Thursday, January 1, 2009

Long Tail Doesn't Apply?

As observers have started to track sales of digital goods more closely, some apparently-contradictory evidence has started to appear about the "long tail" theory of sales in the digital domain. The concept: cheap digital distribution changes the retail sales function, allowing profitable sales of low-volume titles or items on a scale not possible in a physical distribution strategy. Most observers instinctively would agree. 

The key prediction has been that online distribution would allow niche businesses, content and goods to thrive in a digital distribution context impossible to sustain in a physical distribution context. As commonly understood, perhaps an additional 20 percent increase in volume should be feasible, as well as a change in "mass culture" that would fragment demand. 

But some sales data contradicts the notion. A new study by Will Page, chief economist of the MCPS-PRS Alliance, a music royalty collection organization, suggests that online sales success still relies on big hits. The study found that 80 per cent of all revenue came from around 52,000 tracks. For albums, of the 1.23 million available, only 173,000 were ever bought, meaning 85 per cent did not sell a single copy all year.

Frankly, the long tail now appears, in one sense, as the triumph of hope over experience. Many "wanted" online distribution to change purchasing patterns. The notion was that once huge variety was available, tastes would change. It isn't so clear why it is assumed "tastes" will change with different distribution. It is clear why fulfillment will change with cheaper distribution. But efficient, or better, distribution still should result in a "long tail" of demand that is the same as a distribution-induced "short tail" of demand. 

The reason derives from the theory itself. The idea behind the "long tail" is not actually new, and dates back to an Italian economist, Vilfredo Pareto, who in 1906 coined the Pareto Principle, popularly known as the 80-20 rule. Basically, the idea is that in much of life and nature,  roughly 80 percent of the effects come from 20 percent of the causes.

The same concept is known as Bradford's Law. 

As implied by the theory of the long tail, online distribution should allow retailers to sell small volumes of hard-to-find items, instead of selling a smaller number of highly-popular items. Most people can grasp that. What isn't so clear is why that expected distribution curve will be a "new" Pareto curve, instead of validating the existing Pareto curve. 

Under any normal set of circumstances, a Pareto distribution is what one would expect to see. Some have pointed to music sales at Rhapsody, an online music service. Of the 735,000 items for sale, 39,000 account for 78 percent of sales, while 796,000 titles represent 22 percent of sales. 

Likewise, Netflix data suggests 20 percent of total rentals are of "tail" or low-volume titles, while 80 percent of rentals are basically "hit movies" one would expect most people to be interested in. 

Is that confirmation of the operation of a Pareto distribution? Yes. Does it represent incremental sales of 22 percent that might not occur in a physical distribution scenario? Yes. Are the results unexpected? Not if one expects to see a Pareto distribution. 

Is the idea wrong, or useless? Not really. A Pareto distribution can assume a 70-30 pattern, for example, suggesting a bigger role for niche products than before. That represents an important shift of opportunity for providers of niche services and products because of online or Web distribution. 

But the long tail might not mean a revolution. Forrester Research, for example, estimates seven percent of retail sales in 2008 will have been made online, up from 3.2 percent in 2007. What does that mean? Most sales follow a Pareto curve: 97 percent of things sold still are sold the traditional way. There will be further shifts, of course.

But Pareto would suggest online sales will settle in at around 20 percent of total sales, at best, on a sustainable basis. 

3 comments:

Phoebe said...

Thanks, this is an interesting perspective and I agree that the long tail is meaningful but isn't going to cause a sudden revolution. What do you think of Chris Anderson's claim that the results of the study can't be interpreted until the source (music service) is revealed?

Gary Kim said...

I suppose I'd say the same thing, but if the Pareto distribution is truthful (and I think it is), the results should not vary much by providers). The other thing: does the "long tail" apply most meaningfully to sales volume, sales revenue or profits, or all? In some ways, I think people want to believe that the "long tail" will disrupt markets. The theorem itself suggests no such thing.

Anonymous said...

The unique and global commerce channel introduced by the internet does not necessarily mean that low-selling and/or low cost items will boom in sales volume, of course.

Nevertheless, at the fast pace we are seeing, it is true that (for example in the music industry) artists are starting to consider the internet as the preferred medium of selling their work, which, together with the ease / speed of placing their work online, and the continous appearance of new bands, has clear effects on the revealed statistics (1 Britney will always sell more than 20 new "still shadowed" artists, also because there is an "acquaintance curve" with new artists and their works).

I don´t think the music industry is a good example. Isn´t the long tail suppose to apply to goods that are rarely sold, or companies that have trouble in selling outside the web? For the music industry, the web is just another yet-to-come cash cow, which will probably be fed by some provately-owned and copyrighted peer to peer paid content distribution network.

Many Winners and Losers from Generative AI

Perhaps there is no contradiction between low historical total factor annual productivity gains and high expected generative artificial inte...