Data Mining Cookbook: Modeling Data for Marketing, Risk, and Customer Relationship Management by Olivia Parr Rud

From the Publisher

Increase profits and reduce costs by utilizing this collection of models of the most commonly asked data mining questions

In order to find new ways to improve customer sales and support, and as well as manage risk, business managers must be able to mine company databases. This book provides a step-by-step guide to creating and implementing models of the most commonly asked data mining questions. Readers will learn how to prepare data to mine, and develop accurate data mining questions. The author, who has over ten years of data mining experience, also provides actual tested models of specific data mining questions for marketing, sales, customer service and retention, and risk management. A CD-ROM, sold separately, provides these models for reader use.

Product Details

• Table of Contents
• Read an Excerpt

Table of Contents

Read an Excerpt

I am a data miner by vocation and home chef by avocation, so I was naturally intrigued when I heard about Olivia Parr Rud's Data Mining Cookbook. What sort of cookbook would it be, I wondered? My own extensive and eclectic cookery collection is comprised of many different styles. It includes lavishly illustrated coffee-table books filled with lush photographs of haute cuisine classics or edible sculptures from Japan's top sushi chefs. I love to feast my eyes on this sort of culinary erotica, but I do not fool myself that I could reproduce any of the featured dishes by following the skimpy recipes that accompany the photos! My collection also includes highly specialized books devoted to all the myriad uses for a particular ingredient such as mushrooms or tofu. There are books devoted to the cuisine of a particular country or region; books devoted to particular cooking methods like steaming or barbecue; books that comply with the dictates of various health, nutritional or religious regimens; even books devoted to the use of particular pieces of kitchen apparatus. Most of thesebooks were gifts. Most of them never get used.

But, while scores of cookbooks sit unopened on the shelf, a few--Joy of Cooking, Julia Child--have torn jackets and colored Post-its stuck on many pages. These are practical books written by experienced practitioners who understand both their craft and how to explain it. In these favorite books, the important building blocks and basic techniques (cooking flour and fat to make a roux; simmering vegetables and bones to make a stock; encouraging yeast dough to rise and knowing when to punch it down, knead it, roll it, or let it rest) are described step by step with many illustrations. Often, there is a main recipe to illustrate the technique followed by enough variations to inspire the home chef to generalize still further.

I am pleased to report that Olivia Parr Rud has written just such a book. After explaining the role of predictive and descriptive modeling at different stages of the customer lifecycle, she provides case studies in modeling response, risk, cross-selling, retention, and overall profitability. The master recipe is a detailed, step-by-step exploration of a net present value model for a direct-mail life insurance marketing campaign. This is an excellent example because it requires combining estimates for response, risk, expense, and profitability, each of which is a model in its own right. By following the master recipe, the reader gets a thorough introduction to every step in the data mining process, from choosing an objective function to selecting appropriate data, transforming it into usable form, building a model set, deriving new predictive variables, modeling, evaluation, and testing. Along the way, even the most experienced data miner will benefit from many useful tips and insights that the author has gleaned from her many years of experience in the field.

At Data Miners, the analytic marketing consultancy I founded in 1997, we firmly believe that data mining projects succeed or fail on the basis of the quality of the data mining process and the suitability of the data used for mining. The choice of particular data mining techniques, algorithms, and software is of far less importance. It follows that the most important part of a data mining project is the careful selection and preparation of the data, and one of the most important skills for would-be data miners to develop is the ability to make connections between customer behavior and the tracks and traces that behavior leaves behind in the data. A good cook can turn out gourmet meals on a wood stove with a couple of cast iron skillets or on an electric burner in the kitchenette of a vacation condo, while a bad cook will turn out mediocre dishes in a fancy kitchen equipped with the best and most expensive restaurant-quality equipment. Olivia Parr Rud understands this. Although she provides a brief introduction to some of the trendier data mining techniques, such as neural networks and genetic algorithms, the modeling examples in this book are all built in the SAS programming language using its logistic regression procedure. These tools prove to be more than adequate for the task.

This book is not for the complete novice; there is no section offering new brides advice on how to boil water. The reader is assumed to have some knowledge of statistics and analytical modeling techniques and some familiarity with the SAS language, which is used for all examples. What is not assumed is familiarity with how to apply these tools in a data mining context in order to support database marketing and customer relationship management goals. If you are a statistician or marketing analyst who has been called upon to implement data mining models to increase response rates, increase profitability, increase customer loyalty or reduce risk through data mining, this book will have you cooking up great models in no time.

Michael J. A. Berry
Founder, Data Miners, Inc
Co-author, Data Mining Techniques and Mastering Data Mining

Data mining is a term that covers a broad range of techniques being used in a variety of industries. Due to increased competition for profits and market share in the marketing arena, data mining has become an essential practice for maintaining a competitive edge in every phase of the customer lifecycle.

Historically, one form of data mining was also known as "data dredging." This was considered beneath the standards of a good researcher. It implied that a researcher might actually search through data without any specific predetermined hypothesis. Recently, however, this practice has become much more acceptable, mainly because this form of data mining has led to the discovery of valuable nuggets of information. In corporate America, if a process uncovers information that increases profits, it quickly gains acceptance and respectability.

