Just because the Internet makes it possible to offer a near-infinite inventory of goods for sale does not mean that consumers will start wanting more obscure items in any great numbers. That is the conclusion Harvard Business School associate professor Anita Elberse comes to in a recent article in the Harvard Business Review that takes on some of the sacred cows of the Long Tail theory.
The Long Tail is Wired editor Chris Anderson’s theory (based on an article and resulting book of the same name) that as it becomes easier to distribute a wider variety of items, consumers will venture down the long tail of the distribution curve and find the products that exactly match their interests and idiosyncratic needs. Elberse questions this notion:
Is most of the business in the long tail being generated by a bunch of iconoclasts determined to march to different drummers? The answer is a definite no.
. . . Although no one disputes the lengthening of the tail (clearly, more obscure products are being made available for purchase every day), the tail is likely to be extremely flat and populated by titles that are mostly a diversion for consumers whose appetite for true blockbusters continues to grow. It is therefore highly disputable that much money can be made in the tail.
Elberse looks at data from Rhapsody, Quickflix (Australia’s version of Netflix), ans Nielsen for songs and movies. Out of one million tracks she studied on Rhapsody, the top one percent accounted for 32 percent of all plays and the top ten percent accounted for 78 percent of all plays. Similarly, the top one percent of videos on Quickflix accounted for 18 percent of rentals and the top ten percent accounted for 48 percent of rentals. Anderson responds that she defines “head’ and “tail” differently than he would. Even so, he adds, that top one percent of Rhapsody songs is still 10,000 songs, more than what you’d find in a typical record store.
What is more interesting about the study is that Elberse cites evidence that, even given more choice, consumers still flock to the blockbuster products that make up the “head” of the distribution curve. This might be because we are all lemmings or, more likely, that taste in music and movies has a social component. We tend to like a song or movie, in part, because other people like them too. Taste doesn’t form in a vacuum. It is socially reinforced.
Even adventurous consumers who venture into the more obscure realms of inventory tend to buy more hit products than long-tail ones. For instance, QuickFlix customers who rented the most movies from the bottom 10 percent of the distribution curve only did so 8 percent of the time. The largest chunk of their consumption (34 percent) came from the top 10 percent of titles just like everyone else. (In the chart below, the red parts of the bars represent the top ten percent of movie titles, and the black parts represent the bottom ten percent. Each bar, in turn, represents a different set of customers and how their rentals are distributed among each decile of popularity). Elberse concludes:
No matter how I slice and dice the customer base, customers give lower ratings to obscure titles. A balanced picture emerges of the impact of online channels on market demand: Hit products remain dominant, even among consumers who venture deep into the tail. Hit products are also liked better than obscure products. It is a myth that obscure books, films, and songs are treasured. What consumers buy in internet channels is much the same as what they have always bought.
So does this disprove the Long Tail theory? Not exactly. (Lee Gomes’ gleeful grave-digging notwithstanding). All it proves is that blockbusters are more durable than we’d like to think, even in an age of limitless inventory and perfect search.
But to say there is no money in the Long Tail is nonsense. It is just more finely distributed and harder to find. True, there are not many businesses that have figured out how to collect it. Google is one with AdSense and search ads. Each search ad is insignificant in and of itself, but all of those obscure terms add up to billions of dollars.
Is this repeatable in other markets? Elberse herself notes that demand is being pushed down the tail. Even if they can gather up that new demand, Long-Tail businesses may not become the most profitable. The economics have changed. And Google is likely the exception rather than the new rule. But neither can that Long-Tail demand be ignored.
In the end, Elberse presents a false dichotomy. The choice is not head or tail. It’s both.
I think there are plenty of other good examples like online movie and music stores. The long tail has tons of potential in all kinds of areas it is just as you said sometimes hard to find before someone else has figured it out. Then that early adopter better makes a good name of themselves in that niche or profit margins decline rapidly.
This post’s discussion seems a bit misplaced. Perhaps I’m off, but I thought the point of the long tail was that there was a latent demand that wasn’t met through traditional retail channels and internet companies (netflix, etc) could capture that value (not *create* value) by offering the long-tail goods. This discussion sounds a bit superfluous then.
