Within the realm of logit-type random parameter models to address unobserved heterogeneity in preferences there are two dominant approaches: the mixed logit model, which assumes parametric and continuous heterogeneity distributions, and the latent class logit model, which is a discrete and semiparametric counterpart of mixed logit. In addition to offer flexibility benefits, random parameter models allow researchers to make conditional (posterior) inference on preference parameters at the individual-specific level. In this paper we extend the individual-specific experimental approach, that was conducted by Revelt and Train (2000) for the continuous heterogeneity distributions of a mixed logit, to the discrete case of the latent class logit model. Our Monte Carlo study results confirm the expectation that for a given number of individuals, the density of the conditional means converges to the conditional population as the number of choice situations increases. We also add to the analysis the behavior of interval estimates using two methods for the derivation of standard errors of the individual-specific estimates. In general, as we have more information of the choices made by the individuals, we are in better shape to identify individual-specific preferences. Our main conclusion is that accurate individual-specific estimation is possible – including correct assignment to classes, but a large number of choice situations is needed to correctly approximate the true underlying distribution.