Crowdfunding, which started small in the late 1990s, has since exploded into a major source of funding for new ideas. Kickstarter, one of the biggest platforms, saw its total pledges grow from $276 million in 2012 to $7.8 billion in 2024. The stakes are now so high that some people hire experts to help design the perfect campaign.

In this competitive landscape, a good pitch is crucial. This is where machine learning comes in. Researchers from the University of Toronto’s Rotman School of Management tested several machine learning models, including deep learning, to predict which campaigns would hit their fundraising goals. These models didn’t just beat traditional statistical methods; they also helped identify the key factors behind a campaign’s success.

All or nothing

Kickstarter uses an all-or-nothing system, meaning project creators only get the funds if they reach their goal. The researchers looked at data from over 100,000 campaigns and found that the target amount had the biggest impact on success. Social capital (like the number of comments), the number of reward options, and the campaign’s length were also important.

Machine learning provided more detailed insights than traditional methods. For example, campaigns with a goal of up to $100,000 had a decent shot at success, but chances fell sharply after $133,300. Conventional models, on the other hand, wrongly predicted that higher goals always reduce the likelihood of success. Machine learning was better at capturing the complex relationships between different factors.

The models also showed diminishing returns in some areas. For instance, while more comments initially helped a campaign, the benefit stopped after about 750 comments. Likewise, campaigns lasted longer than 15 days saw less chance of success. The ideal number of reward tiers was around 15 to 20, with gains tapering off after 50.

Using text analysis, machine learning could dig into details that standard statistical tools couldn’t. It revealed, for example, that “gadgets” tend to struggle on Kickstarter. And if you’re thinking of crowdfunding a flyable World War II plane, the odds are stacked against you.

In today’s crowded crowdfunding world, data-driven insights like these can help project creators improve their chances. Instead of relying on instinct, the evidence suggests that machine learning should guide the way forward.