A plethora of commercial social robots and social robotics startups have risen over the last few years. At a cursory glance, most such robots are merely conversational agents, essentially offering a similar or subset of the capabilities of a smart communication device embodied in a mobile/semi-mobile robotic platform. This raises the question of the efficacy of such an approach. In this paper, we explore embodiment using a social, in-the-wild interaction scenario, comparing a Human-Computer and Human-Robot context. A public site has been deliberately chosen to highlight the importance of conducting such user studies in unconstrained social settings as opposed to in controlled lab settings. Increasing evidence suggest the lack of generalizability of lab-based results in the wild, which we argue as a reason for misguided commercialization of social robots and their eventual commercial failures. The results have implications for the longterm commercial viability of such social robots.