Online crowdfunding platforms such as Kickstarter and Indiegogo make it possible for users to pledge funds to help creators bring their favorite projects into life. With an increasing number of users participating in crowdfunding, researchers are progressively motivated to investigate on improving user experiences by recommending projects and predicting project outcomes. To prompt the sustainable development of these platforms, understanding backers' behaviors becomes also important, as it helps platforms provide better services and improve backer retention. In particular, studying backers' temporal behaviors allows them to monitor the dynamics of backers' actions and develop appropriate strategies in time. Therefore, in this paper, we analyze a large amount of backer data from Kickstarter and Indiegogo, and do a comprehensive quantitative analysis on users' temporal backing patterns. Employing time series clustering methods, we discover four distinct temporal backing patterns on both platforms. In addition, we explore various characteristics of these backing patterns and possible factors affecting backers' behaviors. Finally, we leverage these insights to build a prediction model and show promising results to identify users' backing patterns at a very early stage. The datasets used in this paper are available at: https://go o.gl/ozgLvP.