It’s interesting to think about this long-tail dichotomy in terms of independent music. If indie music is on the long tail, that doesn’t mean those artists won’t move up the tail - so to speak. With music, I agree that popular music becomes so if not only because other people like it too - as mentioned, taste doesn’t form in a vacuum.
This theory suggests two kinds of consumers and I’m going to relate this to popular vs. indie music. The long tail theory might say that there are mainstream music lovers and then there are indie music lovers but, if I understand it correctly, Anita is saying that the indie music lovers also mostly buy mainstream music but become adventurous in music tastes less as often.
Making it easy to support indie artists is what we’re trying to do at Apricado (http://www.apricado.com) but this is eye-opening when thinking about the kind of folks who might be purchasing indie music.
The important profits from the long tail will largely be made those able to aggregate niche products. The struggling artist will still struggle, until they go mainstream at which point they have by definition escaped the long tail.
fuck the tail man, damn!
It seems that the top selling songs are usually the ones that cost the most money to promote. In order for these songs to be so massively popular, labels usually pour loads of money into marketing and advertising. So yes, some songs are much more popular then others but are they turning enough profit to justify the huge marketing sums?
Which takes us back to the long tail songs that may sell less but did not require expensive videos, tv spots, magazine ads, PR campaigns, etc…
The “Long Tail Theory” itself is nothing more than a regurgitation of Alvin Toffler’s theory of “overchoice”, from “Future Shock” and successive books.
#4 That’s how the game works. LT (yes #5, f the Tail) is simply the next mainstream artist without the marketing $$. I bet you, the next Brit Spears or Kanye is out there in the LT. There are two things that can get LT to compete and sell like mainstream 1) Monies 2) Traction.
Hype the shit out of a LT artist, and I betcha it’s not too hard to make them mainstream. Fun is when you don’t need the big labels or $$ to compete with the current mainstream artists.
How can a few-year-old theory have sacred cows?
Good products will always float to the top.
I don’t see how this is new at all. Just by looking at the power law distribution above you see at the first glimpse that ‘hit products remain dominant’. This is also known as the 80-20 rule or Pareto principle and is far from being just a few years old.
Doug, the post seems to indicate that aggregating smaller niche audiences and directing them to/serving them low demand goods is more difficult and less of a slam dunk than the Long Tail theory would indicate. That even consumers who enjoy the more niche items also partake (with a greater portion of their consumption) in the Head goods.
And, buying into the social/shared aspect of Head goods popularity, the key to success in the Tail is in creating that sense of community of some adequate size n that motivates consumers to purchase and/or spend. Which is Godin’s, Roger’s, et. al. argument about where the music business needs to/is heading.
Erick, thanks for highlighting the article.
I’ve got to call bullshit on this report. Quickflix is a pale imitation of Netflix - and certainly not a very successful one at that. Nielsen Netratings (which is the main source of market data down here) puts them at 61k unique users in May. It also states that the volume isn’t really enough (does not reach the sample size standard), so that’s likely to unaccurate (i.e. probably less than that).
Regardless, the number of rentals actually generated by Quickflix is tiny. Not enough to run this sort of analysis on and get any general conclusions that apply across industries.
This is actually pretty important for readers of this blog to understand and embrace. A lot of Web 2.0 ethos are built on long tail theory, particularly in regards to media consumption. We all flocked to the idea of the long tail because it reflected the buying patterns of a relatively small minority of the consumer populace that wanted to believe it was in fact the silent majority. The rest of the world is buying software from Microsoft, clothes from the Gap, seeing the summer blockbusters, and listening to music from chart topping artists from major labels.
You can make money FROM the long tail but not IN the long tail. That’s it. The rest is execution..no fancy studies needed.
