GET INSIDE YOUR CUSTOMERS' HEADS
by Kayte VanScoy
Your company's most valuable asset is sitting right under your nose. It's
not your inventory or your brand, your trade secrets or your newest product
line. The key to coming out on top in the new economy is hidden in your data.
From daily receipts to banner ad click-throughs, powerful new software and
services can take advantage of the one natural resource every business has:
information.
This process of putting your customer information to work is called data
mining, and it's powering the shift toward personalized customer service
across every industry. Forget broad demographics; the name of the game is
winning and keeping individual customers. You need to know their favorite
colors and whether they prefer paper or plastic just as surely as you know
what was on their most recent order. Why? So you can read their minds,
anticipating what they want and turning them into loyal customers.
Sound complicated? You'll be surprised at how much you already know. But
successful data mining means walking the fine line between serving your
customers and encroaching on their privacy. We'll show you how to do it right.
Mining for Gold
"It used to be that the most important thing you could make was a product,
and you went looking for customers for that product. In the Information Age,
the most important thing you make is a customer. Then you find products for
the customers you have," says Martha Rogers, cofounder of Peppers and Rogers
Group, a Stamford, Connecticut based consulting firm that specializes in
customer relationship management.
The best example of this strategy comes from Amazon.com, which quickly moved
from selling books to offering CDs, toys, videos, electronics, and even
auctions. What Amazon is selling became less important than the fact that
Amazon is the one selling it. Customers who are already comfortable buying
there are likely to stick around and buy more. "The emphasis is shifting from
who can sell more products to who can sell a customer more products over their
lifetime," Rogers says.
The secret to Amazon's success, and that of offline companies like
Blockbuster Video and American Airlines, is an old technology exploited in a
new marketplace. Data mining used to mean dusty, academic algorithms that were
too arcane for the world of business. But with technology and automation
sweeping into every business sector, the amount of data available to
individual companies has exploded. Add to that the Internet's ability to
process and present data in real time, and the result is the ability to create
instantaneous, personalized product offerings. And the future holds more of
the same: "A lot of the ways that we think about e-commerce today are
transactional, but obviously this isn't about transaction marketing anymore.
It's about relationship marketing," says Adam Gross, cofounder and director of
technology for Personify, a data-mining software company.
Direct mail firms and catalog retailers have long relied on targeted
marketing, using early data-mining solutions to track repeat customers'
transactions by way of catalog and coupon codes. At the heart of those
campaigns were highly specialized knowledge workers manipulating complex
software and monolithic hardware. Today's data-mining tools run on desktop PCs
and are easy enough for a call-center operator, regional marketing manager, or
chief technology officer to use. The range of data-mining products available
is broadfrom complete inventory-management packages to services that send out
personalized broadcasts to handheld PCs.
For example, say you know a customer's wedding anniversary is July 12, info
you gathered from an online survey. Before you can turn that fact into a sale,
it may go through a half-dozen software packages: from a database like Oracle
or Informix, to a data-analysis package like the SAS Institute's Enterprise
Miner to a customer relationship management (CRM) package like E.piphany E.4,
and finally into a Web content management platform like Vignette StoryServer.
Once it makes its way there, it could automatically launch a greeting card and
sales promotion that's sent by regular mail, e-mail, pager, or even a
personalized greeting on your home page. It could also trigger banner ads from
other companies where people celebrating anniversaries might want to shop.
(For more details, see "How It Works: Data Mining," below.)
Most data-mining software claims to address every one of these steps, but in
fact most businesses rely on several solutions working together. "At any
particular customer location, there are a minimum of five products coming
together to create the application that you're using on the Web," says Erik
Josowitz, vice president of product management for Vignette.
Know your Options
The first step in choosing the right data-mining solution is figuring out
what kind of data you want to work with and what you hope to accomplish.
"There has been an exponential jump in data collected by businesses because
there are so many new ways to collect it," says Micro Strategy's analyst
relations manager Ian Walsh. E-commerce and Web server information is the most
obvious. But also look at the data you collect in other areas of your
business: records of transactions made at your call center, sales receipts and
customer information from point-of-sale transactions, and mailing lists of
your company's catalogs.
