Artificial intelligence enables modern robots to serve as service and sales assistants. Today's robotic shopping assistants (RSAs) can appear either humanoid or non-humanoid and possess utilitarian and/or hedonic attributes. However, many questions remain unexplored regarding an effective customer-centric RSA design. Do customers prefer a humanoid or non-humanoid RSA with hedonic or utilitarian attributes? To answer those questions, the research deploys a mixed-method approach involving a survey of customers who have interacted with the Pepper Robot, a humanoid robot (Study 1), and follow-up experiments examining customer responses to a humanoid/non-humanoid RSA with hedonic/utilitarian attributes (Studies 2 and 3). The research employs an innovative approach that analyzes both unstructured and structured data simultaneously. Study results suggest that customers prefer humanoid RSAs with utilitarian attributes over those with hedonic attributes. The research contributes to the literature by proposing hedonic (vs. utilitarian) attributes of RSAs as new drivers of anthropomorphic perceptions.