12-12: Hsiu-Yuan Tsao
The Impact of Campaign Personality on the Crowdfunding Project Success
Crowdfunding is currently one of the most important fund raising platforms for individuals and organizations. Because of crowdsourcing’s overwhelming success, the key factors underlying this model demand attention. At the most obvious level, crowdfunding projects succeed because they bring value to investors. Researchers have employed a number of different viewpoints in examining this topic, such as the relationships to project creators’, Twitter followers, quantity of money raised as indicator of success, project descriptions, frequency of progress updates and project duration. Many previous studies use quantitative factors to analyze crowdfunding success, but the predictive power of such approaches is questionable. Aaker (1997) uses consumer association to brand, company image, and other relative attributes to evaluate brands—finding five different dimensions to evaluate brands. Brand personality is key in understand how brands affect consumers. Some researchers further show brand personality affects the perceived quality of a product, thus influencing purchase determination. The current study attempts to explore whether successful crowdsourcing projects have consistent brand personality within their campaigns. We conduct computer-aided scaled directed sentiment analysis of consumer’s comments (text) for 200 projects from two of the most popular categories, Design and Technology, on the crowdsourcing platforms FlyingV and Zec Zec in the year 2018. Previous research focuses attention on updates, duration, social media, and funding levels as predictor of success. We extract the salient features of the five dimensional brand personality, along with other numerical data of previous used. Logic regression and cutting edge ensemble algorithms of machine learning are employed to compare the reliability of prediction and the relative importance of features in determining project success. Ensemble methods train multiple learners to solve the same problems. In contrast to ordinary learning approaches that construct one learner from training data, ensemble methods construct a set of learners and combine them.
Results show sincerity, sophistication, and ruggedness significantly affect project success. While the traditional logic regression has around a 70% prediction rate, the ensemble machine learning algorithm, significantly improves on this with a prediction rate of up to 80%.