If your company offers a shopping card that customers use to receive
discounts, like those used in many grocery stores, you have another valuable
asset. These cards let you track data about purchasing patterns. Also consider
buying third-party information to help you more thoroughly analyze your own
data. Buy a list of potential customers with good credit from a credit bureau.
Purchase weather data to help you make sense of trends in the regions where
you do business for instance, customers in Texas purchase outerwear when the
temperature dips below 55 degrees. Partner with a company that has a
complementary businessan auto-repair shop located near your car dealership,
for example and get access to its customer data. Then feed it all to your
data-mining software.
Sometimes the most useful data is ignored or even abandoned. "We've found
that catalog companies often held only six months of data, and they were
missing all the seasonal information," says Anne Milley, global marketing
strategist for data mining at SAS. "They didn't know who shopped with them
only at Christmas, for example. Seasonality is very important to your
business."
Looking at data in a different way is how Victoria's Secret solved an on
going underwear shortage in its Southwest region. It used MicroStrategy
software to discover that customers in the Northeast liked brightly colored
underwear while Southwesterners preferred neutral colors. But what Victoria's
Secret learned also underscores the importance of being realistic about what
you can do with data mining. The company hasn't integrated data from its Web
site with its call center or retail stores. "It's not a technology issue any
longer. It's more of a return on investment issue," says Marc Reifeis,
director of data warehousing for Victoria's Secret's parent company, The
Limited, explaining that such an undertaking would run into the multimillions
of dollars. (For more on how Victoria's Secret uses data mining, see "Smart
Return on Investment" below.)
Even new businesses, especially e-businesses that haven't collected much
information yet, can leverage data mining. Purchasing the right list of
prospective customers makes more sense than taking the time to build up your
own list over time. That's what More.com did, quickly becoming one of the top
pharmacies on the Web. "There are ways to acquire customers other than brand
advertisement," says Peppers and Rogers' Martha Rogers.
Getting Started
Personalization is really the end game for data mining. "I start with what I
know about this customer and figure out what the customer needs from me next,"
says Rogers. "Ideally, there is a single person whose job it is to get more
business from this customer."
Gathering more information isn't always a straightforward process. You could
ask the customers to tell you more about themselves, but that can backfire.
"People are very sensitive to your collecting information about them, so focus
on how the relationship builds up over time," says Vignette's Josowitz. "Don't
immediately jump in with, 'Fill out this form.'"
Instead, start with what you already know. When visitors come to your Web
site, you can tell what types of computers they are using, what browsers they
have, and who their Internet service providers are. Also look at batch data,
which is collected across broad demographicsin the early stages it can make
the difference between simply getting to know your customer and actually
making the sale.
It works: Amazon uses shopping-cart analysis to predict a customer's next
purchase based on the purchases of others with similar tastes. Not only can
Amazon predict a purchase, it can nearly guarantee it when it offers the CD or
book at a 20 percent discount. It sounds easy, but Rogers warns that it won't
work if your company doesn't change its customer approach. "Until you shift
the focus from the product to the customer, essentially what you've got is
better targeted harassment, not customer relations," she says.
Personify's Gross offers another word of advice for coming up with your
datamining plan: linking. "Make sure that your data is linked across channels,
meaning that you have a consistent way of identifying users. It could be an email
or a home address, or for Web businesses it could be a cookie," he says.
And sometimes you can have too much of a good thing. Don't blindly collect
data if you can't use it; you'll risk alienating customers. "It's critical
that there be a very clear payback for the customer," says John Samuels, vice
president of interactive marketing for American Airlines. "Don't ask questions
until you know how the information is going to be used and make sure there's
plenty of payoff. People will wonder: 'I told them about x, and what am I
getting for that?' " (For more on data mining in action at American, see "SkyHigh
Customer Loyalty" below.)
