<<

. 91
( 137 .)



>>

especially if the customer paid in cash. Such anonymous transactions still have
information; however, there is clearly little opportunity for providing direct
messages to customers who have provided no contact information.
When event-based relationships predominate, companies usually commu­
nicate with prospects by broadcasting messages widely (for instance in media
advertising, free standing inserts, Web ads, and the like) rather than targeting
messages at individuals. In these cases, analytic work is very focused on prod­
uct, geography, and time, because these are three things known about cus­
tomers™ transactions.
Of course, broadcast advertising is not the only way to reach prospects.
Couponing through the mail or on the Web is another way. Pharmaceutical
companies in the United States have become adept at encouraging prospective
customers to call in to get more information”while the company gathers a bit
of information about the caller.
Data Mining throughout the Customer Life Cycle 459


Sometimes, event-based relationships imply a business-to-business rela­
tionship with an intermediary. Once again, pharmaceutical companies pro­
vide an example, since much of their marketing budget is spent on medical
providers, encouraging them to prescribe certain drugs.

Subscription-Based Relationships
Subscription-based relationships provide more natural opportunities to
understand customers. In the list given earlier, the last three examples all have
ongoing billing relationships where customers have agreed to pay for a service
over time. A subscription relationship offers the opportunity for future cash
flow (the stream of future customer payments) and many opportunities for
interacting with each customer.
For the purposes of this discussion, subscription-based relationships are
those where there is a continuous relationship with a customer over time. This
may take the form of a billing relationship, but it also might take the form of a
retailing affinity card or a registration at a Web site.
In some cases, the billing relationship is a subscription of some sort, which
leaves little room to up-sell or cross-sell. So, a customer who has subscribed to
a magazine may have little opportunity for an expanded relationship. Of
course, there is some opportunity. The magazine customer could purchase a
gift subscription or buy branded products. However, the future cash flow is
pretty much determined by the current composition of products.
In other cases, the ongoing relationship is just a beginning. A credit card
may send a bill every month; however, nothing charged, nothing owed. A
long-distance provider may charge a customer every month, but it may only
be for the monthly minimum. A cataloger sends catalogs to customers, but
most will not make a purchase. In such cases, usage stimulation is an impor­
tant part of the relationship.
Subscription-based relationships have two key events”the beginning and
end of the relationship. When these events are well defined, then survival
analysis (Chapter 12) is a good candidate for understanding the duration of
the relationship. However, sometimes defining the end of the relationship is
difficult:
A credit card relationship may end when a customer has no balance
––

and has made no transactions for a specified period of time (such as 3
months or 6 months).
A catalog relationship may end when a customer has not purchased
––

from the catalog in a specified period of time (such as 18 months).
An affinity card relationship may end when a customer has not used
––

the card for a specified period of time (such as 12 months).
460 Chapter 14


Even when the relationship is quite well understood, there may be some
tricky situations. Should the end date of the relationship be the date of cus­
tomer contact or the date the account is closed? Should customers who fail to
pay their last bill be considered the same as customers who were stopped for
nonpayment?
These situations are meant as guidelines for understanding the customer
relationship. It is worthwhile to map out the different stages of customer inter­
actions. Figure 14.4 shows different elements of customer experience for news­
paper subscription customers. These customers basically have the following
types of interactions:
Starting the subscription via some channel
––

Changing the product (weekday to 7-day, weekend to 7-day, 7-day to
––

weekday, 7-day to weekend)
Suspending delivery (typically for a vacation)
––


Complaining
––


Stopping the subscription (either voluntarily or forced)
––


In a subscription-based relationship, it is possible to understand the cus­
tomer over time, gathering all these disparate types of events into a single pic­
ture of the customer relationship.


Stop Complain
Temporarily
Some Channel
Respond from




SUBSCRIBER Voluntary
paying Churn
Stop for
Other
ill
yB
Pa Reason
Not Pay
Pay Bill




SALE ORDER START
Create Deliver
Account Paper
No
t Pa
y SUBSCRIBER Forced
late paying Churn
Stop Paying



Stop Complain
Temporarily




Figure 14.4 (Simplified) customer experience for newspaper subscribers includes several
different types of interactions.
Data Mining throughout the Customer Life Cycle 461


Business Processes Are Organized around the
Customer Life Cycle
The customer life cycle describes customers in terms of the length and depth
of their relationship. Business processes move customers from one phase of
the life cycle to the next, as shown in Figure 14.5. Looking at these business
processes is valuable, because this is precisely what businesses want to do:
make customers more valuable over time. In this section, we look at these dif­
ferent processes and the role that data mining plays in them.


Customer Acquisition
Customer acquisition is the process of attracting prospects and turning them
into customers. This is often done by advertising and word of mouth, as well
as by targeted marketing. Data mining can and does play an important role in
acquisition. Chapter 5, for instance, has an interesting example of using
expected values derived from chi-square to highlight differences in acquisition
among different regions. Such descriptive analyses can suggest best practices
to spread through different regions.
There are three important questions with regards to acquisition, which are
investigated in this section: Who are the prospects? When is a customer
acquired? What is the role of data mining?


Acquisition Activation Relationship Management Retention

Former
Customers
High
Voluntary
Value
Churn

Target New High
Responder Customer
Market Customer Potential

Forced
Low Value
Rest of Churn
World



Winback



Figure 14.5 Business processes are organized around the customer life cycle.
462 Chapter 14


Who Are the Prospects?
Understanding who prospects are is quite important because messages should
be targeted to an audience of prospects. From the perspective of data mining,
one of the challenges is using historical data when the prospect base changes.
Here are three typical reasons why care must be used when doing prospecting:
Geographic expansion brings in prospects, who may or may not be sim­
––

ilar to customers in the original areas.
Changes to products, services, and pricing may bring in different target
––

audiences.
Competition may change the prospecting mix.
––




Y
These are the types of situations that bring up the question: Will the past be
a good predictor of the future? In most cases, the answer is “yes,” but the past




FL
has to be used intelligently.
The following story is an example of the care that needs to be taken. One
AM
company in the New York area had a large customer base in Manhattan and
was looking to expand into the suburbs. They had done direct mail campaigns
focused on Manhattan, and built a model set derived from responders to these
campaigns. What is important for this story is that Manhattan has a high con­
TE

centration of very expensive neighborhoods, so the model set was biased
toward the wealthy. That is, both the responders and nonresponders were
much wealthier than the average inhabitant of the New York area.
When the model was extended to areas outside Manhattan, what areas did
the model choose? It chose a handful of the wealthiest neighborhoods in the
surrounding areas, because these areas looked most like the historical respon­
ders in Manhattan. Although there were good prospects in these areas, the
model missed many other pockets of potential customers. By the way, these
other pockets were discovered through the use of control groups in the
mailing”essentially a random sampling of names from surrounding areas.
Some areas in the control groups had quite high response rates; these were
wealthy areas, but not as wealthy as the Manhattan neighborhoods used to
build the model.

WA R N I N G Be careful when extending response models from one
geographic area to another. The results may tell you more about similar

geographies than about response.




When Is a Customer Acquired?
There is usually an underlying process in the acquisition of customers; the
details of the process depend on the particular industry, but there are some
general steps:
Team-Fly®
Data Mining throughout the Customer Life Cycle 463


Customers respond in some way and on some date. This is the “sale”
––

date.
In an account-based relationship, the account is created. This is the
––

“account open date.”
The account is used in some fashion.
––

<<

. 91
( 137 .)



>>