Purpose – This paper has two purposes: to use polarisation to identify variations in loyalty and to apply polarisation to an important non-brand attribute, price.Design/methodology/approach – A comprehensive revealed preference data set of wine purchases is used to apply polarisation. Polarisation was defined in two ways: as a function of the beta binomial distribution (BBD) to give a measure of loyalty for an alternative; and as a function of the Dirichlet multinomial distribution (DMD) to give a baseline level of loyalty. Variations were identified by analysing the differences between the BBD and DMD.Findings – Polarisation was shown to be one way of identifying variation across price tiers. In the empirical example used, the DMD model is violated with the price tiers not being directly substitutable with one another. Buyers show excess loyalty towards the lowest and highest price tier levels. One tier shows “change-of-pace” loyalty. Small brands do better when they focus on high loyalty tiers, middle brands compete in the change-of-pace tier and large brands do well across all tiers.Originality/value – Very little work has been undertaken into price tier loyalty and no known empirical research has been undertaken into behavioural loyalty to price tiers in wine. Very little empirical research has considered the association between excess loyalty for attribute levels (such as price tiers) and the existence of niche, change-of-pace and reinforcing brands.