Loyalty programs have exploded in popularity in recent decades. In the United States alone, membership has reached 1.3 billion (Ferguson and Hlavinka, 2007). In spite of their continued popularity, the effectiveness of these programs has been long debated in the literature, with mostly mixed results. Verhoef (2003) finds that the effects are positive but very small, DeWulf et al. (2001) finds no support for positive effects of direct mail, Shugan (2005) finds that firms gain short term revenue at the expense of longer term reward payments, and Hartmann and Viard (2008) found no evidence of the loyalty program creating switching costs. Rather than attempt to broadly label loyalty programs as either effective or ineffective, this dissertation instead focuses on how firms can use their loyalty program databases to model customer behavior. In the first essay I investigate how positive and negative casino experiences influence the casino's targeting strategy. In the second essay I study a coalition loyalty program and use the variation in the coalition network size to estimate the value of store participation. Finally, I extend the first two essays and summarize the ongoing debate as to whether the human brain processes information using Bayesian inference. The intent of this research is to both contribute to academic literature and also provide insight for practitioners to improve decision making.
In the first essay I consider the direct marketing targeting problem in situations where 1) the customer's experience quality level varies from occasion to occasion, 2) the firm has measures of these quality levels, and 3) the firm can customize marketing according to these measures and the customer's behaviors. A primary contribution of this paper is a framework and methodology that allows the manager to assess the marketing response of a forward-looking customer with any specific experience and behavior history, which in turn can be used to decide which customers to target for marketing. This research develops a novel, tractable way to estimate and introduce flexible heterogeneity distributions into Bayesian learning models with forward looking agents. The model is estimated using data from the casino industry, an industry which generates �more than $60 billion in U.S. revenues but has surprisingly little academic, econometric research. The counterfactuals offer interesting findings on gambler learning and direct marketing responsiveness and suggest that casino profitability can increase substantially when marketing incorporates gamblers' beliefs and past outcome sequences into the targeting decision.
In the second essay I consider the problem faced by managers of coalition loyalty programs of store network composition. In a coalition loyalty program, managers need to determine which stores to include in the coalition network. The value of a particular store may depend on both the changes in projected member spend at the focal store and the changes in spend at the other stores in the network. The primary contribution of this research is a model that can measure these same-store and cross-store effects. These cross effects can be used to determine the extent of spillover both to and from a focal store from other stores in the network. I estimate the model using data from a coalition network in Europe and find that there are substantial cross-store effects. The findings have substantial implications for managers: even if an individual store is not contributing much revenue to the coalition its presence in the network can positively influence the network as a whole substantially.
In the extension I summarize the recent debate as to whether or not the human brain processes information in a Bayesian fashion and propose potential departures from rational Bayesian inference. I then present models of these departures and review the characteristics of the data sources that would be ideal in estimating these effects. This extension builds on the first two essays by 1) exploring more deeply the learning model presented in the first essay and 2) presenting models that may account for the potential imperfect recall and incomplete knowledge that was assumed in the second essay. The goal of the extension is to highlight how a clear understanding of how the brain operates and processes new experiences could have drastic implications for a loyalty program's design and targeting marketing strategy.