I’m not sure this study is accurate. Considering that both Rhapsody and Netflix are recommendation engines crowd sourced by customers, the inventory wasn’t fairly represented. It seems obvious to me that the content that’s recommended is the content that’s most popular, which starts a snowball effect. I’d be more interested in seeing data from a simple search box like Google where an impressive volume of all searches are “unique”. If you look to hitwise reports, there’s a clear trend that consumers are becoming more particular about what they are searching for and how they are searching.
I think the benefit of the “Long Tail” has been proven, but only if used in the correct context. The big idea behind Long Tail is making already popular goods and services accessible by the masses, not making obscure ones available to small disperse markets…Although, as this post states, that’s still very interesting.
Try to think of Long Tail within market domains. By making things more efficient, convenient, closer, and cheaper, like only the Internet can do, more people can consume and purchase goods and services that were previously not available to them. Take Google Adwords for example. Everybody could buy advertising on the Internet prior, but Google made it much more convenient and lowered the minimum price barrier by removing the expensive sales agent and making it self-serve. Thus, Google accessed the Long Tail in the internet advertising market. Conversely, if I was selling a red number 5 button for my polycom soundpoint ip 550 phone, there might only be 10 people in the world who might buy that, if I’m lucky. Prior to the Internet, I might have only sold 1, but with the Internet, I’ve experienced a 1000% increase in sales. Still, selling 10 is not worth it.
I’m pretty sure most of costumers consume the head of the tail, but there are plenty of people who simply doesn’t consume blockbusters - or consume blockbusters from other eras that are considered the bottom of the tail today, or blockbusters of a very specific genre that’s not in the head of the tail.
It all depends of the tail’s head size and definition.
The statement “It is therefore highly disputable that much money can be made in the tail” is really telling. What is much money? Feeding off the long tail probably won’t fetch too many companies with 5-billion-dollar valuations, but it will more than support someone who works from home and has little overhead. And that is the beauty of the Internet.
Relating my own, actual consumer behavior to this debate, I’m simply buying more altogether. Enabled by the accessibility of the long tail. And maybe that’s the main effect of the web executed well: overall demand increases, regardless whether it’s spent in the head or the tail.
Its called Branding. When there are too many options, people go with what they are familiar with.
As much as we all would like to regard ourselves as “individuals”.. the fact is… advertising works, most people don’t “march to a different drummer”. Unless you live under a rock and don’t watch TV you’re most likely influenced by advertising and therefore are prone to purchase the products that are familiar to you.
If you live under a rock, you probably don’t use the Internet… so the tail doesn’t apply to you too.
As they say, the more things change… the more they stay the same.
different people are in different states of consciousness
the journal writer dealt with lower, chris with higher
no big deal, makes both true
Two-dimensional models often yield false dichotomies.
(I do not treasure her simple definition of “treasured”.)
The long tail can also be seen as an accretion disk.
Incidentally, there are many tails, as with galaxies.
http://www.longtail.com/the_lo.....hbr-p.html
Excellent HBR piece challenging the Long Tail
Agree or disagree - Chris Anderson showed class in his response -
thanks for the book report style summary….Also thanks for the recap of Pareto Analysis (i.e. the 80/20 rule)….this does nothing to disprove that micro transactions are a viable business…the data source is total non-sense and everyone knows that 20 percent of the inventory always makes up 80 percent of the sales or that 20 percent of the clients make up 80 percents of the sales and other Pareto analyses however you want to describe them, they typically prove true statistically….What I learned from this post is that apparently HBS has run out of unique articles and is now regurgitating unrelated nonsense that gets picked up as news
What about fashion products? People hate to wear the same clothes or shoes their peers are wearing…..
in this case, the social reinforcement doesn’t apply
this isn’t what the long tail theory — at least in anderson’s book — is about. it’s not that consumers buy MORE of long tail goods, it’s that proprieters providing digital access to long tail goods can make a living.
People only still buy one obscure ‘object’ a year (or whatever), but, in the old shelf-model, a store can’t survive selling just these type of goods. Only a limited geographic range of people can reach the store, and, there is not enough profit in the one-time sale to justify keeping it on the shelf. The proprietor of a long tail store COULDN’T stay in business by keeping a load of long tail goods; he or she had to add more ‘hits’ or make negligible profits.
