Recent years have seen a growth in customized products and services. As a prerequisite for customization, private information on individual customers’ quality preferences needs to be uncovered. Sellers can listen to customers about their stated or self-reported preferences through direct communication (e.g., conversation, survey). Alternatively, customer preferences can be inferred from their behavior when they are given the rights to self-design the quality. In this research we endogenize the viability of customization by investigating whether and when customers may reveal their stated/inferred preferences truthfully. We find that, for either preference-learning approach, customers would voice their preferences faithfully if and only if they are sufficiently heterogenous. Equilibrium preference revelation, and hence endogenous customization, tend to be sustained by intermediate seller bargaining power or nonextreme production/selling costs. We examine how the preference-learning approaches may differ in the endogenous feasibility of customization, equilibrium qualities, and the parties’ expected payoffs. We show that giving up the design right need not always be harmful for the seller, and gaining it can make the buyers worse off, especially when fixed costs of customization are considered. This paper was accepted by Eric Anderson, marketing. Funding: This work was supported by a DAG grant offered by the Hong Kong Research Grant Council.