Choose Your Solution
Many companies tend to buy first and ask questions later when it comes to
the latest technology. You need to look closely before you jump. Ideally, you
should ask the software providers for other customer success stories so you
can see if the solution would work for you in a similar way. But getting that
information is often tricky: "The thing about data mining is that many
companies don't want people to know that they're using our software, let alone
how they're using it," says SAS's Milley.
Some data-mining products run on stand-alone computers, such as the high-end
packages from SAS. "There is a need to be able to respond very quickly in this
Internet age," Milley says, "but to have an understanding of past trends takes
looking at the data offline." Other solutions, like those from Personify, are
completely Web-based and focus on personalization.
Personify's Gross counters, "We built our product on the tradition of
catalog marketers, who were always relationship oriented." Other solutions,
like those from MicroStrategy, bank on the future and integrate with wireless
devices like the Palm VII and text pagers. For example, MicroStrategy's
software lets you page customers to let them know the video they wanted to
rent is in the store or their stock has dropped. (For more details on the top
data-mining solutions, see "Building Blocks" below.)
Price is another big consideration. Packages start at $50,000, but the
average price for a complete system is $500,000 and sometimes reaches into
seven figures. "This isn't just throwing money into the wind," Gross says.
"This is real, high-precision, analytical marketing. It's a different type of
marketing than classic branding, though."
For an investment that large, you'll want to make sure you've got the right
people on staff to manage it. "A couple of years ago," says Vignette's
Josowitz, "we were talking to a marketing guy who normally had no idea what it
took technically. Then we'd talk to an IT guy who could understand the
infrastructure involved but had no sense of the marketing and business
challenges. Today most of the organizations have anointed someone as the
leader of e-business."
Watch Your Step
So apart from the expense, what's the catch? Just ask Internet advertising
powerhouse DoubleClick, which ended up at the center of a controversy over
consumer privacy earlier this year. (For more details, see "First-Mover
Disadvantage," below.) Privacy watchdog groups and the Federal Trade
Commission are keeping a close watch on data-mining technology and are ready
to pounce on any perceived abuses. In the absence of regulations or even
guidelines, it is up to companies to find a balance between spooking customers
with Big Brother technologies and winning them with promises of convenience.
"Bribe them!" is the simple solution Rogers offersand many businesses seem
to agree. SAS's Milley remembers a time when her father got a discount on the
family's health insurance premiums if she maintained a high grade point
average in college. Her dad saved a little money, and the insurance firm
collected valuable customer data, such as the college Milley attended, which
classes she took, and how well she fared. Providing that information might
have seemed safe enough when data was stored on magnetic tape and accounts
were managed by warm bodies, but the Internet's speed and accessibility now
makes many consumers wary.
"Our privacy laws and principles are based on the concept of a record that
is retrievable only by a person's name. New technologies may mean that all
those principles have to be re-examined," says Robert Ellis Smith, publisher
of Privacy Journal. "Every privacy policy on the Net begins with the line,
'We're sensitive.' ... That's hollow. Just being sensitive is not enough
anymore."
Even if companies don't heed Smith and his privacy watchdog cohorts, the
stock market certainly will. Witness DoubleClick's plunge after its privacy
flap and Intel's dive after a threatened boycott over the tracking capability
of its Pentium III chip. Don't expect your data-mining software to make it any
easier either. "Privacy is a policy that each of our client companies needs to
address," says Paul Rodwick, vice president of product marketing and strategy
for E.piphany. He echoes the sentiment of other data-mining companies.
The international privacy group Junkbusters offers some guidelines for fair
information practices that sites like Dash.com and CDnow follow. First, give
people access to information about themselves and allow them to modify it if
they wish, says Junkbusters founder Jason Catlett, who has a Ph.D. in computer
science. Second, use the information only for the purposes you specify. Third,
keep the data secure from unauthorized use. Catlett and others stress the
importance of moving to an "opt in" model, in which consumers must consent to
having their data mined a practice the Federal Trade Commission may soon
mandate.
Despite the controversies, though, your business's success depends upon data
mining. "In the future, business channels are going to resemble the Web more
than they do traditional channels. Data mining will apply to television, cell
phones, kiosks, direct advertising, you name it," says Personify's Gross.
