In recent years there has been disenchantment with the performance and relevance of customer satisfaction models in applied business contexts. An expected utility framework overcomes many of the weaknesses and provides a possible relevant solution. This paper applies a designed stated choice experiment examining the impact of disconfirmation on customer satisfaction and ultimately switching behaviour. Additionally, the two-stage experiment incorporates both mean and variance of attribute expectations. The impact of expectations, product performance and disconfirmation on satisfaction and in particular, future choices, is estimated using contingency tables and a binary logit model for brand switching. Overall, both mean and variance of attribute expectations are relevant in explaining choices. Pre-experience choice depends on mean and variance of attribute expectations for both brands. Product experience with a given brand impacts on attribute expectations for both brands. Switching behaviour between pre and post experience choices is explained by both Satisfaction and updating of attribute expectations (mean) for both brands in the experiment. The predictive performance of the switching model based on both satisfaction and updated attribute expectations is reasonably good with overall approximately 75% correct predictions.