This paper studies the influence of observational learning and herding in networks of friends versus informants on consumer purchase decisions. We explore how people trade off their needs to belong and to be different by first developing an exponential random graph model to predict online purchasing decision while taking into considerations of product properties, consumer demographics, online rating, as well as consumer social networks. We test our model through collecting panel data on a leading social commerce site in Asia. Contrary to the popular belief that people tend to follow friends’ choices, subjects in our context are more likely to diverge from the popular choice among their friends. As our study shows that the need to be different can dominate the need to be belong in certain contexts, we discuss managerial implications of our results for social media marketing.