Does the Similarity Pattern among Crowdfunding Projects Help Design a Winning Project?
(A2020-58013)

AUTHORS

Jihoon Hong, University of Southern California; Yanhao Wei, University of Southern California; Gerard Tellis, University of Southern California

KEYWORDS

Crowdfunding; Similarity; Word2Vec

ABSTRACT

While crowdfunding has soared in popularity, creating successful projects is a major challenge. We explore the similarity pattern among prior projects to help improve success in raising funds. Applying machine learning techniques, the authors measure the similarity among projects on Kickstarter. This measure allows us to analyze each project as a creative combination of prior projects. We find that (i) the funding success of a new project has an inverted U-shaped relationship with the project novelty, (ii) the optimal funding goal is slightly above the average funds collected in prior similar projects, (iii) a balance between atypical and conventional imitations of prior projects boosts a new project’s success, and (iv) information on prior similar projects significantly improves the out-of-sample prediction of new project success. The authors implement these findings to generate actionable insights for project creators and crowdfunding platforms.