Good Catch Dennis. This graph is very misleading. It’s basically saying that only 10% of youtube videos are seen (in their first month) by more than 1,000 people. Which makes sense; not every video goes viral, and 1,000 views is no where near enough to be considered viral.

My question would be how do first month numbers compare to year long numbers, but that would require another study all together, and I think it would better capture the “long tail” effect that marketers might be interested in.

Are all youtube vidoes eventually seen 100,000 times, or are the only videos that get to that status the ones that get 10,000 views in their first month? As a marketer, I would pay good money for the answer to that question.

The graph uses a perfectly normal logarithmic scale (as used extensively in science and engineering) which is quite appropriate for this type of study. It might be misleading to someone who is more comfortable with the linear maths common in the financial world, but that is not an issue with the graph itself… which is clearly marked. Logarithmic scales are useful if you expect to see a power-law in operation (which are surprisingly common in population studies such as percentage of people in various wealth brackets, and percentage of businesses with various numbers of employees).

I’m a little more concerned about the graph having only one sample point for every 10% of the population, and straight lines between the points. If they indexed a large number of videos then why not also plot a large number of data points?

Thanks for the comments! I was meaning to indicate there was a problem with using that kind of scale – in fact, in this case it was the only option. However, it’s a problem when people might look at the graphic representation and believe it is a “proportional” representation that looks like a long tail story – I’d bet dollars to donuts that the majority of people that look at a graphic like that do not make the logarithmic adjustment in their head.

1-agreed that the graph is misleading
2-what, for you, is a long tail? i was hoping you would define this. of course the ‘length’ of the long tail can only be defined relative to other tails and we should view it as a continuum rather than any sort of blockbuster/long-tail dichotomy. what i am asking is what characterizes a long tail? if there are a few wildly popular “hits” (blockbusters), but also a robust think-tail through much of the rest of the distribution, is that a long tail or not? it seems to me that would be a situation where *both* the long tail and blockbusters co-exist. thus, they should not be discussed as antonyms.
3-why do we call it a ‘long tail’ when we really mean a “thick” tail? a situation, like the music industry in the late 90′s, where there was TONS of stuff put out, but few of them sold much and there were a few major blockbusters was a situation of a very long tail, just very thin. when we say “long tail” these days, we are not talking about that situation, but rather the situation where the tail is thick as one moves down the percentiles.
4-as a comment indicated above, these results are only the first month of viewership. all conclusions should bear that in mind. when you say that “this looks like a blockbuster model”, you should state that “in the first month after posting, this looks like a blockbuster model.” i could imagine that after 3 months, 6 months, 1 year, the shape of the curve will look different. i’d love to see those curves plotted against eachother to see if the tail grown thinner or thicker over time.

“what, for you, is a long tail? i was hoping you would define this”

I think that’s been pretty solidly defined elsewhere. In basic terms (thinking from a business perspective) you can sell a standard product with high volume manufacturing and completely inflexible, or you can work with some form of flexibility in your design to sell a customized product that tailors to a specific market. Building flexibility into the design increases costs and reduces the advantage of high volume manufacture, but it opens up more market opportunities (regardless of whether that might be advertising opportunities or selling product or building community, whatever your objectives might be).

So the balance of the tail versus the “blockbuster” is a question of which region of the curve contains the most potential profit. As a first estimate this is the area under the curve, but a more refined estimate will also consider the level of competition to be found at different parts of the curve and the nature of the customers too (loyalty, etc).

There’s some moderately good Wikipedia articles on long tails and power laws too…

http://en.wikipedia.org/wiki/The_Long_Tail
http://en.wikipedia.org/wiki/Power_law

Strangely, the power law article claims “In general, power-law distributions are plotted on doubly logarithmic axes, which emphasizes the upper tail region” but then the example plot is on a linear axis (looks linear, no actual markings on the graph). That’s Wikipedia for you, a camel designed by a donkey, but never-the-less highly educational.

[...] Tardes political economy of knowledge fits in Heylighens conceptual schema. See e.g. Trust in Communication between Individuals: A Connectionist Approach (2008). In recent analysis of Web 2.0 we found also indications that Adam Smith’s and Marx’s analysis of the market and commodities cannot be applied on information. One of the misleading concepts in Web 2.0 is the long tail, a marketing based concept (consumer demographics). It points to the context of Smiths free market as a driving force in the development of the World Wide Web. This is a cliché that easily can be undermined. Denis Hancock did put the long tail of Youtube in perspective: [...]

[...] Tardes political economy of knowledge fits in Heylighens conceptual schema. See e.g. Trust in Communication between Individuals: A Connectionist Approach (2008). In recent analysis of Web 2.0 we found also indications that Adam Smith’s and Marx’s analysis of the market and commodities cannot be applied on information. One of the misleading concepts in Web 2.0 is the long tail, a marketing based concept (consumer demographics). It points to the context of Smiths free market as a driving force in the development of the World Wide Web. This is a cliché that easily can be undermined. Denis Hancock did put the long tail of Youtube in perspective: [...]

More great comments… I wanted to respond to Nathan J’s #3 & #4 specifically.

#3 – I think you are on to something with the long / thick tail distinction. Where markets vary from the blockbuster model is not when the tail is long, but when it is thick… interesting to think about.

#4 – I thought about this 1 month issue, and I personally believe it’s not a big deal. For the distribution to look any different over a long period of time (in a way that makes a long tail story), one would have to believe that videos that were relatively unpopular when first released somehow become more popular later on. That wouldn’t make much sense to me. My hypothesis would be that if you extend the data set out over multiple months, the blockbuster model would actually intensify – those videos that were popular still having a shelf life of some sort, while those that weren’t die off. Again, hypothesis. Would love to have it proven right or wrong…

[...] bandwidth.� Playing host to the world’s home videos is expensive, and the long tail (see:� Denis’s previous blog on this topic) means that the vast majority of videos lack potential to generate ad [...]

[...] host to the world’s home videos is expensive, and the long tail (see:� <a href=”http://www.wikinomics.com/blog/index.php/2009/03/04/putting-the-youtube-long-tail-in-perspective/&#8221; target=”_blank”>Denis’s previous blog on this topic</a>) means [...]

[...] [...]