Abstract: Recent years have witnessed the phenomenal success of a new form of social e-commerce platforms, which transforms users into agents by motivating them with monetary rewards to promote products and invite new agents through their social network. Despite their rapid growth, there is still inadequate evidence on how such agent invitation works. This research examines what factors affect the agent invitation process. We first conduct a qualitative user study, where we identify four potential mechanisms related to the agent invitation: social conformity, social enrichment, refusal avoidance, and benefit-cost trade-off. Leveraging the empirical data collected from one of the largest social e-commerce platforms in China - Beidian, we operationalize a set of behavioral indicators of these mechanisms and further develop machine learning models to predict users' reactions to invitations. We found that the identified four mechanisms contribute to the high success rate of agent invitations differently. We conclude by discussing the implications of our findings and their potential benefits to real-world applications.