Last Updated Oct 13, 2009 3:51 PM EDT
David Bell, associate professor of marketing at University of Pennsylvania's Wharton School.
Plenty of marketers pull mountains of data from the Web, such as
search habits and online surveys, to figure out how to reach new customers. The
assumption is that many of the retail rules that apply to physical stores don't
apply to Internet commerce, or at least not in the same way. href="http://marketing.wharton.upenn.edu/people/faculty.cfm?id=227">David Bell, an associate
professor of marketing at University of Pennsylvania's Wharton
School, says that's not so. In fact, his research shows that any
Internet retailer looking for customers would be wise to use some pre-Web
We all understand that people who live in the same community shop at the
same physical stores. But you’ve actually found that people in the
same neighborhoods also tend to shop at the same online stores?
Yeah, we call it contagion, or neighborhood effects. We
looked at Netgrocer.com and built a statistical model. It turns out that
customers of Internet retailers are not scattered randomly on the map. So even
when your business is on the Web, you’re going to get more new
customers in areas that are contiguous to areas where you already have
Why would this happen when people are shopping in private, from their
There are two reasons. One is people may talk to each other.
You and I are having lunch and I just happened to buy my groceries on Netgrocer.com and somehow it comes up in conversation. Or, we live in the same apartment
building, you come home at night, the doorman’s got a box waiting
there that says Netgrocer.com and somehow that cues you.
So a business could target a specific neighborhood, or zip code, because it
wants word to spread in a certain demographic?
Absolutely. Knowing this can help you seed the market and
target your marketing. You can take old economy data, like census data, and
look at who’s living in a neighborhood, for example, or how many
stores are there. And it is very informative for how well your Internet store
is going to do. ... It also helps to have some buzz-worthy benefit, such as
fast and free shipping.
I trust you’d advise companies to combine this data with more
We worked with Diapers.com and they asked people to self-identify
when they first sign up at the Web site: “Did you hear about us
because of word of mouth, search advertising, magazine?” That’s
really valuable information.
If this social contagion helps online stores, wouldn’t it always
make sense for Internet stores to target densely populated areas?
Not necessarily. There’s the concept of a
preference minority. Another Wharton professor, Joel Waldfogel, wrote a book called,
The Tyranny of the Market:
Why You Can’t Always Get What You Want. It starts off with Milton
Friedman’s line about the best thing about capitalism is that any man
can have any tie in any color that he wants.
That sounds like a very appealing statement, but it turns
out not to be true if there are high fixed costs to providing stuff. So let’s
imagine you like red ties and you live in a neighborhood of 100 people but the
other 99 people like blue ties. The tie-making guy’s going to say, ‘Well,
screw it, I’m only going to make blue ties, and so you’re
out of luck.’ And so your preference is in some sense hurt by the
majority preferences of other people.
So the Internet business can figure out how to try to reach those
minorities — the so-called Long Tail customers?
Exactly. Say there are two neighborhoods, both with 100
people that have young kids. In some sense, they both need the same amount of
diapers. But if in one neighborhood, most of the people are old, the local
stores are going to pay less attention to diapers. What you have is two
markets with the same total potential, but the second market, where your
customers are more of a preference minority, is the market where you’re
actually going to get a much higher level of sales.
How do you figure out the preference minorities in what locations?
Well, with diapers it’s not that difficult because
you can go to the census and you’ve got this very good information
about each neighborhood regarding how many people live there, how many people
have kids under four years of age. And so you come up with pretty good proxies
for the baby population.
The relevant size of your target population is also a good
proxy for how well the local offline competitors are going to treat those
people. And an Internet firm can come up with a measure for each location of how
many potential customers are there, and how many are majority preference
people, and how many are minority preference people. What you learn is that the
people who are in the middle of nowhere might be your best customers.
For all the great data that companies are getting from Web activity, it
sounds like you’ve discovered there’s a ton of value in
I think the use of the offline data is really important. The
customers for Internet businesses are, of course, in unlimited locations, and
this data can help you be much more efficient at targeting. You look at the
numbers, but you also have to understand that the way that you acquire
customers also varies a lot by location, whether you’re getting them
from word of mouth or organic search or advertising. So maybe in San Francisco
you can get something going by word of mouth, and it’s going to be
fantastic. But in Philadelphia, it just won’t work.
More on BNET: