The crowdfunding space is now a multi billion dollar industry, and nailing the pitch is hugely important. Using data science, researchers from the University of Toronto’s Rotman School of Management have discovered the top reasons for crowdfunding campaign failures — information applicable to both project creators and startups.
The Power of Machine Learning
In the competitive world of crowdfunding where every dollar pledged counts, project creators must line up a few ducks in order to run an efficient campaign.
Conventional statistical tools have their own set of drawbacks, usually adhering to linear linkages among the variables and the output. The U of T paper integrated machine learning with this wider view of crowdfunding, which the Rotman School of Management researchers chose to embrace in their research.
As it turns out, these machine learning models were hands down better at predicting campaign success than any conventional methods we tried and more importantly they identified exactly what the most responsible factors are in succesful proposal products In considering the functional relationship between these factors, the researchers were able to uncover insights that would be imperceptible using regular regression analysis.
The Secrets of Crowdfunding Done Right
Revealed in a study published recently in the Journal of Business Venturing Design, these 7 crucial factors effectively determine the success or failure of any crowdfunding campaign from among over 100k Kickstarter projects analyzed.
Perhaps the most interesting of all was the actual goal of the campaign itself. Funding success was surprisingly stable up to the $100,000 goal level, contrary to expectation while showing a more pronounced drop at larger goals starting around $133,300.
Another key element was the social capital of the creator, as quantified by how many comments the project had. The machine learning models suggested a relationship in which success increased with social capital up to about 750 comments, at which point additional comments ceased appearing to have any meaningful impact — an effect more complex than the linear one found using standard regression.
The researchers also explored the optimum number of rewards an individual is offered and how long a reward crowdfunding campaign should be open for. Their interpretation of the data pointed to campaign length being most effective at 10 days (around 25% funded) but almost as successful if extended up to 15 days, and rewards increasing in success moderately between a single reward and around 15-20 rewards, then a marginal decrease potentially up until around 20 (from personal experience this seems logical since after your first few it gets harder to think of something original and useful), followed by a cycle of increased chance thereafter leading them to conclude that having between about 20 and maxing out at just over fifty reward options was generally best for campaigns.
Conclusion
This study underscores the value of machine learning in revealing some of the key drivers behind crowdfunding success. This paper highlights sweet spots for campaign goal, social capital and reward options, offering a roadmap to optimize crowdfunding strategies of project creators and startups towards winning funding target. With the crowdfunding market expanding, these insights may be invaluable to entrepreneurs and innovators who want to bring your dreams to life.
