Service marketers have access to many forms of acquisition channels. Therefore, understanding which channel is the most effective at attracting loyal customers is of growing importance. However, evaluation of a customer acquisition channel is not always straightforward. Ideally, the true contribution can be estimated in a controlled setting where a firm may assign channels randomly to each customer and analyze the relationship between the channel and the corresponding retention behavior. In practice, on the contrary, customers may preferentially select their subscription channels. Thereby, a particular channel may attract disproportionately more (or less) loyal customers, resulting in biased over- (or under-) estimation of the channel impact on customer retention, regardless of the true contribution of the channel per se. This bias is known as self-selection bias.
To our knowledge, no empirical study in the acquisition channel research has incorporated this self-selection bias into the model. Our model addresses this challenge by allowing that the customer acquisition channel use is endogenously determined in the model specification. Further, our model takes into account the sample selection bias that arises from the dependent nature of customer decisions. Empirical illustration shows that the self-selection bias exists in the sample, and confirms that our model provides an unbiased estimate of an acquisition channel effect on customer retention and customer duration, compared to competing models in which the self-selection bias was not considered. Future research may relate our model to other marketing problems such as evaluation of a customers service channel use (i.e., a call center and an email center) and a customers multi-channel use.