The long tail book merely says that since a digital proprietor doesn’t need to a) cover the overhead of a product sitting-on-the-shelf-over-time, and, b) be confined to consumers in a geographic region, the proprietor can carry enough longtail goods to make an excellent profit.
This study’s finding are pretty silly. So, the hits continue to get the majority of the audience? Well, duh. That’s why they are the hits. The concept of the long tail is not that the tail will outsell the head — its that there is untapped potential within the tail that is now accessible because of the internet.
Erick: Thanks for reflecting a lot of the things I was thinking as I was reading the piece (only sometimes I was screaming some of the points in my head — to no avail).
i don’t understand why some people think hits are random (or just a product of huge marketing). everything in the long tail which is great will eventually be a hit.
If you want to analyze all the theory around long tail, you can’t miss Erik Brynjolfsson (et al) papers: http://ebusiness.mit.edu/erik/ (e.g.: “Goodbye Pareto Principle, Hello Long Tail: The Effect of Search Costs on the Concentration of Product Sales” )
Currently the challenge is how to move up from the tail spending less resources (companies: less resources in marketing, consumers: less time for search). In the short term the most promising technologies are collaborative filtering and social networks. See the interesting effect under http://www.problogger.net/arch.....than-digg/
OK, let’s ignore the QuickFlix sample for a moment.
Let’s look at web traffic. Thanks to services like Google Ad Planner and Quantcast, we have even greater transparency into the Long Tail. Is the traffic acceleration in the biggest sites being mirrored by smaller sites?
At the same time, the average search query is growing in both number of terms and complexity. Does it mean that people are getting better at finding niche sites in the Long Tail? Or does it suggest people are getting better at using search tools to sample new experiences and once visited, that they may not necessarily return to a site?
I cover search and how it affects the risk associated with developing new forms of media on my blog. I’d appreciate your comments.
http://connectme.typepad.com
Elberse seems to only back up the Long Tail theory rather than disprove it. The point of the long tail is that it now exists in mainstream distribution whereas it didn’t before the internet. From my interpretation, Anderson doesn’t claim that the tail is highly profitable in comparison to the head. It’s just that there are profits to be had in the tail.
Yeah just to reiterate, thinking back to my first reading of LT about three years ago, this report seems to have missed the point. Tying into what Sean and Zinger have said above, LT theory proves that in aggregate you can make a decent business on the LT content/product/goods. In fact, her data actually proves that - in the Quickflix example (even though this is poor), they’re making 50% of their revenue (rentals) on things outside the top 10%. That’s exactly where the long tail starts… For QF to stop supporting long tail content, not only would they kill half their business, they would likely kill their whole business as selling or distributing both types of goods go hand in hand on the web (otherwise you’d have a Youtube that wasn’t searchable and only showed you the top videos). Ok I think we all get the point.
Anderson’s theory is interesting. Actually long tail (or its formal name , Power Law distribution) is a concept that has been found out to exist in the real world in many domains. Power law shows up in many problems in Physics and Biology, etc… In Economics/Finance, it has been found out that the market prices follow a power law or long tail distribution rather than the previously accepted log-normal distribution (such as Black-Scholes pricing). So, economists have now modified theories of the market with power law derivation so that theory agrees with observations.
The most interestingly of all, is that the way people surf the internet follows a power law distribution, ie, long tail at certain conditions only. This is described here:
The Distribution of PageRank Follows a Power-Law only for Particular Values of the Damping Factor
I think that Silicon Valley is booming with hyped Web 2.0 overvalued tech companies which follow a power law probability distribution, ie, the evaluations of these Web 2.0 start ups is skewed so much to the super-hyping tail.
Hi Erick Schonfeld,
I think my last comment went straight to your spamfolder (2 URLs in the message) or perhaps your awaiting-moderation folder.
Cheers.
Well, maybe the reviewer missed out one detail - what is the ratio of music and movies downloaded illegally and then will need to analyze the preferences of the bittorrent / emule crowd. I think it will definitely show that the long tail works.
