<<

. 12
( 137 .)



>>

aggregations,
information
views




Historical
Data whose
format and
content change Transaction
over time Data with
missing and
incomplete
fields




Data from multiple
competing sources



Data Mart

Marketing Summaries
Operational System
Figure 2.2 Data is never clean. It comes in many forms, from many sources both internal
and external.


A wireless telecommunications company once wanted to put together a
data mining group after they had already acquired a powerful server and a
data mining software package. At this late stage, they contacted Data Miners
to help them investigate data mining opportunities. In the process, we learned
that a key factor for churn was overcalls: new customers making too many
calls during their first month. Customers would learn about the excess usage
when the first bill arrived, sometime during the middle of the second month.
By that time, the customers had run up more large bills and were even more
unhappy. Unfortunately, the customer service group also had to wait for the
same billing cycle to detect the excess usage. There was no lead time to be
proactive.
However, the nascent data mining group had resources and had identified
appropriate data feeds. With some relatively simple programming, it was
30 Chapter 2


possible to identify these customers within days of their first overcall. With
this information, the customer service center could contact at-risk customers
and move them onto appropriate billing plans even before the first bill went
out. This simple system was a big win for data mining, simply because having
a data mining group”with the skills, hardware, software, and access”was
the enabling factor for putting together this triggering system.


Take Action
Taking action is the purpose of the virtuous cycle of data mining. As already
mentioned, action can take many forms. Data mining makes business deci­
sions more informed. Over time, we expect that better-informed decisions lead
to better results.
Actions are usually going to be in line with what the business is doing
anyway:
Sending messages to customers and prospects via direct mail, email,
––

telemarketing, and so on; with data mining, different messages may go
to different people
Prioritizing customer service
––

Adjusting inventory levels
––

And so on
––


The results of data mining need to feed into business processes that touch
customers and affect the customer relationship.


Measuring Results
The importance of measuring results has already been highlighted. Despite its
importance, it is the stage in the virtuous cycle most likely to be overlooked.
Even though the value of measurement and continuous improvement is
widely acknowledged, it is usually given less attention than it deserves. How
many business cases are implemented, with no one going back to see how well
reality matched the plans? Individuals improve their own efforts by compar­
ing and learning, by asking questions about why plans match or do not match
what really happened, by being willing to learn that earlier assumptions were
wrong. What works for individuals also works for organizations.
The time to start thinking about measurement is at the beginning when
identifying the business problem. How can results be measured? A company
that sends out coupons to encourage sales of their products will no doubt mea­
sure the coupon redemption rate. However, coupon-redeemers may have pur­
chased the product anyway. Another appropriate measure is increased sales in
The Virtuous Cycle of Data Mining 31


particular stores or regions, increases that can be tied to the particular market­
ing effort. Such measurements may be difficult to make, because they require
more detailed sales information. However, if the goal is to increase sales, there
needs to be a way to measure this directly. Otherwise, marketing efforts may
be all “sound and fury, signifying nothing.”
Standard reports, which may arrive months after interventions have occurred,
contain summaries. Marketing managers may not have the technical skills to
glean important findings from such reports, even if the information is there.
Understanding the impact on customer retention, means tracking old market­
ing efforts for even longer periods of time. Well-designed Online Analytic Pro­
cessing (OLAP) applications, discussed in Chapter 15, can be a big help for
marketing groups and marketing analysts. However, for some questions, the
most detailed level is needed.
It is a good idea to think of every data mining effort as a small business case.
Comparing expectations to actual results makes it possible to recognize
promising opportunities to exploit on the next round of the virtuous cycle. We
are often too busy tackling the next problem to devote energy to measuring the
success of current efforts. This is a mistake. Every data mining effort, whether
successful or not, has lessons that can be applied to future efforts. The question
is what to measure and how to approach the measurement so it provides the
best input for future use.
As an example, let™s start with what to measure for a targeted acquisition
campaign. The canonical measurement is the response rate: How many people
targeted by the campaign actually responded? This leaves a lot of information
lying on the table. For an acquisition effort, some examples of questions that
have future value are:
Did this campaign reach and bring in profitable customers?
––

Were these customers retained as well as would be expected?
––

What are the characteristics of the most loyal customers reached by this
––

campaign? Demographic profiles of known customers can be applied to
future prospective customers. In some circumstances, such profiles
should be limited to those characteristics that can be provided by an
external source so the results from the data mining analysis can be
applied purchased lists.
Do these customers purchase additional products? Can the different
––

systems in an organization detect if one customer purchases multiple
products?
Did some messages or offers work better than others?

