Determining the importance that customers place on the product and service attributes that drive their satisfaction with, and loyalty to, service providers is an essential part of a firm’s resource allocation process. An unsettled issue is whether importance measures should come directly from customers or be derived statistically and, if so, how. The authors compare direct importance ratings with a variety of methods for statistically deriving attribute importance in a customer satisfaction model. Using three data sets, the methods are compared on criteria that include their ability to explain variation in satisfaction, to identify customers’ more important attributes, and to be interpretable. The findings suggest that because each of the tested methods has its strengths and weaknesses, it is essential to choose a method that is compatible with the research goals and context.