Data mining is the automated search for hidden previously unknown and interesting knowledge from large databases. This paper describes the use of data mining in the domain of customer satisfaction studies more specifically to tackle the problem of identifying latent dissatisfied customers. To do this an association rule algorithm is being used to identify the characteristics of dissatisfied customers. Satisfied customers with approximately identical characteristics might have a high probability to become dissatisfied in the near future. This knowledge can be used by domain experts to develop action plans in order to prevent these customers from actually becoming dissatisfied.
In customer satisfaction research, analysis methods are needed that can provide a clear insight into the complex relationships among satisfaction ratings and various performance attributes. The purpose of such an analysis is to identify the areas (processes, attributes) that need improvement and that most significantly enhance the business relationship with the customer. This paper describes a novel method to examine customer satisfaction survey data. The method includes a refinement in assessing attribute importance, by differentiating between value enhancing and decreasing attributes, satisfiers and dissatisfiers. Information gain measures and statistical tests are used for this purpose. The refined relevance measures result in a modified importance-performance chart that is used to suggest a suitable strategy for the attributes under investigation. The method is demonstrated for a sample of 2504 customers of retail banking services.