In this project we investigate how entrepreneurs should describe their project in online crowdfunding platform. We scrape data from Kickstarter for our analysis and used Latent Dirichlet Allocation to identify topics from the project description. We then investigate if the appearance of these topics can be related to the comments of the backers, the nature of these comments, and eventually the success of the project o be funded. We also explore if the efficacy of these topics would differ across different project types.