|
At 50-plus years old,
direct merchant Fingerhut Co. has been around longer than most Internet
executives have been alive. The launch of Fingerhut.com in 1998 simply
added another channel to a successful catalog and telemarketing operation,
and more customer information to a database already so big it boggles
the mind.
Every transaction, every promotion received and every payment made,
every bit of personal information volunteered, every questionnaire filled
out and more are in the database: a motherlode of customer intelligence.
And in fact, Fingerhut has long mined the data for marketing guidance,
developing a reputation as a savvy database marketer in its catalog and
telemarketing channels. But what about the Internet?
“It’s all about how we use our database to put in front of customers
offers that are going to appeal to them,” says Michael Sherman, president
of Fingerhut, a division of Federated Department Stores. “It’s been part
of our business from the beginning. Now we’re just doing it in another
channel—the Internet.”
Fingerhut estimates its database is as large as seven terabytes. How
big is seven terabytes? “I can put it this way,” says Randy Erdahl, Fingerhut’s
director of business intelligence. “We have 65 million customers who’ve
bought from us at one time or another over the years, with an average
3,000 data elements on each one.”
Like most retailers collecting data about their customers, Fingerhut
has known that collecting data is good; the question is, what do you do
with it once you have it? “In a new channel like e-commerce, there’s a
lot of data that we’re stockpiling but haven’t figured out how to utilize
yet,” Erdahl says. “The software data mining tools that allow you to search
and mine the data efficiently weren’t available until recently.”
But now they are. And this year, Fingerhut, with annual sales of more
than $1.5 billion, is migrating its legacy systems into the data warehouse
it started building in 1995. It expects within the year to
have full capability not only to do database marketing and queries for
decision support for its web channel, but also to build models, score
and select customers for and execute integrated marketing campaigns across
all sales channels.
Bottom line:
$3.5 million
Working
with IBM’s horizontal marketing technology, it’s already stripped redundancy
from a cross-channel customer database grown so unwieldy that some customers
received as many as 60 catalogs a year. And it has segmented customer
groups most likely to respond to various promotions. This exercise in
database mining cut catalog mailings by 7% and added $3.5 million to the
bottom line in its first year of implementation, a combination of saved
costs and increased sales. Next up: applying those analytical capabilities
to cross sell and upsell to customer groups on the web, which the company
plans to do this year.
As Fingerhut goes, so goes e-retail. While some web merchants are out
ahead and others behind the curve, the trend is clear. Multichannel marketers
are getting a new stream of customer data from their web operations, and
they’re just beginning to turn it into web-based and integrated marketing
campaigns.
 Not,
however, without some challenges. “The joke is that everyone says they
need data, but half the time they don’t do anything with it,” says Elaine
Rubin, chairman of e-retail trade association Shop.org. “It’s not that
they don’t want to, or that they don’t realize that knowledge is power.
Everybody plans to do something with the data, but it’s a matter
of priorities. A vast number of e-retailers have been given just enough
resources at this point to get their sites up and prove a business.”
Indeed, the challenge for many is not collecting data but figuring out
the best way to use it for marketing. “The issue is not the accumulation
of data,” says Robin Green, executive vice president of the ASP business
unit at Xchange Inc., a provider of analytic software. “Retailers have
for years amassed an enormous amount of customer information as transactional
data, product registration data, direct marketing data—reams of information.
We’re seeing now that many of those retailers have not found a way to
translate that information into their online channels. Their biggest challenge
now is to convert this offline asset—this customer database—into something
that’s useful for online purposes and keep all of the information integrated
and synchronized in real time.”
A bigger window
Until
recently, retailers have been able to collect only purchasing data that
breaks down sales. But now, with the advent of the web and a whole new
generation of analytic technology, they have a potential window on a range
of customer behavior beyond buying that’s not trackable through other
channels. “If a retailer’s customer gets a catalog, he might spend a week
looking through it, circling items, showing them to other people. But
the retailer didn’t know that— all the retailer knew was what he bought,”
says Trevor Rubel, director of product strategy at Personify Inc., an
analytics software and services provider. “But online you can see what
they looked at, what they checked on the sizing charts or that they checked
the color availability.”
Rethinking
the market
For
online merchants with access to the right analytics and the smarts to
use it, there’s gold to be mined in that data. Ofoto.com is an online
photography service offering digital and film processing, prints, private
online image storing and sharing and other services and products. The
Emervyville, Calif.-based company has been monitoring its visitor activity
with Personify profiling and analytic technology since it launched in
1999, and is already mining its 1-year-old database to segment customers
and support personalization efforts.
