Patterns of Customer Exit in a Contract-Based Subscription Service

Abstract

Customer exit is one of the most extreme forms of complaining behavior, as pointed out by Hirschman (1970). It may be precipitated or accompanied by some negative affect toward the company supplying the service, although it may also be caused by a more distanced calculation of price and value. The timing of the customer's intended exit is extremely important, and has been studied heavily. That timing is complicated for a service which has been initiated by a contract with a definite lifespan, and attention centers on the importance of the contract expiration date and the customer's post- (and sometimes pre-) expiration behavior. The supplier is typically very interested in such exit patterns, their relationship to available internal information, and their consequences for such customer retention tactics. In this paper, we outline a method for viewing the exit propensity of cellular telephone customers, most of whom initiate service through the signing of a contract for a specified term. The production of these estimated propensities follow classical statistical survival analysis, and are augmented by the output of a neural net model for customer lifetime data. These propensities for exit then take the form of a statistical hazard function, whose components estimate the conditional probability of a customer's departure in the month of his/her service. These hazard functions will be categorized and related to internal company database information. This determines a way of segmenting customers, and creating indicators of that segmentation. Finally, we suggest a way to use this information to understand the customer's situation and to develop retention tactics. This concept is an extension of the classical use of lifetime value (LTV) as developed for the mail order industry, in that it affords a way of quantifying the value of allocating marketing strategies to customers with different valuations.