We investigate various dynamics characterizing the crowdfunding process: stagnation after friend-funding, gradual increase through crowd participation, and acceleration in the last phase. We propose three mechanisms as major drivers of the crowdfunding dynamics: forward-looking delaying investment behavior, contempora- neous social interactions, and forward-looking social interactions. We apply the rational expectations equilibrium of the approximate aggregation approach to model the under- lying mechanisms. Using the Bayesian IJC method, we analyze individual-level investment data from a crowdfunding platform, Sellaband. We find strong evidence for the three mechanisms and confirm that they contribute to the contrasting dynamic patterns ob- served in our data. We also simulate counterfactuals to derive optimal policy decisions for both fundraisers and platforms. For fundraisers, we infer the optimal goals that ensure goal completion while raising the maximum capital. For platforms, we suggest an optimal targeting strategy that identifies those crowdfunders who contribute the most to the crowding process and, ultimately, goal success. Also, we provide critical input for various resource allocation decisions by accurately predicting whether the project will succeed and when it will succeed at the time when 50% of the goal has been achieved.