Can we predict top-performing products, services, or businesses by only monitoring the behavior of a small set of individuals? Although most previous studies focused on the predictive power of "hub" individuals with many social contacts, which sources of customer behavioral data are needed to address this question remains unclear, mostly due to the scarcity of available datasets that simultaneously capture individuals' purchasing patterns and social interactions. Here, we address this question in a unique, large-scale dataset that combines individuals' credit-card purchasing history with their social and mobility traits across an entire nation. Surprisingly, we find that the purchasing history alone enables the detection of small sets of ``discoverers" whose early purchases offer reliable success predictions for the brick-and-mortar stores they visit. In contrast with the assumptions by most existing studies on word-of-mouth processes, the hubs selected by social network centrality are not consistently predictive of success. Our findings show that companies and organizations with access to large-scale purchasing data can detect the discoverers and leverage their behavior to anticipate market trends, without the need for social network data.