Should retailers take product returns into account when choosing their assortments? And, when doing so, should they consider assortment selection as a popularity contest - by carrying products that they think will be popular among consumers? Or, is there ever a case for carrying eccentric products - those that are least likely to be purchased by a typical consumer? In search of answers to these questions, we explore in this chapter the interactions between product assortment, return policy, and pricing decisions of a retailer. We consider a category of horizontally differentiated products delivered in two alternative supply modes: make-to-order (MTO) and make-to-stock (MTS). In the MTO mode, products are supplied after demand materializes, whereas in the MTS mode, the retailer stocks products prior to the selling season. Underlying our demand model, consumer choice behavior follows a nested multinomial logit model, with the first stage involving a product choice, and the second stage involving a keep-or-return decision. We show that the structure of the optimal assortment strongly depends on both the return policy, which we parameterize by refund fraction (percentage of price refunded upon return) and the supply mode (MTO vs. MTS). For relatively strict return policies with a sufficiently low refund fraction, it is optimal for the retailer to offer most eccentric products in the MTO mode, and a mix of most popular and most eccentric products in the MTS mode. For relatively lenient return policies, on the other hand, conventional thinking applies: the retailer selects most popular products. We also numerically study three extensions of our base model to incorporate: (1) endogenous price, (2) endogenous refund fraction, and (3) multiple periods.We demonstrate that interesting aspects of our results regarding strict return policies prevail under all of these extensions.