The Long Tail of Time

Why are some things less popular than others? There are lots of reasons, of course, but the two I've been focusing on most in my research are the breadth of their appeal and their age. Things of broad appeal tend to sell better than things of narrow appeal. And new things tend to sell better than old things.

If you've been reading any of my work you'll know that I'm quick to emphasize that there's little connection between those two factors and the underlying quality of the product in question or the extent to which someone will be satisfied by it. Indeed, I argue that products aimed at narrow audiences, while not for everyone, tend to satisfy that minority that they're aimed at better than one-size-fits-all fare (think niche blogs or documentary films). And older products that still come up in recommendations are often those that have passed the test of time (classics).

When you look at a basic demand curve, the reasons why some things sell less well than others are lost in the merged rankings. Popularity is a multi-dimension thing: factors that determine an album's rankings, for instance, can include not just the quality of the music but also its genre, its release date, the fame and/or nationality of the band, similarity to other artists and so on. But it's all lost in the single dimension of a bestseller list, which obscures all that in a mushed-together melange of apples and oranges. (You can read my earlier rant about popularity lists here.)

The following is a conceptual chart of the primary factor accounting for a song's position pretty far down a demand curve. As you can see, looking at popularity alone it's hard to tell whether something isn't very popular because it's of narrow interest or because it's old (or any combination thereof, although I've only shown those two for simplicity's sake):

If you tease apart those two dimensions, you'll get something like the following (again, the lines are just conceptual; different markets will exhibit different actual shapes):

Or, to put it another way, you can plot those two dimensions in quadrants like this (darker = more sales):

I'm still wrestling with how to visualize the actual topology of the surface connecting those two dimensions. I suspect it's kind of a spout shape, but without actual data I'm just guessing.

If you disregard quality, a rough equation of popularity might go like this:

Sales = Breadth of Appeal/Age

...where Breadth of Appeal goes from 100 (% of the population) to 0 and Age goes from 1 to infinity. So if something is of broad appeal and brand new that's 100/1 = sales factor 100. If it's niche and new that might be 10/1 = sales factor 10, which happens to be about as popular as something of broad appeal but older (100/10 = 10). And so on. Totally oversimplified, but it's a start.

Complicating matters, Kevin Kelly reminds me that "the fun thing, of course, is that narrow things can become broader as they age (Mississippi blues, which I would bet are more popular now then at their birth), or start broad when young and narrow as they age (opera)."

So there's lots here for the math geeks to chew on. Meanwhile, I'd love to hear any other ideas of better ways to visualize the many dimensions of popularity.

UPDATE: I thought about it a bit more and roughed out the following conceptual topology. Now to crunch some actual data and see if reality fits the theory:

Comments

[SoundRater]

I presume that *sales* means sales over a given time period, such as sales per annum.

Cumulative sales may show an interesting picture in relation to *quality* as the fall in sales per annum may be slower for *quality* items, eventually resulting in greater cumulative sales over the life of the product. This could be regarded as greater *popularity*.

Is it possible that a connection between the age of the item and underlying quality could be seen in the rate of decline of sales p.a. over time with *quality* items tending to show a lower rate of decline. Items in the longtail would show such a connection more strongly as they are less dependent on marketing budgets to fuel initial sales (they may even show an increase over the long-term as superior quality becomes recognised).

[Francis Hwang]

I touched on this back in January, in the midst of discussing the idea that all online newspapers should completely free their archives:

... as we study the long tail more seriously in the years ahead, I suspect we will discover that it’s most applicable to content that cannot be evaluated quickly, like essays, novels, records, and films. News articles, by contrast, are engineered for quick consumption. They announce themselves with straightforward headlines, focus on concrete events, and avoid theoretical speculation. If you didn’t read a news article on the day it came out, the odds that you’ll rediscover it later are slim.

