This dissertation studies problems of managing the operations in online platforms, especially when the behavior of agents on the platform is leveraged by users/platform managers to make better decisions. We study three distinct online platforms with unique challenges: in online crowdfunding platforms, how should a creator of campaigns take advantage of the heterogeneity in backers (independent backers vs. herding backers) to maximize the expected revenue; in online spot labor platforms, how employers decide whom to hire, and how experiential learning could help the platform create better matches; in online social media platforms, how people are affected by misinformation. We use both mathematical modeling and data-analytic tools to tackle these problems: we propose optimal pledge levels and campaign durations for campaign creators; we uncover the role of experiential learning in hiring decisions, and design platform's policies to exploit the informational advantage of employers; we find that consumers tend to co-consume mainstream information and unverified information, but substitution is not necessarily required for consumers to get hurt from unverified information. We hope this dissertation sheds lights on the management of different types of online platforms.