Unobserved preference heterogeneity is usually represented in discrete choice models by treating utility function coefficients as random variables. In this paper we propose a different approach to explicitly account for preference heterogeneity - we investigate the possibility of estimating individual-level choice models. To deal with required data intensity we designed a choice experiment in which each respondent faced 16 choice-sets, and was asked to provide a full ranking of the four alternatives in each choice-set. The exploded logit formula was used to estimate individual-level utility parameters which were used to calculate WTP of each individual. Using data at the individual level we explored respondents’ decision rules. In general, our results indicate that pooled models may conceal many types of irregular behaviour. Our analysis also indicates that many respondents exhibit lexicographic or near-lexicographic preferences.