In the immediate future, data mining will allow businesses to become more
integrated. "What we hope to see is, when you're stuck on a Web site you call
the phone number and the operator will have your record in front of him. Or if
I'm shopping on a site, the software will remember those things in my shopping
cart and allow me to move from my PC to my wireless phone to complete the
transaction," says Vignette's Josowitz.
Najib Abboud, associate professor of computational mechanics at New York's
City College, goes one step further: "A real person will pop up on your screen
and say, 'Can I help you figure out what is the best product for you?' They're
not going to have to say, 'What can I interest you in today?' That person will
have available to them your immediate track record. They'll already have a
scope, a context, of what you're interested in or what you're trying to find
out about."
First-Mover Disadvantage
Among the fairy-tale success stories of the new economy, DoubleClick is
destined to become the cautionary tale, giving nightmares to e-commerce CEOs
and NASDAQ investors alike. Once the darling of Wall Street, DoubleClick saw
its stock drop 30 percent in February following the New York based Internet
advertising firm's purchase of the offline direct mail giant Abacus Direct.
Both companies are built on foundations of mined data, but with two primary
differences: DoubleClick collects its data when customers surf one of its
11,000 client Web sites and keeps that data completely anonymous. Abacus
compiles its information from purchasing histories of department stores and
catalog marketers and attaches a name and street address to each of its 88
million listings.
From DoubleClick's perspective, it was a match made in heaven, but privacy
watchdogs and the Federal Trade Commission immediately protested DoubleClick's
newfound ability to tie consumers' names to their Web surfing habits. In the
end, DoubleClick backed down and agreed to postpone the launch of Abacus
Online, which would combine and then sell the data of the two merged
companies. "The mistake Double Click made was getting out ahead of the
industry in trying to link up names in the absence of clear, agreed-upon
guidelines from industry and government," admits Jonathan Shapiro, vice
president of DoubleClick and head of Abacus Online. "Being the leader,
sometimes you step out too far, and you set yourself up."
How It Works: Data Mining
Data mining is a process that allows computers to look for patterns in
information. It helps companies make sense of data they collect about
customersto offer personalized service, anticipate consumer behavior, uncover
trends, and spot the unexpected. Here's how it works.
- Information Capture -- A customer fills out a survey while surfing your Web
site, which specializes in flowers and gifts. The survey information
including spouse's birthday and computer's IP addressis stored inside an
Oracle database along with data from other respondents.
- Data Routing -- The administrator of the database, a real person, reviews
the data and tags information to be sent to a data-analysis program like
that from SAS.
- Data Analysis -- Once there, the program looks for patterns and statistical
information. It discovers that 10 percent of its customers' spouses have
birthdays in the current month. It also sees that for three years in a row,
a particular customer has bought flowers during the week of his wife's
birthday.
- Rules Creation -- That data then goes to a customer relationship management
(CRM) package, such as E.piphany, which creates rules from those patterns,
such as, "If people normally buy flowers in June, then send them a coupon
in May."
- Rules Application -- A marketing analyst (a real person) puts the rules in
a content management platform like Vignette StoryServer. The system creates
a 10 percent discount offer for a floral arrangement and e-mails it to the
customer. It also displays gift-related banner ads when the customer
returns to the Web site.
- Customer Notification -- Since the customer has signed up for wireless
alerts,his cell phone receives a text page reminding him of his wife's
birthday and suggesting several gifts he can buy.
Building Blocks
Setting up a data-mining solution isn't a matter of buying just one product.
Instead, you'll most likely use a handful of specialty applications that work
together to give you the big picture. Here are five areas to consider.
Database
This is the heart of your data-mining system, where your raw data is stored
and extracted. Choose products from one of the following vendors:
- Brio (www.brio.com)
- IBM (www-4.ibm.com/software/data/db2)
- Microsoft (www.microsoft.com/sql/?RLD=183)
- MicroStrategy (www.microstrategy.com)
- Oracle (www.oracle.com)
Knowledge-Discovery Tools
Use these to find patterns in your data and suggest rules you can act on.
