most people happened to want to wear blue ties, and you wear a red tie.
Now think about the problem faced by
manufacture and seller of ties if there's
a cost associated with producing more variety.
And typically there, there is.
The manager of that factory and that store might say, you know what?
most people want blue ties.
Only one fellow wants a red tie. We're not going to make any red ties.
And as a result, the person with the taste for the
red tie, the preference minority, is not going to get what they want.
So I'm going to explain in the next few slides how this problem can
be overcome through online intervention, particularly
through sellers who operate on the internet.
to make a more personal example here,
I often wonder myself in myself in Philadelphia
why it is that I can't get Vegemite when I go to the local supermarket.
Let me show you a picture of Vegemite.
I'm sure my friends in the UK Australia,
and New Zealand and other places have seen it.
Here's a little kid with Vegemite all over his face.
It's a delicious black paste that you have
on toast with cheese and avocado and stuff like that.
I look for it all the time when I go
to the supermarket in Philadelphia, but I can never find it.
Why?
Because I'm probably one of the only few people that would actually buy it.
So the store manager who wants to stock items that are profitable
and sell frequently, is not going to pay attention to my preferences.
I'm a preference minority.
And therefore I'm not going to be able to get what I want.
The internet, however, could solve this problem.
So let's go again,
a little bit of background. Here's a slide of our friends at Quidsi.
And, the research for this particular article
that I wrote and published with a friend
of mine, Jeonghye Choi, at Yonsei University, is
based on data that we got from diapers.com.
here's the article itself, the title of the article.
if you want to you can always go and read it for
more details, but I'm going to give you the flavor of the main findings.
The article is called Preference Minorities and the Internet.
So, let's see how this works.
The notion is that if you're selling things
online, that gives you the ability to aggregate people.
Maybe there's only two people in
Philadelphia that like Vegemite, but in all
the towns in America maybe there's 100,000 if we added all those people together.
And even though it wouldn't be efficient for us to serve them in
individual shops, we could sell that product
over the internet, that's the basic idea.
So preference isolation is going to bring
shoppers to the online marketplace instead of the offline marketplace where
they're locked out, and it's going to do that in a systematic way.
So now let me get into the details and conclusions of this particular study,
again using the data from diapers.com that
we're all by now, pretty much familiar with.
So, in order to test out our idea, that people who were different from
their neighbors weren't getting served adequately by
offline stores, Jonghei and I went out, and
we did a little bit of a field study, kind
of a fun field study, and here's what we did.
There's a chart on the slide that you can see.
Where we went out and we visited different supermarkets in the Philadelphia area.
Three supermarkets were from the Fresh Grocer chain.
And two of the supermarkets or, or stores were Walmarts.
Now, what's interesting is all of those stores, all
the Fresh Grocers, they're all the same size stores.
But their local markets were different with respect
to the number of households in the local market
that had kids.
So again, if you look at the chart, you can see that Store 1,
about 10% of the population in the
immediate area around the store had young children.
They were households with babies.
Store 3, on the other hand, about 16 percent of the
households in that neighborhood where that store was located, had children.
So, what does this mean for our idea or our theory?
Well, the people who live in the market where there's only 10%
of households with kids are going to be relatively more neglected by the
supermarket than in the market where there are 16% of households with kids.
So if my friend Chris is the manager of the store, he's going to
say, you know what, not that many people in the 10% market have kids.
I'll just have Pampers and a few leading brands on the
shelf, and I won't worry about having a lot of variety.
If, on the other hand, he's managing the store where 16% of the target market has
children, he's going to say, you know what a lot of people in
this local neighborhood have kids I'd
better cater to their tastes and preferences.
And so what you see in the chart is
in neighborhoods where there's a higher fraction of households with
kids, the actual stores have more shelf space, more linear
square feet, and more variety of product of the shelf.
Now of course this doesn't prove our theory, but it does indicate
that local stores pay attention to the composition of the people who
live in the neighborhoods, and then they stock merchandise accordingly.
