Learning the long tail

It’s been about two years of a state of somewhat attentive yearning for evidence of the idea I unfolded in my last writing. That I’ve hunted for in Spotify and body art. The idea that, by by accessing an unlimited catalogue of digital goods, with its recommendation engines and various accelerating accoutrements, users might discover greater variety in said category of goods. That in discovering this variety, they may come across goods that respond to aspects of their personalities that are not strictly unique, but that make them different from most people around them. That in experiencing this recognition, or expression, of what sets them apart from the daily mass of others, they may experience more excitement, fulfillment, transcendance, or something else that feels good inside, than when consuming something served to everyone by traditional distributors of only the most enormous sources of supply, like Wal-Mart or Hot 97. That this more intimate and personal experience of consumption might draw more of their time, attention, and associated dollars away towards the creators of these niche products. That this growth in dollars towards creators of niche products might provide livings for a greater number of creative people. And finally, that these creative people might be important to the “Future of Work”: that frustrating branding for all the people who think about whether and how quickly computers will automate jobs. Every warm and cold study of job loss potential agrees that workers in the future will need to use their creativity. Maybe this is how?

But only in the last week did I sheepishly discover that this is an established idea in, I don’t know what the name of the field would be, but the field of serious people who think about what consumers want and why and how the internet changes that. And that it has a name, and that at least fifteen years of surrounding literature. That name is “The Long Tail.” 

I don’t want to patronize my very small audience of people that are much smarter and more graphically inclined than I am, but just in case anyone is on my level: the long tail refers to the graph of total consumption of a category of thing – say, music – on the y-axis, and the raw number of distinct things available in that category on the x-axis, in order of popularity. For a lot of things, some of the distinct things – in this case, specific songs – are way more popular than all the rest, so the line starts high on the y-axis, and slopes downward as is moves along the x. Because the number of unpopular songs is much larger than the number of popular songs, the “tail” of seldom-sung songs extends for longer along the x as the “head” of hits does up the y. Fans / proponents of the long tail theory ( I guess count me in) believe (hope?) that the tail will “fatten” as the internet expands variety and our capacity for finding it. 

Since I was already feeling sheepish that I’d been walking around for two years feeling like I had some precious insight when a best-selling book was written about this exact thing, it was extra exposing of overconfidence that the very article that formalized this hypothesis in my brain then concluded without qualification that it was wrong. 

Enter this article, where in a 2008 essay, an HBS prof wrote that the long tail is a myth, using data so high quality that discovering today’s version of it two weeks ago would have made me feel like the Emperor of Knowledge. I felt checkmated by their authority, still felt foolish at being fifteen years late to the debate, and as it was at the end of a Friday, my brain had no resilient attention with which to read critically. And she clearly showed that consumer taste in music and fashion, two things whose consistent daily interplay with consumers’ sense of personal meaning and emotion made me count on them to be cesspools of long tail activity, were heavily concentrated. 

But even late-Friday brain could take refuge in the fact that her data was pre-iPhone. And when I returned on Saturday to think about what her data actually said, I didn’t think her conclusion was exactly wrong; but the long tail lived anew. One source at a time:

Starting with Rhapsody: a music subscription service with unlimited listens and no marginal cost to discovery, making it a solid landscape for comparison to today’s dominant streamers, minus recommendation engines, playlists, and other great machines of niche discovery. Within a library of one million songs, the top 1 percent most popular accounted for 32 percent of total listens, and the top 10 percent accounted for 78 percent. 

The bottom 90 percent of tracks getting only 22 percent of listens? That long tail is rat-thin, you might say. But then, as the author notes, in a tonally out-of-place twist of argument: that top 1 percent of the one million song sample, is about the limit of what you might expect to be carried at Wal-Mart (I’ll have to trust her; haven’t darkened one’s door since I last got run off by a security guard for registering voters without permission). That means, in theory, that 68 percent of listens on Rhapsody in some year of the 2000s went to tracks nobody had access to in the 90s unless they lived near a cute record store. Let’s say in the 1990s, a full 50 percent of all music listens nationwide went to the bottom 99 percent of tracks: if all today’s streaming companies did was increase that share to Rhapsody’s 68 percent, it would represent an enormous shift in attention towards smaller artists. 

An Australian DVD mail subscription service called Quickflix (lawsuit, anyone?): Same vibes. In a library of 15k movies, the most popular 10 percent got 48 percent of views. If, as the author says, 150 movies is greater than the annual output of major studios, then 52 percent of consumption went to films beyond what Hollywood can make in a decade. Those films were probably available at Blockbuster, but it’s more than a kiddie pool of eyeballs for people who didn’t feel like taking a meeting with Harvey Weinstein. 

Nielsen SoundScan data on “home video” sales between 2000 and 2005: Wth are home video sales? Is that people buying VHS’s? Did anybody do that? This data doesn’t say anything about how the internet changes behavior, since buying a VHS is among the more un-internet things I can imagine doing. But it could reflect extra-web changes in consumption patterns, which I’m still interested in. The verdict: the total number of films selling a handful of copies popped up like a cork. At the same time, the market got MORE concentrated. How could this happen? A massive increase in overall sales, powered by a large increase in niche purchases and an even lager spike in buying blockbusters. The long tail got way longer; but the head got WAY taller. 

Nielsen SoundScan data on “recorded music” sales between 2005 and 2007: Interesting time window, because it covers a period in which digital music sales rose from one-third to two-thirds of total recorded music revenue: the rise of iTunes and the fall of Tower Records. The impact on the consumption graph of having a large, instantly searchable library, if only with 10 second samples and wimpy discovery algorithms. To wit: a longer and flatter tail, yet more concentration among the most popular titles than before; which must mean a taller head driving overall consumption to new heights. (Right? But she only mentions the first half of that semicolon-divided phrase; an explosion in overall music consumption doesn’t track with my memory of the time and seems like a pretty big wave for the writer to skip; this is making me concerned I’ve fallen down a logical crevasse). 

Assuming I didn’t. I don’t know how to feel. The last two examples kind of spook me. The most fulsome fantasy of the long tail is one where the practice of popularity breaks and created goods are as flat and wide as Kyrie Irving’s earth. Not one where more dribbles of strange sauce leak out the back of a huge machine that sings fewer and fewer songs to more and more people. 

But: the long tail theory, as presented for slaughter in this refutatory paper, says that “online channels” will fatten the tail. And even in this old data, in 2008, when “online channels” meant your laptop, when real “online channels” didn’t exist, when the multiplying possibilities of networks were merely spun sugar in the minds of early adopters, the long tail still grew. So there’s that.