The purpose of this paper is to discuss the difficulties raised by the question of cultural differences in consumption behaviour studies, and proposes the best-worst method as a tool for comparing data from a cross-national survey. Design/methodology/approach – Data were collected from samples of wine consumers from Australia (n?=?283), the UK (n?=?304) and France (n?=?147), using the BW procedure, where respondents have to assess what are the most and the least important attributes when choosing a wine at a restaurant. Findings – Results show differences between the countries, with a clear contrast between the French, on the one hand, and the Australians and British, on the other. They confirm empirically the idea that the country may be a valid “culti unit” in cross-cultural research. Research limitations/implications – The authors suggest that the method works well in the context of this research, but does not avoid some of the uncontrollable biases of declarative data, nor the question of the relevance of some choice items in cultural contexts that can be very diverse (such as the role of a waiter in a restaurant). Practical implications – The paper shows that the method can be used quite easily in a large number of studies where it is important to hierarchize choice cues and compare different segments of a population. Originality/value – This fairly large scale study contributes to the marketing literature dealing with country of origin as a segmentation criteria, and exploring the concept and measurement of “cultural difference”. Finally, it fosters a scarce literature dealing with consumer behaviour in “on premise” situations.