As a little side note, this was kind of a fun thing to do in Philadelphia.
Jeonghye and I had a pink measuring tape, we were running around trying not to
get caught by the store managers measuring
how much space was allocated to these things.
If you've been to Philadelphia, it's kind of a tough town.
It's a place where they boo Santa Claus, at least the football fans do, so you have
to be a little bit careful if you're running
around with a pink tape measuring stuff in stores.
Okay, so let me elaborate a little bit more on this next slide
with the actual theory, that we built up to try and explain this concept.
So imagine we have two different markets.
this is just purely a conceptual argument, so the first market, Market A, is a
market where there are 100 households with
babies or 100 households in the target population.
The total population of this particular market is 200 people,
so half of the people in this local neighborhood have
the characteristic that we want, in this case it's whether or not
they have children, but you could think about this idea much more generally.
It could be half the people want Vegemite, or
any other product that you could come up with.
And so notice in that market there is one
store, and the store is 200 square feet in area.
And so the manager of that store says, gee, half the people in the market have
this particular characteristic, let's say households with kids
So I'm going to allocate half of my store
to products that cater to those people.
And again this is just a stylized example to make the
main point but hopefully you can see where this is going.
Now in Market B, again there are 100 people who have
the characteristic we're looking for, in this case households with children.
But the total population of the market is 1,000 people.
So these people with kids are a little bit more rare in this case.
They're only 10% of the population.
Now notice however, because Market B
has more people in total. It has a 1,000 instead of 200.
there's more stores.
And we're just assuming that the number of stores grows with the population.
So it, in a market of 200 people, if there's one store,
in a market of a 1,000 people, there will be five stores.
Now again my friend Chris who does a lot of store managing I guess.
He's in market B and he says gee, I'm running these fives
stores, and each store is two-hundred square feet in size, and I notice
that 10% of the households in the
local market have children, therefore I'm going to
allocate 10% or 20 feet of the space in my store to serving that group.
So what you notice here is even though in both Market
A and Market B, the target market is the same size, 100
households with kids in A, 100 households with kids in B, the
extent to which the offline market is paying attention to them is
very, very different.
The customers who live in Market B,
everything else held constant, should be more likely
to want to buy their products online, in this case from our friends at diapers.com.
So let's see if that is in fact true.
So now Jeonghye and I went and we looked at the real data.
This is just a map, it's a black and
white map, but hopefully you can get the idea here.
This is an area of Los Angeles county in the United States,
and what you notice is in the top map there's an indicator
of how isolated the customers are.
We call this the Preference Minority Index or PM Index.
And the darker the color, or the darker
the shading, that means customers are more isolated.
That's in the top part of the map.
In the bottom part of the map, these are the sales to diapers.com.
And again, you notice the fifth quintile, or
the area that's having the most sales, the darker
areas, are sort of the same dark areas in the bottom part of the map as they
are at the top part of the map.
So this is indicating some support for our theory that when customers
are isolated, they're more likely to
use online merchants instead of offline merchants.
So the next thing that we did after looking at the
raw data is do what a lot of us here at the
Wharton School will do, whether it's Pete or Barbara or myself, is
we ran some statistical analysis or some econometric analysis on the data.
And what we did was we tried to see if it was in fact the case
that sales at diapers.com were higher in markets where
the customers that they were focused on were more isolated.
Everything else held constant.
So we controlled for the education level.
We controlled for the number of stores in the area.
We controlled for the population density.
We controlled for the income.
All the things that you might think would effect online verses offline buying.
So all of those things were held constant in our study.
And what we
found was, yes, there was a highly
statisically significant effect of isolation on sales.
And markets where customers were more isolated, they were more
prone to go online, and sales at diapers.com were higher.
So now I'm going to explain the magnitude
of the effects, which I think is really interesting.
It was also very useful for the company, for the guys at Diapers.com.
So most of you, I think, are familiar with the idea of a percentile.
But let me explain that,
because that's going to be important for understanding these results.
