4.1 Introduction

After customers have been acquired, it is important to discuss the second step of CRM – retaining customers. Customer retention strategies are used in both contractual (where customers are bound by contracts such as mobile phone subscription or magazine subscription) and non-contractual settings (where customers are not bound by contracts, such as grocery purchases or apparel purchases). Reichheld and Sasser (1990) stated that a 5% improvement in customer retention can cause an increase in profitability between 25 and 85%, in terms of net present value, depending upon the industry. Since then, companies have constantly allocated resources on customer retention management and researchers have put much emphasis on studying customer retention.

Research on customer retention has two main streams. One stream is interested in investigating the effects various marketing variables on customer retention, which in turn influences a firm's performance. The other stream is interested in building econometric and statistical models to estimate or predict the customer retention decisions from both the customer and company prospective. Figure 4.1 shows an integrated framework which describes the various relationships examined in different studies across many industries. These industries have included telecommunications, financial services, hairdressing, restaurants, and retailing, among others. The common thread of these relationships is the mindset that (a) increased product and service quality lead to increased customer satisfaction, (b) increased customer satisfaction leads to increased customer retention (which is mediated by relationship quality such that a higher relationship quality positively enhances the link between satisfaction and retention), and (c) increased customer retention leads to increased firm performance.

Figure 4.1 Integrated relationships addressed in customer retention.

Besides understanding the general linkages between product and service quality and firm performance, it is also important to understand the antecedents of each key construct which help drive the value of the construct. For instance, research has shown that several different factors can influence customer satisfaction level including staff performance, disconfirmation, expectation, convenience, and marketing culture (which can also influence customer retention directly). Furthermore, research has shown that customer characteristics and switching costs are two factors most commonly examined as moderating links between customer satisfaction and customer retention.

To examine the effects of various factors highlighted in Figure 4.1 on customer retention, the most commonly used method is confirmatory factor analysis. Confirmatory factor analysis is a multivariate statistical method used to test the understanding researchers have of the constructs of interest. A strong empirical or conceptual foundation is necessary for building up a factor model. Usually, researchers propose a set of hypotheses to explain the phenomenon of interest and confirmatory factor analysis is used to test these hypotheses. For example, one might hypothesize that satisfied customers tend to buy more from the company. Then the researcher would use confirmatory factor analysis to create latent values for customer satisfaction and customer retention, often using surveys of customers, to test this hypothesis. The most commonly used model-fitting procedure for confirmatory factor analysis is maximum likelihood estimation. Statistical software like AMOS, LISREL, EQS, and SAS can be used for confirmatory factor analysis.

The second stream of research on customer retention studies issues (Figure 4.2) involved in decisions made by managers on current customers. There are often several key questions that managers are interested in answering after a customer has been acquired. These include:

• Will the recently acquired customer repurchase or not in the future?
• What will be the lifetime duration of the customer (i.e., when will the customer churn)?
• Given the customer is going to repurchase:

• Does the customer mainly purchase from my firm (high share-of-wallet) or from many different firms (low share-of-wallet)?
• What is the long-term impact of this customer's purchase behavior on firm value?

Figure 4.2 Issues addressed in customer retention modeling.

Many different research studies have been conducted and many models have been developed to try and answer these questions. Table 4.1 contains a representative set of studies that have considered these issues and accounted for many of the problems that might occur in the model-building process. To provide a comprehensive understanding of how to model customer retention, we will review the issues in the studies one by one along with the related modeling techniques. We will also provide empirical examples at the end of each subsection to demonstrate how to apply this knowledge to a representative sample of customers from a B2C firm.

Table 4.1 Review of customer retention models.

Similar to customer acquisition models, the first question that needs to be answered when considering model selection is whether the firm's customers purchase in a contractual versus non-contractual manner. In most instances this will determine the type of statistical model that needs to be used in order to gain any insights from the data.

4.1.1 Data for Empirical Examples

In this chapter we will be providing a description of the key modeling frameworks that attempt to answer each key research question raised at the beginning of the chapter. We will also be providing at least one empirical example at the end of each subsection which will show how sample data can be used to answer these key research questions. For all the empirical examples in this chapter we provide a dataset titled ‘Customer Retention’ which is broken down into two related data tables. In this dataset you will find two tables of data that include a representative sample of 500 customers from a typical B2C firm, where all the customers are from the same cohort. In this case, the cohort consists of a random sample of 500 customers who all made their first purchase with the firm in quarter 1. In the first data table we provide the transaction information for each customer over the course of 12 quarters. Thus, the data table here consists of 6000 rows (500 customers × 12 quarters) and 8 columns. In the second data table we provide the demographic information for each customer. Thus, the data table here consists of 500 rows (500 customers) and 6 columns.

The first data table (labeled Transactions) includes the following variables, which will be used in some combination for each of the subsequent analyses:

The second data table (labeled Demographics) includes the following variables which will be used in some combination for each of the subsequent analyses:

These two data tables will be used for each of the examples presented at the end of each of the sections. These examples will cover the topics of repurchase or not, order quantity, order size, cross-buying, share-of-wallet (SOW), and profitability (CLV).