new artists, a model which seems to be succeeding somewhat in the hip-hop genre. It's a natural creative outlet for the "elders" who tire of touring and recording themselves. Especially of performing live, which will increasingly become the means by which artists make the bulk of their income, ironically bringing the music industry back to where it started in the days before the technology which allowed for mass-market distribution of recordings. The established artists have useful "been there, done that" wisdom, and would be a patron more likely to preserve artistic freedom/integrity than a corporate underwriter.

An exciting side benefit for artists pooling their resources will be the natural "cross-pollination" of their ideas as they work in such an arrangement, whether physical or virtual. Think of all the many instances of the germination of bands and musical styles in the days of yore when artists shared studio time.

Very enlightening article, even though the results are not surprising.

I think that the current structure of the digital content market makes the 'long tail' theory irrelevant. The reason is the cost structure of these products – a large portion of the cost is fixed. While in some kinds of content the fixed costs are actually caused during the production process (movies, television programs, video games), in most of them the main fixed cost is the marketing expenditures. The consumption of digital content is mostly driven by a former exposure. Therefore the blockbusters are predetermined by the producers who are ready to spend a considerable amount of money on publicity. Not every piece of content that is intended to become a hit will actually achieve this goal, however most hits were intended to be such from the beginning. Along with the low variable costs, this is a classic example of a natural oligopoly. As long as the fixed costs remain high, the market will remain dominated by only a few products. Even the Internet revolution can't make a change, simply because it only lowers the variable costs keeping the fixed cost high enough to maintain the oligopoly market. The fact that the content is more accessible now has a minor effect because people keep consuming the content that they were previously exposed to and the exposure costs remain high. Even those who look for an obscure content won't be easily detached form the mainstream (as the data shows) and the reason is quality.

While most of the income in this market is concentrated within a handful of producers of popular content (leading record companies, movies studios, TV networks, publications), a relatively small portion of the income is distributed among large number of small producers of the obscure content. Consequently there are much more resources available in production of each 'main-stream' product relatively to the 'niche' products; as a result there is a quality gap (I'm referring to the objective aspect of quality) between those two groups of content. The fact that the consumers have rated popular content higher than the obscure content is evidence supporting this notion. Eventually, those are substitute products, and even though one of them is better suited to the subjective preferences of some of the consumers, the other is just of better quality. Not being competitive enough may partially explain the failure of the 'long-tail' products to capture a significant share of the market.

Interesting analysis, but I think Anita's recommendations are off the mark: while the analysis has the benefit of hindsight as to which items became hits and which ones stayed obscure, this understanding can not be developed a-priori in order to be used in any action plan. I have blogged about this on my blog.

For example, Anita segments the 3.9 million tracks from 2007 into the 3.6 million which sold 100 copies or less, and the other 300,000. Unfortunately, this segmentation was not available to anyone at the beginning of 2007, and if you'd tried to guess at which tracks would end up in the winning 300,000, you'd have missed a lot of the blockbusters.

My guess is that the amount of money made by "surprise" winners (maybe a few thousand tracks) is much more than the amount of money lost (if any) on the long tail of titles (the 3.6 million) that according to Anita add very little profitability. By following the blockbuster strategy where you're betting in advance on some likely winners, you'd miss out on these surprise winners.

Anita also tries to segment consumer preferences for the obscure versus the blockbuster, e.g., "customers give lower ratings for obscure titles". NO, NO, NO!! Instead, these titles stayed obscure because users gave them low ratings. The "obscure titles" that got good recommendations went on to become hits, and hence moved out of Anita's sample set.

That's why, while I like Anita's analysis (of past data), I don't agree with her user segmentation and recommendations (for future planning).