So if you've ever taken a standardized test like an SAT test or a GMAT test,
any test at the end of high school or to get into college, those kinds of things.
You'll remember that when you get your score back, in addition
to the raw score, you typically get a measure of percentile.
So where is it that you ranked, relative to everybody else that took the test.
So if you were in the 90th percentile, that's pretty good.
That means only 10%
of the test takers beat you, and you beat you know, 89 90% of everybody else.
If you're in the tenth percentile, which I'm sure that none of you were, that means
that you only beat 10% of the people and in fact 90% of the people beat you.
So, what Jeonghye and I does, we used this same
concept, but we applied it to our preference minority index.
So, we looked at all of the locations in the United States.
All of the areas, all of the zip codes that
were really, really isolated in terms of people with kids being relatively rare.
And what we found was, for the zip codes of the 90th
percentile on that index, the online sales at diapers.com were 50% higher.
So think about that result for a moment.
If we have two zip codes that were absolutely identical in all respects and
in particular, these two zip codes had
the same total number of households with children,
100 here and 100 here, just like in the example.
If this was a zip code that was more
isolated, the online demand at diapers.com was 50% higher.
So we think that's a very interesting finding and also one that
internet retailers can actually use when
they think about online offline interaction.
Now if you think back to the long tail you might
remember the different brands also have different levels of sales so the
most popular brand has the highest level of
sales and the baby diaper category that's Pampers.
And the second one is the second most popular and so on
down all the way out into the tail or into the niche products.
So the same thing happened here.
What we found was if we looked at
particular products that were niche products, and we compared
an isolated market versus a non-isolated market, the
sales in the isolated market online for niche products
were about 125% higher.
And I'm going to show you this in a diagram to make it easier to remember.
So now here's the diagram that explains everything that
we've discussed and brings it all together in one place
and also, importantly, relates it back to the other
key idea, the long tail that we've all ready discussed.
So let me show you what's going on.
There are three pieces to this diagram that
are important in terms of understanding the overall concept.
So first on the left-hand side
we see Market A and Market B, those are just those markets that
we had previously where we looked at the fraction of households with children.
So notice that Market A is a market where offline retailers are paying a lot of
attention to our target customers and households with
babies, because they are 50% of the market.
And Market B, the offline retailers, aren't paying much attention
to our target customers who are the households with children.
And that's reflected at the bottom of the slide with the
little thumbs up and thumbs down in the two markets.
So what does that mean for the way
products are sold and bought online versus offline?
So for that, we have to go to the top
of the diagram, which is our old friend, the long tail.
And so to just remember, by way of review,
that the long tail is an idea or concept that
has a plot of all of the products that
are available from a particular seller or a particular merchant.
The Y-axis
is always the sales of the product.
And the X-axis is all products lined up
from the most popular to the least popular.
So it just so turns out in the diapers category
the most popular brand is Pampers, followed by Huggies, then Loves.
And then there's a whole bunch of other products literally thousands
and thousands of different varieties of styles and brands and so on.
And one that I've highlighted here is called Seventh Generation.
That's a niche product
out in the tail that has lower sales than the other tree.
Seventh Generation is not available at every offline retailer.
So how do we relate this long-tail idea back to
preference isolation and see how the two things come together?
Well if we think about the online retailer, that's our
friends at diapers.com, they carry the entire distribution of products.
They offer everything, probably have the
largest assortment of baby-related products and diapers.
Probably of anyone in the world actually. Certainly bigger than any physical store.
Now if we turn to Market A, in Market A there's also pretty good variety.
Maybe not everything is there because physical stores have space constraints
and so forth, but Market A has a decent amount of variety.
Meaning that the offline sellers are quite attracted.
Market B, however, most of the sellers are just stocking
the popular brand and not really catering to a full range.
So in Market B, the amount of product available is much
more limited and is just really focused on the most popular brand.
That's why, in Market B, the customers are more prone to shop online versus offline.
50% overall, and up to 125% more online shopping when they're looking for those
niche brands that are really impossible to find in those preference-isolated markets.
[MUSIC]