“In the end, Elberse presents a false dichotomy. The choice is not head or tail. It’s both.”
Elberse, and frankly most media, demonstrate the same myopic view on this issue as that was applied to the “clicks vs. bricks” arguments from the 1998-2000 period.
Perhaps, it is the nature of journalism, or more fundamentally, a human characteristic that we are looking for 1 answer. The same is going on right now when you hear stories about understanding the current oil crisis or sub-prime crisis.
Charlie Rose, one of the best in journalism, still often suffers from the same oversimplification of analysis. He usually asks questions as if “OK, I know that there are lots of explanations, but what is really causing the issue xyz…?”.
The world is not A/B it is multivariate.
The Long Tail is a powerlaw curve - demand falls off exponentially. Comparing the obscure to the blockbusters results in very small numbers when shown as percent.
A thousandth of a percent of $500 million is $5000 - one is the take for Indiana Jones 4, the other is a decent side income for a personal hobby creating how-to videos online.
Even at a millionth of a percent, $5 in ad revenue is more than it costs to deliver that content online. Aggregate that over billions of videos and you’ve got Youtube. Maybe HBR can look into them next.
Academics are academic.
When looking at the longtail, don’t look at “social media” - films / music.
Look at other items to get the picture.
DUH!
I can’t understand how the statement “that top one percent of Rhapsody songs is still 10,000 songs, more than what you’d find in a typical record store.” can be true. If there are 10 songs per cd and they all are unique (optimistic, I know, but still) you only need 1000 cds, which I think a medium sized shop easily could have…
Danny Robinson’s point (#17) is an interesting one:
“Conversely, if I was selling a red number 5 button for my polycom soundpoint ip 550 phone, there might only be 10 people in the world who might buy that, if I’m lucky. Prior to the Internet, I might have only sold 1, but with the Internet, I’ve experienced a 1000% increase in sales. Still, selling 10 is not worth it.”
This is true if your only product is red number 5 buttons but if you are are selling thousands or even millions of products each year then selling 1 button a year is fine because you’ll be making thousands or millions of single digit sales every year and this will add up to significant income. Amazon and eBay have got particularly smart to this by essentially aggregating sellers and taking a cut along the way. They don’t have to pay for warehousing (which is a hassle if you have to store large, obscure items that might only sell once every few months) or packaging and they don’t even have to provide significant customer support. That, for me is where the long tail really comes into its own. These two can offer pretty much any product I might want even though they don’t actually sell it themselves. Effectively they’ve become product search engines rather than retailers in the traditional sense.
I think that we also forget that sites that sell both blockbusters and obscurities also editorialise and advertise to their users so purchases are not simply determined by individual tastes. This is important because clearly there will be greater ROI for DVD retailers who push the latest blockbuster to all their users rather than an obscure reissue. Quite simply more people have seen (and are therefore more likely to buy) Spiderman than Le Cercle Rouge. This creates a positive feedback loop that pushes these products to top of sales and review charts.
Yay, I knew if I waited long enough I’d see Quickflix featured somehow on Techcrunch
La gets half the story right when he/she says, “…they’re making 50% of their revenue (rentals) on things outside the top 10%. That’s exactly where the long tail starts… For QF to stop supporting long tail content, not only would they kill half their business, they would likely kill their whole business.”
Exactly. The other thing their LT tells them is how many copies to stock, of which products - what’s sometimes referred to “depth” of items. Quickflix (and Netflix) use the aggregated future rental requests of customers to forecast exactly how many copies of which titles they need to have in stock to meet customer demand for rentals.
Just because Quickflix displays a product for rental, that doesn’t mean it owns a copy of that product. It’s putting it out there to test whether it *can* rent that product, and if so, how many copies it should have in stock to satisfy demand.
DVD titles for home-use rental vary greatly in price according to a number of factors, and an online DVD rental business can greatly influence its profitability by buying the right depth of copies, at the right time, to optimise cost of inventory.
