Metric hybrid conjoint models have received a fair amount of industry application to date. The purpose of these models is to reduce data collection time while still retaining individual differences in part-worth functions. The present paper extends this class of models to include categorical conjoint analysis in which the criterion variable is classificatory. This model is applied to an extremely large conjoint problem involving over 40 attributes and over 100 attribute levels. The study results support the viability of the model for dealing with extremely large conjoint problems. The study also shows evidence of the inability of simple functions of self-explicated utilities for components of a bundle of hotel amenities to predict respondents' preferences for the total bundle.