Choice experiments are an effective way of obtaining objective information regarding the voice of the customer. They can be used to obtain the relevant customer attributes and importance rankings used in the first step of quality function deployment. They are also used extensively in marketing research. Optimal designs for choice experiments have been discussed in the literature. However, optimal designs are only optimal for a particular model. In this article we borrow ideas from quality engineering and industrial experimentation to develop designs for choice experiments that are model-robust (in the sense that they are efficient for fitting a model involving main effects plus a few interactions that need not be specified in advance). A case study is presented to illustrate the use of a model-robust design for a choice experiment. Two unsuspected interactions were discovered in the case study, and this discovery led to added insight regarding customer preferences and importance rankings of product attributes. These insights would not have been possible if an optimal design for the main effects model had been used. Copyright © 2011 John Wiley & Sons, Ltd.