Data Mining Techniquesin CRM

Inside Customer Segmentation

Konstantinos TsiptsisCRM & Customer Intelligence Expert, Athens, Greece

Antonios ChorianopoulosData Mining Expert, Athens, Greece

A John Wiley and Sons, Ltd., Publication

ayyappan9780470685822.jpg

Data Mining Techniquesin CRM

Data Mining Techniquesin CRM

Inside Customer Segmentation

Konstantinos TsiptsisCRM & Customer Intelligence Expert, Athens, Greece

Antonios ChorianopoulosData Mining Expert, Athens, Greece

A John Wiley and Sons, Ltd., Publication

This edition first published 2009 2009, John Wiley & Sons, Ltd

Registered officeJohn Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, United Kingdom

For details of our global editorial offices, for customer services and for information about how to apply forpermission to reuse the copyright material in this book please see our website at www.wiley.com.

The right of the author to be identified as the author of this work has been asserted in accordance with theCopyright, Designs and Patents Act 1988.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, inany form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted bythe UK Copyright, Designs and Patents Act 1988, without the prior permission of the publisher.

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not beavailable in electronic books.

Designations used by companies to distinguish their products are often claimed as trademarks. All brand namesand product names used in this book are trade names, service marks, trademarks or registered trademarks oftheir respective owners. The publisher is not associated with any product or vendor mentioned in this book. Thispublication is designed to provide accurate and authoritative information in regard to the subject matter covered.It is sold on the understanding that the publisher is not engaged in rendering professional services. If professionaladvice or other expert assistance is required, the services of a competent professional should be sought.

Library of Congress Cataloguing-in-Publication Data

Record on file

A catalogue record for this book is available from the British Library.

ISBN: 978-0-470-74397-3

Typeset in 11/13.5pt NewCaledonia by Laserwords Private Limited, Chennai, India.Printed and bound in Great Britain by Antony Rowe Ltd, Chippenham, Wiltshire.

www.wiley.com

To my daughter Eugenia and my wife Virginia, for their support and understanding.And to my parents.

Antonios

In memory of my father.Dedicated to my daughters Marcella and Christina, my wife Maria, my sister Marina andmy niece Julia and of course, to my mother Maria who taught me to set my goals in life.

Konstantinos

CONTENTS

ACKNOWLEDGEMENTS xiii

1 DATA MINING IN CRM 1The CRM Strategy 1What Can Data Mining Do? 2

Supervised/Predictive Models 3Unsupervised Models 3Data Mining in the CRM Framework 4

Customer Segmentation 4Direct Marketing Campaigns 5Market Basket and Sequence Analysis 7

The Next Best Activity Strategy and Individualized Customer Management 8The Data Mining Methodology 10Data Mining and Business Domain Expertise 13Summary 13

2 AN OVERVIEW OF DATA MINING TECHNIQUES 17Supervised Modeling 17

Predicting Events with Classification Modeling 19Evaluation of Classification Models 25Scoring with Classification Models 32

Marketing Applications Supported by Classification Modeling 32Setting Up a Voluntary Churn Model 33Finding Useful Predictors with Supervised Field Screening Models 36Predicting Continuous Outcomes with Estimation Modeling 37

Unsupervised Modeling Techniques 39Segmenting Customers with Clustering Techniques 40Reducing the Dimensionality of Data with Data Reduction Techniques 47Finding What Goes with What with Association or Affinity Modeling

Techniques 50Discovering Event Sequences with Sequence Modeling Techniques 56Detecting Unusual Records with Record Screening Modeling Techniques 59

Machine Learning/Artificial Intelligence vs. Statistical Techniques 61Summary 62

viii CONTENTS

3 DATA MINING TECHNIQUES FOR SEGMENTATION 65Segmenting Customers with Data Mining Techniques 65Principal Components Analysis 65

PCA Data Considerations 67How Many Components Are to Be Extracted? 67What Is the Meaning of Each Component? 75Does the Solution Account for All the Original Fields? 78Proceeding to the Next Steps with the Component Scores 79Recommended PCA Options 80

Clustering Techniques 82Data Considerations for Clustering Models 83Clustering with K-means 85

Recommended K-means Options 87Clustering with the TwoStep Algorithm 88

Recommended TwoStep Options 90Clustering with Kohonen Network/Self-organizing Map 91

Recommended Kohonen Network/SOM Options 93Examining and Evaluating the Cluster Solution 96

The Number of Clusters and the Size of Each Cluster 96Cohesion of the Clusters 97Separation of the Clusters 99

Understanding the Clusters through Profiling 100Profiling the Clusters with IBM SPSS Modelers Cluster Viewer 102Additional Profiling Suggestions 105

Selecting the Optimal Cluster Solution 108Cluster Profiling and Scoring with Supervised Models 110An Introduction to Decision Tree Models 110

The Advantages of Using Decision Trees for Classification Modeling 121One Goal, Different Decision Tree Algorithms: C&RT, C5.0, and CHAID 123