In a recent test, for example, Ofoto identified customers who had visited
frame and card product pages relating to babies, children, weddings or
pets. The theory was that this group would be more inclined to respond
to a particular e-mail promotion than customers who didn’t visit these
pages. E-mailed to both groups, the promotion generated five times more
orders and 10 times more dollars in the targeted than in the control group.
The virtual vineyard Wine.com is another Personify client, one of its
earliest. The site was built to appeal to enthusiasts already equipped
with a fair knowledge of wine. Analysis showed that the site did attract
that customer segment and that they used features such as complex tasting
charts that rated wines on multiple scales. But it also showed a customer
segment that marketers hadn’t anticipated. Wine novices were visiting
the site and accessing features that defined terms and suggested basic
food/wine pairings.
After segmenting out food shoppers, gift shoppers, and “drop-ins” who
left the site after a few clicks, Personify’s analysis categorized a solid
18% of visitors as wine enthusiasts and 6% as wine novices. But when further
analysis compared these behavioral categories to shopping carts, it revealed
that 86% of total sales from all segments were from the wine novices,
one of the smaller customer segments. “They decided to use the information
to market different messages to different groups,” says Rubel. “To this
day, if you look at wine.com advertising online or off, you see messages
like ‘wine shouldn’t be that hard,’ or ‘wine for the rest of us.’”
With new analytics technology, the possibilities for slicing, dicing
and cooking up marketing programs out of customer data are seemingly endless.
Check out how many of your customers who bought the black sweater also
looked at but didn’t buy the gray pants, for instance, and target those
shoppers with an e-mail offer of 25% off gray pants. That’s not to say,
though, that hypersegmentation is the best marketing use of customer data,
or that it’s even practical for every e-retailer or product category.
Starting at
$500,000
For
one thing, it can be expensive. Developing later and even more sophisticated
generations of software, Personify, for example, now counts a number of
Fortune 500 companies among its clients—it’s currently scaling up an analytics
program big enough to handle customer data at L.L. Bean, for example.
And it’s charging prices to match. Software starts at about $325,000 and
a company may spend another $50,000 to $100,000 to deploy and integrate
it. “A large multichannel retailer will spend about half a million with
us on consulting support services, deployment and integration costs,”
says Rubel. Personify also offers analytics as a hosted application, with
a one-time upfront fee of about $250,000 and an annual hosting fee of
$60,000.
Another barrier is that supersegmenting customer data to drive highly
personalized marketing adds layers of complexity to marketing and IT operations
that can be difficult to manage. Amazon.com wowed the rest of e-retail
a few years ago when it applied collaborative filtering technology to
drive product recommendations to individual shoppers. But what might make
sense for books and CDs doesn’t necessarily make sense in other product
categories. Take apparel; to be specific, the classic, casual apparel
that’s the mainstay at Eddie Bauer. Though Eddie Bauer’s offering includes
a selection of trendier clothing and accessories, the customer’s choice
is often much simpler: it’s khakis with pleats or without. When the assortment
is basic, attempting to segment customers into multiple layers based on
product preferences is less apt to pay off, believes Michael Boyd, director
of customer relationship management for Eddie Bauer.
“Every time you slice the pie into smaller pieces, you have less and
less difference between the customers who end up in the different segments,”
says Boyd. “Our experience tells us that if you take a pool of customers
and segment them into two groups by whatever most differentiates them,
and talk to the two groups differently, you’re going to get a better return
on investment for that activity than you’ll get on any subsequent breaks.”
Different
strokes
For
now, drawing customer groups broadly is affording retailers a balance
between true one-to-one marketing, e-marketing’s much-discussed but little-seen
Holy Grail, and a one-size-fits all approach that doesn’t segment at all.
CDNow is developing profiles of six customer segments characterized by
such parameters as volume of past purchases and whether they’re new to
the site. The segmentation will be used to develop different approaches
online, says Amy Belew, vice president of customer service and operations.
“We might tell a customer with a long wish list that there’s a sale coming
up,” she said. “Or you might not want an e-mail from a high-volume customer
to have to wait long in a queue of e-mails from other customers.”
Eddie Bauer has for now segmented its customers into two groups, but
not, as is frequently the case, based on what they buy. Rather, it bases
the segmentation on how customers shop—whether they are convenience buyers
or like to take the time to assemble an outfit (see box).