––

Did customers reached by the campaign respond through alternate

––

channels?

32 Chapter 2


All of these measurements provide information for making more informed
decisions in the future. Data mining is about connecting the past”through
learning”to future actions.
One particular measurement is lifetime customer value. As its name implies, this
is an estimate of the value of a customer during the entire course of his or her rela­
tionship. In some industries, quite complicated models have been developed to
estimate lifetime customer value. Even without sophisticated models, shorter-
term estimates, such as value after 1 month, 6 months, and 1 year, can prove to be
quite useful. Customer value is discussed in more detail in Chapter 4.


Data Mining in the Context of the Virtuous Cycle




Y
A typical large regional telephone company in the United States has millions




FL
of customers. It owns hundreds or thousands of switches located in central
offices, which are typically in several states in multiple time zones. Each
switch can handle thousands of calls simultaneously”including advanced
AM
features such as call waiting, conference calling, call-forwarding, voice mail,
and digital services. Switches, among the most complex computing devices
yet developed, are available from a handful of manufacturers. A typical tele­
TE

phone company has multiple versions of several switches from each of the
vendors. Each of these switches provides volumes of data in its own format on
every call and attempted call”volumes measured in tens of gigabytes each
day. In addition, each state has its own regulations affecting the industry, not
to mention federal laws and regulations that are subject to rather frequent
changes. And, to add to the confusion, the company offers thousands of dif­
ferent billing plans to its customers, which range from occasional residential
users to Fortune 100 corporations.
How does this company”or any similar large corporation”manage its
billing process, the bread and butter of its business, responsible for the major­
ity of its revenue? The answer is simple: Very carefully! Companies have
developed detailed processes for handling standard operations; they have
policies and procedures. These processes are robust. Bills go out to customers,
even when the business reorganizes, even when database administrators are
on vacation, even when computers are temporarily down, even as laws and
regulations change, and switches are upgraded. If an organization can manage
a process as complicated as getting accurate bills out every month to millions
of residential, business, and government customers, surely incorporating data
mining into decision processes should be fairly easy. Is this the case?
Large corporations have decades of experience developing and implement­
ing mission-critical applications for running their business. Data mining is dif­
ferent from the typical operational system (see Table 2.1). The skills needed for
running a successful operational system do not necessarily lead to successful
data mining efforts.

Team-Fly®
The Virtuous Cycle of Data Mining 33


Table 2.1 Data Mining Differs from Typical Operational Business Processes

TYPICAL OPERATIONAL SYSTEM DATA MINING SYSTEM

Operations and reports on Analysis on historical data often
historical data applied to most current data to
determine future actions

Predictable and periodic flow of Unpredictable flow of work
work, typically tied to calendar depending on business and
marketing needs

Limited use of enterprise-wide data The more data, the better the results
(generally)

Focus on line of business (such as Focus on actionable entity, such as
account, region, product code, minutes product, customer, sales region
of use, and so on), not on customer

Response times often measured in Iterative processes with response
seconds/milliseconds (for interactive times often measured in minutes or
systems) while waiting weeks/months hours
for reports

System of record for data Copy of data

Descriptive and repetitive Creative


First, problems being addressed by data mining differ from operational
problems”a data mining system does not seek to replicate previous results exactly.
In fact, replication of previous efforts can lead to disastrous results. It may
result in marketing campaigns that market to the same people over and over.
You do not want to learn from analyzing data that a large cluster of customers
fits the profile of the customers contacted in some previous campaign. Data
mining processes need to take such issues into account, unlike typical opera­
tional systems that want to reproduce the same results over and over”
whether completing a telephone call, sending a bill, authorizing a credit
purchase, tracking inventory, or other countless daily operations.
Data mining is a creative process. Data contains many obvious correlations
that are either useless or simply represent current business policies. For exam­
ple, analysis of data from one large retailer revealed that people who buy
maintenance contracts are also very likely to buy large household appliances.
Unless the retailer wanted to analyze the effectiveness of sales of maintenance
contracts with appliances, such information is worse than useless”the main­
tenance contracts in question are only sold with large appliances. Spending
millions of dollars on hardware, software, and analysts to find such results is a
waste of resources that can better be applied elsewhere in the business. Ana­
lysts need to understand what is of value to the business and how to arrange
the data to bring out the nuggets.
34 Chapter 2

<<

. 12
( 137 .)



>>