Offline event marketing has become increasingly popular. As a large amount of data from location-based social networks (LBSNs), such as Foursquare, Gowalla, and Facebook, becomes available, how to make use of these data to analyze users’ social behaviors is an important issue for offline event marketing. To provide some valuable guidance for businesses, this paper presents a statistical inference approach to optimally selecting participants who have a high probability of visiting an offline event. Technically, we formulate participant selection as a constraint optimization problem. In particular, our marketing cost model takes into account key factors such as distance, loyalty influence, and recommendation index. In addition, four participant-based strategies and a detailed algorithm are presented. Experiments on real-world datasets have demonstrated the effectiveness and efficiency of our proposed approach and the quantitative model.