But Eddie Bauer—or any retailer—can’t market to the segments without
the systems to make that marketing possible. And so Eddie Bauer has been
focusing on creating the single view of a customer. “For us, back end
integration has been a higher priority than sending out targeted emails
in the near term,” Boyd says. “The email marketing tactic is new enough
that most of us are still building infrastructure that lets us integrate
those contacts with the rest of our contacts. Ultimately the big win is
having the single view of all outbound communications to the customer
and being able to manage those centrally.”
Is it really
new?
Eddie
Bauer’s choice between wanting to do the marketing and making sure it
has the appropriate technology, techniques and systems in place illustrates
the dilemma that faces e-retailers as each new wave of technology hits:
is it really a leap, or merely a fancy new package for just good merchandising
practices they’ve already got covered?
The answer: it depends. Data mining and new analytics technology can
be viewed as a speedier, more powerful way to feed the kind of database
marketing direct merchants have always done. It’s already delivering bottom
line results for companies that use it to manage growing amounts of customer
information across channels. Fingerhut and others are effectively using
data mining that focuses largely on purchase history to zero in even closer
to what customers want, and to target outreach accordingly.
But industry observers say buying behavior is just the start, and an
even more robust use of the tons of data being gathered online and across
channels is yet to come. “It’s now possible, for example, to measure customers’
viral propensity,” says David Daniels, an analyst with Jupiter Media Metrix.
“If I’m a loyal customer, I’m influencing other people to buy from the
brand. There’s software that has the ability to track whether customers
send e-mails from the merchant to someone else. You can also measure behavior
that indicates the customer isn’t satisfied and work that into a retention
strategy. AOL, for example, monitors how much time people spend online
with them. When they see a customer’s use going down, the customer will
receive incentives to come back, like $10 off at one of their affiliate
merchants.”
In other words, if strip mining the data is already
producing results, digging deeper into databases could yield even more
for merchants, as they gain new ways of defining value among customer
segments and target marketing approaches to match. For many, for now,
that will wait for a day when staff and budgets catch up with what technology
makes possible. And if increasingly sophisticated data mining and analysis
can deliver the formula retailers have been looking for, it’s a day that
will likely arrive soon. “The object here is to find the right buyer,
match them with the right products and right price and the right channel.
That’s the ultimate problem of retailing, and technology today, if used
in the right way, can make huge advances in that direction,” Xchange’s
Green says. “We’re just beginning to learn how powerful it can be.”
mary@verticalwebmedia.com
It’s
not what they buy, it’s how they buy it
While
most retailers segment their customers based on what they buy—if they
bought a $1,000 lamp maybe they’d be interested in an oriental rug—Eddie
Bauer is segmenting customers based on how they shop. “Our merchandise
assortment is so much more narrow that we might only be promoting three
men’s sweaters at one time. We don’t need a complex algorithm to recommend
one of the three sweaters,” says Michael Boyd, director of customer relationship
management for Eddie Bauer.
To come up with segments based on shopping style, Eddie Bauer conducted
an exploratory data mining and analysis exercise, looking at behavioral
data that defined shopping preferences. It considered such factors as
what percentage of a shopper’s total purchases were made on mark-down,
what colors a given customer buys, and what time of day a customer is
most likely to do a transaction. All provided clues linked to larger predictive
patterns. Eddie Bauer also mined its database of attitudinal research
looking for more ways to classify customers in terms of how they approached
shopping.
Two segments emerged from the data. One, dubbed Too Busy to Shop, doesn’t
particularly enjoy shopping for apparel, is willing to pay more for a
speedy shopping experience, and likes to buy outfit solutions rather than
apparel pieces. The Professional Shopper, on the other hand, loves to
shop, is more price-aware and prides herself on the ability to assemble
a look from a variety of different sources.
“We’ve changed our focus to using all of the data we have to draw only
one conclusion: which type of shopper is this?” Boyd says. “And we develop
different versions of that set of communications.” Before the holidays,
the company tested three versions of an e-mail, including a version targeting
Too Busy to Shop, one targeting The Professional Shopper and a control
group. Tests on and off target resulted in sales 7% to 8% higher when
customers got communications targeted for their type. When they received
communications targeting the group they didn’t belong to, the message
actually depressed sales, making them even less likely to buy than the
control group.
The two-segment approach is already in use at Eddie Bauer in the form
of segmented catalog mailings. A roll-out of segmented e-mail campaigns
is likely later in the year—not, however, without additional testing and
validation of the concept.
|
