Modeling customer satisfaction and loyalty: Survey data versus data mining

The loyalty literature has investigated the association between customer satisfaction and customer loyalty and revealed mixed results. Some studies have indicated that the relationship is linear, whereas others have found it to be non-linear. This study examines the nature of this association in retail banking, an issue that has not been tested empirically. A survey study examined bank customers' attitudes, perceptions, and behavior. Bivariate and multivariate testing was applied to develop two loyalty models: one based only on variables typically known to a bank, such as demographics and recent consumer behavior, and the other based on additional survey data. A non-linear relationship between customer satisfaction and customer loyalty was found, and a model explaining 56.9 percent of the variation in customer loyalty was developed. Predictors of loyalty beyond the attitudinal dimensions traditionally tested for their association with loyalty were found to be associated with customers' intentions to remain with their bank. In particular, market conditions such as switching costs and benefits as well as recent consumer behavior were found to add explanatory power. Further, this study contrasted a full model explaining 56.9 percent of the variation in loyalty with a model based only on variables known to banks, which explained only 8.4 percent. Profiling customers based on survey data can thus provide additional explanatory power compared to data mining models The models can be used by bankers to profile customers who are likely to remain loyal, allowing practitioners to implement proactive marketing action to reward such loyalty. Customers least likely to defect have high satisfaction levels, perceive switching as an unattractive option, and typically have a long-established banking relationship.