Understanding the mechanisms underlying participants’ performance in crowdsourcing contests is a critical issue for practitioners to extract maximum value from crowdsourcing contests. However, as a vital factor, research on the role of past successful experience in crowdsourcing contests lacks consistent conclusions and further analysis of its specific factors. Using data collected from Epek.com, we bring together two disparate but related streams of literature and provide a novice understanding of the complicated, non-linear relationship between past successful experience and future performance. Then, we use text mining to capture the specific factors of tasks and experience and verify their moderating effects. The results show that the solver’s success experience has an inverted U-shaped relationship with future performance in crowdsourcing contests. There is a weaker inverted U-shaped relationship when the solver’s experience is more diverse or their past success experience is more similar to the current crowdsourcing task. The implications of our findings for researchers and practitioners are discussed as well.