This book provides a managerial overview of the application of data mining to Customer Relationship Management. It is one of the few available books that organizes much of its material around CRM functions (i.e. Segmentation, Retention, Cross-selling, etc.) This organization is appropriate for the person interested in understanding the marketing context of various data mining projects.
Chapters 1-3 are light reading and may be skimmed for purposes of this course.
Chapter 4 discusses the
relationship between data mining and data warehousing. I will be augmenting this
material in the first lecture.
Chapter 5 contains many
important, relevant concepts regarding the nature of data mining and its relationship to
other forms of data analysis, most notably OLAP and statistics. I will be
augmenting this material in the first three lectures. WARNING: There are several points made by the authors
in this chapter that I disagree with. I will be discussing these points in
class.
Chapter 6 presents what the authors call "classical" data mining
techniques and chapter 7 presents what the authors call "next
generation" techniques. In my opinion their distinction is arbitrary. Our
other text, Data Mining Techniques by Berry and Linoff, does a better job of
introducing the major techniques in their chapter 7.
Chapter 8 touches on several important, practical aspects of data mining. Perhaps the chapter would have been better titled "Data Mining Best Practices." The section "Using the Right Technique" will be covered in the "Overview of Data Mining Techniques" lecture. It overlaps with Berry and Linoff's chapter 17, "Choosing the Right Tool for the Job."
This section comprises the core concepts of our case studies and should be read carefully. Unfortunately, each section is rather light in its treatment. Once again, our other textbook provides valuable examples and supporting material.
Chapter 9 discusses Customer Profitability and its effect on the data mining project. While customer profitability is increasingly important in today's competitive landscape, it really represents a more sophisticated level of customer information - information that many companies are just now beginning to collect. The basic CRM functions, customer acquisition, expansion (cross-selling), and retention, are refined when customer profitability is considered. That is why I have chosen to cover this topic last.
Chapter 10 discusses Customer Acquisition. This is perhaps the most difficult CRM function to deploy data mining solutions for because it relies on data about non-customers (i.e. prospects). Obtaining good prospect data is expensive and time-consuming. Successfully acquiring new customers is also notoriously expensive. Many companies have relied on brand equity and mass marketing to achieve market penetration and market share. Data mining is useful for improving mass marketing programs and also for executing targeted marketing campaigns.
Chapter 11 discusses Cross-Selling. This is an active area of data mining and recently there have been a number of case studies publicized. There are a large number of modeling approaches being actively developed. It is an especially important tactic for mature markets where competitors are essentially competing for the same customers and want to expand their relationships with them. This is especially true for companies looking to compete on service quality and satisfaction rather than price or product innovation. The catch-phrase here is "Sell more stuff!"
Chapter 12 discusses Customer Retention. Cell phone customer churn (or customer turnover) is a popular case study in data mining. We will also be discussing customer retention in retail banking. The appeal to this tactic is that it is much cheaper (and therefore more profitable) to retain customers than it is to go out and develop new ones. The task of the data miner is to identify those who are most likely to leave, which ones are the most desirable to keep, and how best to keep them.
Chapter 13 discusses Customer Segmentation. Segmentation, like Profitability, is a key marketing technique that overlaps the previous three activities. In the past, segmentation schemes have been developed using statistical analysis, empirical observations, and industry-specific tradition. Data mining provides a new twist to the mix by allowing what the authors call "data-driven" segmentation. Properly used, this approach can yield improved marketing results over traditional methods, but may also be harder to explain.
Chapters 14 - 18 cover major milestones in a typical data mining project. What's missing from this material is a more complete discussion of data preparation. We will have an entire lecture devoted to that topic alone. Berry and Linoff calls their process framework "The Virtuous Cycle of Data Mining." An industry consortium that includes SPSS has published the CRISP-DM methodology. And SAS follows their SEMMA framework. We will integrate material from chapters 14 and 15 in our discussions of the Data Mining Process. Chapters 16 - 18 may be read as supplemental material.
Chapter 19 is a somewhat outdated overview of the major data mining packages. A lot has happened in the past two years. But this may be of interest to those that are considering evaluating data mining packages and want a little perspective on where some of these systems came from. We will not cover this chapter in class.
Chapter 20 looks at major trends. For the most part, this material is still relevant and may be of interest to students with specific interests that coincide with one of these specializations. We will not cover this chapter in class.