Conjoint experiments are fast becoming one of the dominant experimental methods within the social sciences. To date our methods of estimating effect heterogeneity within these designs is limited, despite clear benefits to researchers across a wide gamut of different research areas. We approach this problem from a potential outcomes perspective, demonstrating two natural extensions to the estimated parameters in conjoint experiments: the observation-level and individual-level marginal component effects. To recover these quantities, we propose a new estimation strategy and show how it can be used to evaluate treatment effect heterogeneity. Our proposed method exhibits good statistical properties, and we find interesting heterogeneity when re-estimating existing conjoint studies. To accompany this paper, we provide new a R package, cjbart, that allows researchers to model heterogeneity in their experimental conjoint data.