Recommended CHAID Options 125Summary 127

4 THE MINING DATA MART 133Designing the Mining Data Mart 133The Time Frame Covered by the Mining Data Mart 135The Mining Data Mart for Retail Banking 137

Current Information 138Customer Information 138Product Status 138

Monthly Information 140Segment and Group Membership 141Product Ownership and Utilization 141Bank Transactions 141

Lookup Information 143Product Codes 144

CONTENTS ix

Transaction Channels 145Transaction Types 145

The Customer Signature from the Mining Data Mart to the MarketingCustomer Information File 148Creating the MCIF through Data Processing 149Derived Measures Used to Provide an Enriched Customer View 154

The MCIF for Retail Banking 155The Mining Data Mart for Mobile Telephony Consumer (Residential) Customers 160

Mobile Telephony Data and CDRs 162Transforming CDR Data into Marketing Information 162

Current Information 163Customer Information 164Rate Plan History 165

Monthly Information 167Outgoing Usage 167Incoming Usage 169Outgoing Network 170Incoming Network 170

Lookup Information 170Rate Plans 171Service Types 171Networks 172

The MCIF for Mobile Telephony 172The Mining Data Mart for Retailers 177

Transaction Records 179Current Information 179

Customer Information 179Monthly Information 180

Transactions 180Purchases by Product Groups 182

Lookup Information 183The Product Hierarchy 183

The MCIF for Retailers 184Summary 187

5 CUSTOMER SEGMENTATION 189An Introduction to Customer Segmentation 189

Segmentation in Marketing 190Segmentation Tasks and Criteria 191

Segmentation Types in Consumer Markets 191Value-Based Segmentation 193Behavioral Segmentation 194Propensity-Based Segmentation 195Loyalty Segmentation 196Socio-demographic and Life-Stage Segmentation 198

x CONTENTS

Needs/Attitudinal-Based Segmentation 199Segmentation in Business Markets 200A Guide for Behavioral Segmentation 203

Behavioral Segmentation Methodology 203Business Understanding and Design of the Segmentation Process 203Data Understanding, Preparation, and Enrichment 205Identification of the Segments with Cluster Modeling 208Evaluation and Profiling of the Revealed Segments 208Deployment of the Segmentation Solution, Design, and Delivery of

Differentiated Strategies 211Tips and Tricks 211

Segmentation Management Strategy 213A Guide for Value-Based Segmentation 216

Value-Based Segmentation Methodology 216Business Understanding and Design of the Segmentation Process 216Data Understanding and Preparation Calculation of the Value Measure 218Grouping Customers According to Their Value 218Profiling and Evaluation of the Value Segments 219Deployment of the Segmentation Solution 219

Designing Differentiated Strategies for the Value Segments 220Summary 223

6 SEGMENTATION APPLICATIONS IN BANKING 225Segmentation for Credit Card Holders 225

Designing the Behavioral Segmentation Project 226Building the Mining Dataset 227Selecting the Segmentation Population 228The Segmentation Fields 230The Analytical Process 233

Revealing the Segmentation Dimensions 233Identification and Profiling of Segments 237

Using the Segmentation Results 256Behavioral Segmentation Revisited: Segmentation According to All Aspects of

Card Usage 258The Credit Card Case Study: A Summary 263

Segmentation in Retail Banking 264Why Segmentation? 264Segmenting Customers According to Their Value: The Vital Few Customers 267Using Business Rules to Define the Core Segments 268Segmentation Using Behavioral Attributes 271

Selecting the Segmentation Fields 271The Analytical Process 274

Identifying the Segmentation Dimensions with PCA/Factor Analysis 274Segmenting the Pure Mass Customers with Cluster Analysis 276Profiling of Segments 276

CONTENTS xi

The Marketing Process 283Setting the Business Objectives 283

Segmentation in Retail Banking: A Summary 288

7 SEGMENTATION APPLICATIONS IN TELECOMMUNICATIONS 291Mobile Telephony 291

Mobile Telephony Core Segments Selecting the Segmentation Population 292Behavioral and Value-Based Segmentation Setting Up the Project 294Segmentation Fields 295Value-Based Segmentation 300Value-Based Segments: Exploration and Marketing Usage 304Preparing Data for Clustering Combining Fields into Data Components 307Identifying, Interpreting, and Using Segments 313Segmentation Deployment 326

The Fixed Telephony Case 329Summary 331

8 SEGMENTATION FOR RETAILERS 333Segmentation in the Retail Industry 333The RFM Analysis 334

The RFM Segmentation Procedure 338RFM: Benefits, Usage, and Limitations 345

Grouping Customers According to the Products They Buy 346Summary 348

FURTHER READING 349

INDEX 351

ACKNOWLEDGEMENTS

We would like to thank Vlassis Papapanagis, Leonidas Georgiou and IoannaKoutrouvis of SPSS, Greece. Also, Andreas Kokkinos, George Krassadakis, KyriakosKokkalas and Loukas Maragos. Special thanks to Ioannis Mataragas for the creationof the line drawings.