Offering both the head and the tail of the market helps establish exactly where the most *profitable* products are for an online DVD rental business. The reason why Blockbuster’s physical stores are a losing business is they are unable to forecast demand and tweak profitability - they have to stock mostly expensive blockbuster titles and then hope they can merchandise them hard enough to make a profit.
nice, but what do the numbers (1-10) on the graph mean & why are there 10 columns?
% is percent of rentals, i guess
- imma (in a confused mood)
Much of the Long Tail is just there because it’s undiscovered country. If marketing and placement were behind some of those obscure long tail pieces and people knew to search for it then those will be at the top.
Pareto’s Principle is that there will ALWAYS be a top that significantly outperforms the rest (and it doesn’t matter the content of the list, check out the Forbes 500 Richest lists). The Long Tail will always exist too. The Internet just allows selections from the Long Tail to move up faster/easier with the right kind of marketing.
The TV show “Friends” used an unknown lead-in song for the show, and the band was put on the map where they sold a lot more songs/CDs that they otherwise would not have. Many times unknown songs put in movies become hits. It’s just that people have a chance to experience the material, not that the material is good/bad on its own.
So go discover the Long Tail.
@Let Me Explain — I like your brevity and precision.
I think the head gets shorter and that’s what really matters now.
I also think that the tail will wag the dog at least in a “countervailing” fashion. The problem with such analyses is that overstretches a small deficiency to make an extreme case.
Elberse doesn’t say that you can’t make money in the Long Tail, but that since the tail is really l o n g a successful long tail business must avoid paying for the merchandise until it is sold. This is easier on the Internet where the cost of storing the product is approaches zero.
The long tail can work for an intermediary who aggregates a lot of different items and pays for them on consignment. It doesn’t work so well for the creators of those products who may only sell a few untits, or none at all.
Google is at first glance, a good counter-example to the long tail is dead argument but its success does not repudiate the anti-long-tail thesis that Anita has put forward. Her point is not to deny demand for the long tail, nor to say that such demand cannot be addressed. Her point is that the marginal revenue from catering to such demand will be on average very slim and therefore only business models engineered with even smaller marginal costs can succeed.
Google validates - not challenges - this thesis:
1) Its marginal cost for each additional page indexed or search query delivered must be vanishingly small while its revenues per click should be larger.
2) The perception that Google covers the long tail / everything under the sun makes users far likelier to use Google for mainstream searches. So not only is Google’s marginal cost for covering the long tail small, actually that cost could be allocated not just to the long tail but also to its mainstream business.
3) Google’s coverage of the long tail seems to have grown inversely with the marginal compute and storage costs of doing so. Google no longer displays the size of its index on its homepage but I recall seeing that number move from a billion to 8 billion over the years when it did publicise that information. In other words, it appears plausible that Google only increased long tail coverage as the marginal cost of doing so became affordable relative to expected marginal revenues.
What this seems to mean for other businesses trying to monetise the long tail is that if the marginal cost vs marginal revenue component of the business model does not work & scale, there is no long term future for the business. Not an encouraging thought but sobering nonetheless.
Her examples are arbitrary. She should have looked at the state of television as there are over 100mm households in the US with a set. Broadcast tv is being replaced by numerous niche cable channels. 100% long tail economics that are actually quite profitable.
Makes you wonder what a Harvard Business School education is worth nowadays.
I have to chime in on this topic having just published an article on the same.
I have not read Ms. Elberse’s paper so I cannot comment on the validity of her findings or conclusions. What we are all really responding to is TechCrunch.com’s own conclusions drawn from her paper (which I assume and hope that they have read).
That being said, Anita Elberse claim is immediately questionable due to her own admission that “She defines head and tail differently than Anderson”. You cannot disprove an argument or experiment using different definitions of the key terms. It is in this violation of the rules of both rhetoric and scientific method that her argument starts on shaky ground - or at least as her argument reads in TechCrunch.com’s article.
The use of relative terms like “much money” also tends to weaken Ms. Elberse’s presentation of her findings. As an entrepreneur running a business from a home office, “much money” has a very different meaning than it does if I am CEO of an international corporation.