DECISION TECHNOLOGIES IN DATABASE MARKETING: PART I
by Gene M. Ferruzza

DEFINING THE SCOPE AND APPLICATIONS OF DATA MINING

INTRODUCTION

In the 1990s, database marketing came into its own, bringing innovations to business that were not possible at the advent of the computer revolution. With the continuing advances in database technology, our ability to collect and store massive amounts of data from daily business processes and "crunch" these data to derive new knowledge -- that is, to "mine" data for new information -- has become a mainstay of strategic business initiatives.

In many business operations, data and derived information specifically related to individual customers form the foundation of database marketing programs or campaigns. Our ability to leverage customer data to improve marketing decision processes is becoming more sophisticated. The competitive edge businesses gain through successful database marketing is providing the momentum for continual advancement in database marketing techniques and technologies.

Database marketing programs most commonly are designed to increase customer retention and cross-sell ratios, calculate and provide an understanding of customer value, and increase acquisition of high-value customers. More sophisticated applications integrate the marketing function with other business applications, such as customer service and product development. For companies with millions of customers, it is difficult to understand the needs or the value of an individual customer or to predict that customer's future behavior. Such understanding of individual customers is the core of database marketing. Determining the scope of a strategic database marketing initiative requires a complex assessment of customer behavior.

A database marketing initiative that bypasses this in-depth assessment will likely serve as an expensive lesson, yielding disappointing results. My goal in this paper is to outline the critical uses of customer data in database marketing programs and define the scope and applications of data mining, from the perspective of executive management. Through a review of data mining techniques and complementary technologies in database marketing, I outline a definition of data mining that many will find surprisingly broad.

WHY DEFINE DATA MINING?

Current usage of the term "data mining" is highly fluid and has as many variations as there are individuals working in information technology. In fact, the term currently is not specific to a product, technology, methodology, or practice.

Data mining is loosely defined in part because it is relatively new. We have been able to mine data only since we have been able to store it and operate on it with computers -- in other words, within the last 20 years or so. Realistically, it has been only about 15 years since computer power has been placed in the hands of most business users, and algorithms to derive new knowledge from data have been converted to affordable, commercially available software. Since that time, forecasting and predictive modeling techniques for business environments have become steadily more advanced and sophisticated. And finally, only in about the last 5 years has the use of data-driven decision technologies in marketing become widespread.

The looseness with which the term "data mining" is used leaves a high degree of flexibility for data-mining software product providers and service and consulting companies in describing what they have to offer. Although I believe that most data-mining providers have been accurately representing their services and products, the users of these services and products often are confused by competing claims. When faced with the need to acquire data-mining products or services, the user often wonders, What actually are "best practices" in data mining? Is data mining a product or service that can be purchased? Is it a process or methodology that can be learned? Is it an art that should be practiced?

Classically, data mining has been defined as a process consisting of data analysis and the use of algorithms to extract patterns in data or identify similar patterns in data. In database marketing, the term has been used narrowly to refer to particular advanced modeling technologies (especially for non-parametric modeling, discussed later). At this critical point, as the role of data mining in information processing is rapidly expanding, I believe an expanded working definition of data mining is needed. This is a time when the cookbooks are being written, new approaches and technologies appear at a regular frequency, and the use of data mining in business applications is still undefined and immature.

Furthermore, a better definition of data mining will help ensure that data miners deliver what users really need. This is clearly defined, actionable, application-specific knowledge derived from data. This knowledge needs to be delivered in an automated, user-friendly environment. We clearly are not there yet. To get there, we need to appreciate the full scope of data mining, understand the needs of the end user, and define a course of study in this discipline. In addition, we need to deliver knowledge in ways that make the highly complex tasks of data mining transparent to the user. Only then will data-mining results meet or exceed the expectations of business users.

My purpose is to define the discipline of data mining -- including its scope and approach -- as it applies to the use of decision technologies in marketing applications.

For more information, contact gmf@cmsnet.com
Part II of this series will be published in next week's D S * .