I suppose, also, that some kinds of content are made to last more, and build audiences more slowly over time. It's easy to get a used copy of, say, a Nelly CD or "The Devil Wore Prada", because that stuff is designed to be more disposable. But you're less likely to see people selling off, say, Marilynne Robinson's latest novel. If you bought that at all, odds are good you're the sort of person who buys books and likes keeping them around.

[csven]

And as niche markets grow, the whole system gets flatter... and flatter... and flatter......

[Chris Anderson]

SoundRater,

Yes, that's correct. Although the above charts are conceptual, most of my data sets are sales per month or sales per quarter.

Chris

[Eric Colson]

Would be interesting to see if there are "ripples" along the age axis. Depending on the industry, some products experience declining popularity to a point at which time they become "vintage" or "retro". Products that achieve this classification would likely experience a temporary bump. This might be a good way to distinguish the classification of "classic", which could be characterized as a "sustained increase".

[Norman]

I can't see how knowing age/new-old/appeal relationships and being able to put it into a special presentation will make anyone any money which must be the end result. Can anyone answer this?

[Francisco Gutierrez]

I like the idea of separating the two sources of popularity, but your model fails to account for the adoption curve of a new product. A product starts by being used by a few "early adopters", and then its usage increases with its age, until it finally reaches the top of the Bell curve and usage decreases. So the decay part only applies after the early majority has adopted. That coupled with the broadening of appeal from niche to general makes reality a lot more complicated.

[(the other) Chris Anderson]

There are certain phenomena, such as "cult" classics, which may not follow those sales curves. I don't know exact figures, but I imagine that films like Bladerunner, or others more obscure when released, have had increases of sales long after the inital marketing campaigns wore off. I suppose this is somewhat like Francisco had in mind. The interesting cases to me are those where a product is abandoned by its backers, only to later surprise them by its success.

[Justin Kownacki]

Age and popularity seem to be unequal as far as tangents go, especially considering the myriad variables that could affect popularity -- advertising, zeitgeist, need vs. want, competition, pricing, delivery method...

For example, Blade Runner on film in the '80s is a different beast from Blade Runner on DVD in the '00s -- or (egad) Blade Runner on LaserDisc in the '90s...

[nkuriakos]

Couple of comments:

Are sales in units or $? Unit sales allow you to make more valid comparisons.

Newness is relative. Look at music genre fads. We are at the cusp of 80s music coming back into vogue. Is this old or new? Or popularity of classical or Jazz music among newer generations of listeners. Perhaps awareness or 'learned newness' would work.

[Jake Lockley]

I think your forgetting the relevance of the product to the consumer. Take for example the sond "Unchained Melody" whose real popularity only came after it was used in the film "Ghost". It became relevant to a lot of people who wanted to use it in their wedding. What you describe is sound for a product whose life cycle lives and dies with it's release and the 'marketing' associated with it, but it doesn't address the relevance of the product to the consumer throughout it's entire lifetime in the wild.

One size fits all could be better describe as 'greater relevance' to the masses versus greater relevance to individuals. Old movies that young kids see for the first time have greater relevance to them because it's all new. Resurgence of fashion trends is a great example of this. It really targets mass appeal using old product. Another interesting example is a company that made wand metal detectors before and after 9/11. The product has been in existence for decades, suddenly it becomes relevant to more people.

So I guess what I am saying is that age of a product in the wild is less important than relevance to individuals at a given point in their life and history, which is directly tied to it's marketing whether it be an event like 9/11 or an advertising campaign that brings it's relevance to bear. Usually there is more marketing when a product first comes out than anytime in it's history, but like I said, that's is not always the case.

[David]

In reading the above comments, it occurs to me that what's most relevant is not so much the age of the product itself but the age of the buzz about it.

The example cited above of an old song being used in a new film is a great example. And I'll bet sales of old Dylan albums are up because of the recent documentary about him.