Select one of the top players:
- HNC Software (www.hnc.com)
- SAS Institute (www.sas.com)
- SPSS (www.spss.com)
Customer Relationship Management
Apply the rules from your knowledge discovery to your customer software.
Here are the best bets:
- Broadbase (www.broadbase.com)
- E.piphany (www.epiphany.com)
- Personify (www.personify.com)
- Pivotal (www.pivotal.com)
- Remedy (www.remedy.com)
- Siebel (www.siebel.com)
- WebAssociates (www.webassociates.com)
Content Management
This software typically runs your Web site and will create the personalized
content for your individual customers. Here are the top three:
- Art Technology Group (www.atg.com)
- BroadVision (www.broadvision.com)
- Vignette (www.vignette.com)
Wireless Solutions
If you want to communicate with customers over wireless devices, such as
cell phones or handheld PCs, you'll need applications from one of these
companies:
- IBM (www-4.ibm.com/software/data/db2)
- MicroStrategy (www.microstrategy.com)
Sky-High Customer Loyalty
American Airlines' Web site www.aa.com offers its AAdvantage frequent
fliers something no other site in the crowded travel-portals market can:
individualized airfares and service. "The AAdvantage program provides a rich
source of customer information to begin with," says John Samuels, American's
vice president of interactive marketing. On top of that data, American
overlays an online survey that asks everything from members' favorite
destinations to their children's school district. In return for the
information, AAdvantage members get e-mail messages with customized offers and
flight discounts to their preferred destinations. American relies on three
products to pull it off: Broadbase Software for data mining, E.piphany for
applying business and marketing rules, and BroadVision for content management.
Before American implemented its personalization campaign, the airline's
NetSAAver mailing list, which simply sent out a list of that week's discounted
fares, had roughly 100,000 subscribers. After the site launched the
personalization features in late 1998, that number steadily rose to 2 million
of the approximately 11 million active AAdvantage members.
Smart Return on Investment
Sometimes mining all of your available data just isn't possible. But that's
where savvy companies like Victoria's Secret know how to pick their marks. The
company, which began as a catalog retailer, has long appreciated the value of
its data. It maintains two databases: a product-focused one, which details
purchasing demographics and inventory, and a more customer-centric one, which
tracks individual purchasing histories. Victoria's Secret manages its
inventory with MicroStrategy software that combines data from the point of
sale, financial systems, inventory systems, and more.
"The question is, how can you leverage and act on this information?" says
Marc Reifeis, director of data warehousing for Victoria's parent company, The
Limited. Victoria's Secret's Web site www.victoriassecret.com tracks
click-throughs and makes offers based on previous purchases, but surprisingly,
the company doesn't overlap its two rich data storehouses.
"When a new location opens up, we leverage the names and addresses to send
out coupons to people in the area," says Reifeis, but he adds that a fully
integrated call center/Web site/storefront system is not yet cost effective.
Reifeis questions the usefulness of further extending into wireless solutions
when the technology is not yet widespread. He says he's less concerned about
being beaten to the punch by a competitor than about his company's bottom
line. "Amazon is a terrific case study, but I don't feel that we all play with
the same set of rules. They don't have to make a profit. We return real value
to our shareholders."
Data Mining 101
Ready to put data mining to work for your business? Follow these guidelines
to get up and running quickly.
Target Your Approach
Decide what process you want to improve in your business: marketing,
customer service, sales, or product development.
Build Your Team
Put together a data-mining team that includes the key people from the
following areas of your business: management, IT, marketing, and call center
workers.
Choose Your Project
Begin with a targeted goal that has a high return on investment, such as a
specific sales promotion.
Get The Right Data
As you create your data-mining model, work with at least 12 months of
detailed data, not just data summaries.
Put The Plan Into Practice
Once you've made your plan, implement it as soon as possible. Data-mining
models lose value over time some even in a matter of weeks, because of
external factors and the changing nature of the information.
|
