How Comments Can Make or Break Your Crowdfunding Campaign
With the rise of internet financing, crowdfunding has become an important channel for many small and medium-sized enterprises and individual entrepreneurs to obtain funding. However, the common “all-or-nothing” outcome of crowdfunding campaigns is often a painful hurdle for entrepreneurs. Professor Yen-Chun Jim Wu from National Taiwan Normal University’s Graduate Institute of Global Business and Strategy used algorithms to conduct sentiment analysis on campaign comments made before and after fundraising. His research explored the predictive power of online comments for crowdfunding outcomes. The findings show that incorporating early-stage comments during a campaign can significantly increase the likelihood of crowdfunding success.
Wu specializes in research on supply chain management, innovation and entrepreneurship, and technology management. In this study, he led his research team to focus on analyzing comments posted on crowdfunding platforms, with particular attention to how “pre-campaign comments” influence fundraising outcomes. The team drew on data from 126,593 projects and more than 5.4 million comments on the well-known U.S. crowdfunding platform Kickstarter. Using sentiment analysis tools such as fastText, and leveraging deep learning algorithms at the core, they enhanced the accuracy of their predictions.
The crowdfunding process is fraught with uncertainty, and figuring out how to succeed is a critical challenge every entrepreneur is eager to solve. According to Wu, the timing and content of comments play a decisive role in crowdfunding. Typically, comments on crowdfunding platforms fall into three stages: first, those made before a project is launched; second, those made during the fundraising period; and third, those made after the campaign ends, which often reflect user feedback and experiences. This study focused on the first two—comments made in the “pre-funding” stages.
Wu explained that although most comments occur after a campaign ends, “pre-campaign comments” are often based on backers’ initial impressions or expectations of a project and are therefore more effective in terms of predicting its likelihood of success. The study found that as crowdfunding progresses, user sentiment toward projects tends to become increasingly negative. This pattern appears in both successful and unsuccessful campaigns, although with varying intensity. However, incorporating early-stage comments significantly improves both prediction accuracy and recall.
The “all-or-nothing” model of crowdfunding means that if a project fails to reach its target, both the entrepreneur and the platform suffer a loss of resources. This study offers clear guidance for entrepreneurs: greater emphasis must be placed on managing comments during the project preparation stage and early fundraising period, while also considering whether early comments convey a sufficiently positive sentiment. On the other hand, the research also provides significant value to crowdfunding platforms. By analyzing shifts in sentiment within pre-campaign comments, platforms can more accurately predict a project’s likelihood of success, identify high-potential campaigns, and in turn allocate resources more effectively, thereby improving overall operational efficiency.
In addition, the study also yielded new insights for researchers. The team tested different algorithms and found that the deep learning algorithm Feedforward Neural Network (FNN) performed best in sentiment extraction and prediction accuracy. Wu noted that in previous studies, there were various approaches to extracting sentiment-related terms and applying algorithms. Through multiple rounds of testing, his team was able to offer recommendations to other researchers regarding algorithms and text-mining tools—constituting the study’s contribution on the academic front.
The research process was not without its setbacks. Wu admitted that the study carried a high degree of uncertainty. From defining the research questions and hypotheses to processing massive amounts of data, the team could not feel entirely at ease until the analysis was completed. He explained that the study itself was also “all or nothing”: if the results lacked statistical significance or turned out to be too ambiguous, no matter how the report was written, it would attract little attention.
Many factors influence the success or failure of crowdfunding campaigns. Speaking about more advanced research, Wu Yen-Chun noted that people often hope to solve most problems with minimal effort—for example, whether the campaign’s theme alone can predict outcomes, or how entrepreneurs respond to online comments. In some cases, such as a restaurant receiving negative reviews, an appropriate response can generate greater returns and contribute to larger success. Conversely, a poor response can make a bad situation even worse. These are all potential directions for further, focused research.
Source: Wang, W., Guo, L., & Wu, Y. J.(2022). The merits of a sentiment analysis of antecedent comments for the prediction of online fundraising outcomes.Technological Forecasting and Social Change, 174, Article 121070. https://doi.org/10.1016/j.techfore.2021.121070
Yen-Chun Jim Wu Professor | Graduate Institute of Global Business and Strategy
Ph.D. from the University of Michigan, Ann Arbor, and a faculty member of the Graduate Institute of Global Business and Strategy. He has been appointed by Taiwan’s Ministry of Education as a visiting scholar at UC Berkeley and as a senior Fulbright scholar in the U.S. To date, he has published nearly 350 papers, including approximately 200 SSCI/SCI articles in 70 leading international journals. He has also served as a reviewer for nearly 50 SSCI/SCI journals and as (associate) editor for five international journals. His recent research focuses on supply chain management, innovation and entrepreneurship, and technology management.