[coturnix]

The short excerpt from Hwang above reminds me of a discussion we recently had about blogs themselves, i.e., how longer, more thoughtful posts, although invisible from the front page, keep getting linked, or arrived to by Google, months or even years after they were initially posted. They are most certaonly NOT burried in the archives never to be seen again. People who tend to write such posts are also more likely to try to turn their blogs into books, for their own sake, or in order to enter the Blooker Prize contest.

EdCone.com has an interesting take on this post - the recurring popularity of seasonal stuff (e.g., Christmas-related movies and music right around this time of year). I am wondering if some other products also have regular oscillations in popularity, tied, for example to the Olympics (or Soccer World Cup) every four years, or US Presidential elections every four years, or celebrations of anniversaries of important events or birthdays (I can think of various Darwin-related events happening every ten years).

[georganna]

Do the curves smooth spikes in popularity as cyclical fads, such as in clothing, return every decade or two?

[MC Stephens]

I'm late to the party, sorry, so wouldn't be surprised if you've decided how to turn your long-tail-over-time idea into a po' man's econometric model. It's a little kinky now... because you've started with an image rather than defining "Appeal" characteristics. Keeping the series comparisons simple and the data set easy to locate, I'd choose one content title, say, the bible, which you know has a loooong publishing history and diverse "Appeal" -- many languages, media, geo-parts. Then you version the bible by publishers (e.g. Gutenberg, Johovah's, etc.) which are a proxies for "Appeal" (i.e. product characteristics). Because the set also includes Q (vol or $) over time, the spreadsheet can easily generate a linear image or bubble chart to dramatize scope (more than one ratio). As always, the interest is in the details of one event (the bible) and the way you interpret or predict the evolution of its markets.

[MC Stephens]

PS. No one can express all the data relationships in one chart. A longitudinal analysis of "popularity" -- some measurable difference between sales and aesthetic? -- is better served in chunks. I'm no math geek, but I do think a good stat story always needs to be explained.

[Michael Liubinskas]

Sales volume may be higher for new products of broad interest, but prices are often higher for older and narrowly focused products. The challenge has been that to find people who want the old and narrow marketing and distribution costs make them very expensive.

The long tail is about a market where the margins are interesting for a new set of customers.

I'd like to see;
a. How the margins change as you go down the curves?
b. How the total pie changes when you combine longtail with leverage (selling digital things that can be duplicated for no cost) and scalability (selling digital things via P2P with very little distribution cost)?

If you answer these questions then you work out whether the movie, music and software industry are better off 'letting go' and becoming a high volume, low margin business or if they should continue to hold on tight?

P.S. Interesting posting about markets;
http://www.slate.com/id/2131573/fr/rss/

[csven]

For some of the people bringing up cyclical issues and the limitation of a 2D curve, something I posted a while back might be of interest: Link

There's an additional, more recent image showing the evolution of the shape I call an "ecoToroid" here. The lowest shape goes back to ancient history and the topmost shape is more current.

Just an FYI.

[Niklas Dahlin]

But... isn't there a risk of conflating "sales" and "appeal"? After all, how to measure a value like "appeal" in a non-trivial sense?
Take poetry post-1945 as an irritating example, say sales figures for Ed Dorn's "Gunslinger" and the collected works of J.H. Prynne. Both are "of high appeal" for a given group, but is there any way of telling them apart given that criterion without refering to actual sales?

[csven]

I just happened to see the updated image adding a third dimension. It's good to see that it looks very much like how I expect the 3D shape I've created to morph as it changes. In other words, as the profile revolves it changes over time and consequently the overall form changes (which yields the image of the various snapshots in time I linked to above).

I'm very much looking forward to seeing some data posted.

[Tatevik Patatanian]

Hello,
My name is Tatevik. I just found your blog online because I am doing a research paper on your Long Tail theory. I'm doing this research paper for my Mass Communications class. I would like to ask you some questions about your book.1)How might The Long Tail cause changes in mass media? 2)How might The Long Tail cause changes in advertising? 3)What types of meida opportunities might arise in a Long Tail world?. If you can please e-mail me back, it would be a great help.

Thank You