Another form of data mining began gaining popularity in the marketing arena in the late 1980s and early 1990s. A few cutting edge credit card banks saw a form of data mining, known as data modeling, as a way to enhance acquisition efforts and improve risk management. The high volume of activity and unprecedented growth provided a fertile ground for data modeling to flourish. The successful and profitable use of data modeling paved the way for other types of industries to embrace and leverage these techniques. Today, industries using data modeling techniques for marketing include insurance, retail and investment banking, utilities, telecommunications, catalog, energy, retail, resort, gaming, pharmaceuticals, and the list goes on and on.

What Is the Focus of This Book?

There are many books available on the statistical theories that underlie data modeling techniques. This is not one of them! This book focuses on the practical knowledge needed to use these techniques in the rapidly evolving world of marketing, risk, and customer relationship management (CRM).

Most companies are mystified by the variety and functionality of data mining software tools available today. Software vendors are touting "ease of use" or "no analytic skills necessary." However, those of us who have been working in this field for many years know the pitfalls inherent in these claims. We know that the success of any modeling project requires not only a good understanding of the methodologies but solid knowledge of the data, market, and overall business objectives. In fact, in relation to the entire process, the model processing is only a small piece.

The focus of this book is to detail clearly and exhaustively the entire model development process. The details include the necessary discussion from a business or marketing perspective as well as the intricate SAS code necessary for processing. The goal is to emphasize the importance of the steps that come before and after the actual model processing.

Who Should Read This Book?

As a result of the explosion in the use of data mining, there is an increasing demand for knowledgeable analysts or data miners to support these efforts. However, due to a short supply, companies are hiring talented statisticians and/or junior analysts who understand the techniques but lack the necessary business acumen. Or they are purchasing comprehensive data mining software tools that can deliver a solution with limited knowledge of the analytic techniques underlying it or the business issues relevant to the goal. In both cases, knowledge may be lacking in essential areas such as structuring the goal, obtaining and preparing the data, validating and applying the model, and measuring the results. Errors in any one of these areas can be disastrous and costly.

The purpose of this book is to serve as a handbook for analysts, data miners, and marketing managers at all levels. The comprehensive approach provides step-by-step instructions for the entire data modeling process, with special emphasis on the business knowledge necessary for effective results. For those who are new to data mining, this book serves as a comprehensive guide through the entire process. For the more experienced analyst, this book serves as a handy reference. And finally, managers who read this book gain a basic understanding of the skills and processes necessary to successfully use data models.

How This Book Is Organized

The book is organized in three parts. Part One lays the foundation. Chapter 1 discusses the importance of determining the goal or clearly defining the objective from a business perspective. Chapter 2 discusses and provides numerous cases for laying the foundation. This includes gathering the data or creating the modeling data set. Part Two details each step in the model development process through the use of a case study. Chapters 3 through 7 cover the steps for data cleanup, variable reduction and transformation, model processing, validation, and implementation. Part Three offers a series of case studies that detail the key steps in the data modeling process for a variety of objectives, including profiling, response, risk, churn, and lifetime value for the insurance, banking, telecommunications, and catalog industries.

As the book progresses through the steps of model development, I include suitable contributions from a few industry experts who I consider to be pioneers in the field of data mining. The contributions range from alternative perspectives on a subject such as multi-collinearity to additional approaches for building lifetime value models.

Tools You Will Need

To utilize this book as a solution provider, a basic understanding of statistics is recommended. If your goal is to generate ideas for uses of data modeling from a managerial level then good business judgement is all you need. All of the code samples are written in SAS. To implement them in SAS, you will need Base SAS and SAS/STAT. The spreadsheets are in Microsoft Excel. However, the basic logic and instruction are applicable to all software packages and modeling tools.

The Companion CD-ROM

Within chapters 3 through 12 of this book are blocks of SAS code used to develop, validate, and implement the data models. By adapting this code and using some common sense, it is possible to build a model from the data preparation phase through model development and validation. However, this could take a considerable amount of time and introduce the possibility of coding errors. To simplify this task and make the code easily accessible for a variety of model types, a companion CD-ROM is available for purchase separately.

The CD-ROM includes full examples of all the code necessary to develop a variety of models, including response, approval, attrition or churn, risk, and lifetime or net present value. Detailed code for developing the objective function includes examples from the credit cards, insurance, telecommunications, and catalog industries. The code is well documented and explains the goals and methodology for each step. The only software needed is Base SAS and SAS/STAT.

The spreadsheets used for creating gains tables and lift charts are also included. These can be used by plugging in the preliminary results from the analyses created in SAS.

While the steps before and after the model processing can be used in conjunction with any data modeling software package, the code can also serve as a stand-alone modeling template. The model processing steps focus on variable preparation for use in logistic regression. Additional efficiencies in the form of SAS macros for variable processing and validation are included.

What Is Not Covered in This Book

A book on data mining is really not complete without some mention of privacy. I believe it is a serious part of the work we do as data miners. The subject could fill an entire book. So I don't attempt to cover it in this book. But I do encourage all companies that use personal data for marketing purposes to develop a privacy policy. For more information and some simple guidelines, contact the Direct Marketing Association at (212) 790-1500 or visit their Web site at www. the-dma. org.

Summary