We propose an approach to the predictive validation of conjoint models that overcomes several limitations of current approaches. There are situations in which a company, after completing a conjoint study, can compare the actual sales of two (or more) of its products in the same product-market, in approximately the same timeframe, and with approximately the same level of marketing support. We assess predictive validity by comparing the ratio of the actual sales of product A vs. product B with the corresponding ratio of choice shares predicted by the conjoint study. The results from four empirical tests confirm the predictive validity of the self-explicated approach to conjoint modeling.