Tip:
Highlight text to annotate it
X
We've been talking about the distribution of popularity
and the way in which people's attention is distributed
very unevenly across items, across books and movies and celebrities and ideas
and so forth.
Because some things command a lot of attention.
They're very, very popular.
And most things command relatively little attention.
And that's the extreme imbalance that characterizes
the distribution of popularity.
Now the distribution of popularity also plays
a huge role in markets that sell media, things like books and songs and movies
and similar kinds of creative work, right?
These are all things that amass an audience, that command attention.
And hence, they're really driven by the kind of imbalances
that we see in popularity.
And the point in particular is that these kinds of markets
contain a complicated mix of blockbusters, things that sell many,
many copies, and niche items, large numbers of smaller items that
have limited followings, right, limited number of fans.
And thinking about the distribution of popularity
will actually help us think about how all these different types of items
interact with each other and form the market.
Now this is a complicated picture.
And in 2004, Chris Anderson wrote a hugely influential article--
it later became a book-- in which he coined a term
to describe this ecosystem of popularities
and tried to bring some clarity to this picture.
He referred to it as the long tail, the term
that has since become very, very popular itself.
And it was based on the following kind of picture.
In this picture, we're going to draw a distribution
on x-y axes that tries to capture how popularity is distributed.
It's going to look a bit different from the curves
we drew a bit earlier when we were looking for power laws.
But it will actually capture the same information, just sort of presented
in a different form.
So on the x-axis, we're going to imagine the following--
that you're a seller of books.
And you carry a very, very large inventory, right?
So you're an online internet bookseller.
And on the x-axis, you have ordered all of your books in order of popularity,
right?
So the very first thing you encounter as you go out the x-axis
is the most popular book that sells the most copies.
And the next one is the second most popular book.
And you just keep going.
And you could go out to 100,000, a million books.
And as you go further out the x-axis, they
get less and less and less popular.
And on the y-axis, we're simply going to say how many copies each of these books
sold, right?
So the first point up there on the y-axis
is going to be how many copies did your most popular book sell.
And the next is the second most popular and so forth.
And so it's going to be this downward-sloping curve that
continues way far out down the x-axis, right?
And it falls off pretty fast at first.
Because you're carrying a few best sellers that just sell a lot of things.
But already by the time you're out to the 20th most popular thing,
the 50th most popular thing, you're at much, much lower sales volume.
OK, so in particular, a point on this curve, right,
a point that has say x-coordinate equal to j, y-coordinate equal to k,
that by definition says, the jth most popular book has sold k copies.
Now when we look at this curve-- and this was Chris Andersen's argument--
we can parse it into the part of the curve that sort of goes out
the long arm of the x-axis and the part of it that
climbs up the high side of the y-axis.
And this part that goes out to the long arm of the x-axis,
he referred to as the long tale of popularity.
And it sort of comes from this picture, thinking
about this downward-sloping curve almost as sort
of like a prehistoric reptile emerging out of a swamp.
And right there is its long tail.
And this corresponding-- this big head that rises up at the front,
right near the y-axis.
OK, and the point in the long tail is it contains a huge number
of items, each of which is individually not very popular.
But there sure are just a huge number of them.
And of course, at the other end of the curve
is what you could call the big head of the distribution--
that part that climbs way up the y-axis.
And in the head of the distribution live a small number
of books in this case, small number of items, but each of them
is just extremely popular.
And so that's a way to sort of parse this sloping curve
into two very important aspects of it.
The long tail, many items, not very popular.
The big head very few items, each of the very, very popular.
And if you're a company that sells books, you're in the media business
then an important question you have to ask yourself--
and this was the crux of the long tail argument--
where is most of your revenue coming from, right?
Is it coming from the big head?
Did you make most your money off selling a few best sellers
to huge numbers of people?
Or are you making it off the long tail, by selling many things but each of them
to only a few people?
Either of those is a strategy.
And in fact, you could think of different companies
could pursue different strategies.
One could be a head company that really puts
its effort into creating blockbusters.
And another could be a tail company, right,
that amasses very small followings for a huge number of distinct niche items.
And another part of the long tail argument
was that marketing to the tail, selling all these niche items,
becomes much easier if you're amassing an audience online,
using the internet and the web.
Because it enables you to assemble a very, very large customer base
that's able to find obscure items.
And you in turn aren't necessarily maintaining a physical store.
You're selling stuff over the internet.
So you don't need to carry a very large inventory
of any one of these tail items.
So in some sense the fact that we're thinking
about the tail and its importance has really
been sort of accelerated by the rise of the interest and the web.
It's made a much more easy to do business that way.
Another key component of a long tail argument
is the power of recommendation.
So if I think about companies that are actually
selling these gigantic numbers of items, right-- Amazon selling books, Netflix
streaming movies to people over the internet--
they're making a lot of money off the tail.
And they're driving people to items in the tail using their recommendation
systems.
So Amazon has a recommendation engine that says if you like these items,
you bought these items, here are some other things you might like.
Netflix says, you've watch these movies.
You've rated them.
Here's some other things you might like.
These aren't just nice features that keep users happy.
They're actually essential to the strategy of selling items in the tail.
Because when you have a gigantic mass of obscure items,
you need to be able to efficiently direct people to the ones that
are going to interest them.
And that means you need to watch their first few actions on the site-- what
they liked, what they didn't like-- and develop a model of them
so that you can then recommend stuff that they're going to like.
Without that, they won't be able to find the things
in the tail that actually appeal to them.
And this whole strategy of marketing to the long tail is not going to work.
So all this is wrapped together.
The distribution of popularity and the way
in which technologies on the internet, selling things online, and using
recommendation, automated recommendation, to steer people
toward what they're interested in, makes it much, much easier
to actually have a strategy based on selling out
in the tail of the popularity distribution.
And there's a final point here of course,
which is that items eventually can move sometimes
occasionally from the tail to the head
An item can start out with just a niche following,
and it can amass a very large audience as it becomes more popular.
But the point is often that process is based on certain accidental things that
happen very early in the life cycle of the item.
It just happens to acquire a few devoted fans who tell their friends.
And so in that way, certain accidental things that happen
amplifies their popularity.
And what that means, if it's really based on these accidents,
is that popularity has an aspect that's inherently unpredictable.
And researchers like Duncan Watts have really
promoted this argument in recent years.
That when we look at these very, very popular items,
sometimes we're seeing something that was not at all inevitable.
But they're really the result of these accidental things happened very early
in its life, when it was first attracting its first few fans that
somehow snowballed into this mass of popularity that we now see.
So in the end, we see that thinking about the distribution of popularity
taps into a number of important issues.
It first of all is closely related to the rich-get-richer dynamic,
by which people imitate each other's decisions.
And it has implications for the design of markets, markets
particularly in settings where the fundamental driving constraint
is attention, and how it's distributed.