Got a Kickstarter project? This man can predict within four hours if it will fail

Within four hours of launching a Kickstarter project, three researchers from the Swiss university EPFL can tell whether it will succeed – with 76% accuracy (PDF).

According to their paper, presented at the Conference on Online Social Networks, Vincent Etter, Matthias Grossglauser and Patrick Thiran beats the current best estimates by 8 percentage points. An earlier paper by an American team can predict success with 68% percent accuracy.

By combining information on the interest shown on the project on Twitter, the previous projects backed by supporters, and the number of first-time backers – as well as the all-important figure of how quickly money is coming in – the Swiss team achieve a marked boost in accuracy just a few hours into the project.

The academics have also built a tool, "Sidekick", applying their theories to live Kickstarter projects. HexFlex, a project to build a snowboarding adventure tool, launched eight hours ago and has already achieved 17% of its $10,000 funding, and Sidekick gives it a 92% chance of success. But a four-song EP by indie rock artist Terri Moeller, launched at the same time, is only given a 16% chance, despite having already raised nearly a third of its desired $2,157.

“The model reaches 85% accuracy after only 10% of the campaign [duration],” says Etter. “The accuracy is of course close to 100% near the end of the campaign, but this is pretty useless, as there is not much you can do at that point.”

In a way, higher accuracy is just a way of breaking artist’s hearts quicker. If a project has failed to take off in the first few hours, the chance of it recovering is low. “It is not impossible to recover from a slow start,” Etter adds. “There are projects that took a while to take off, but that eventually made it. Nevertheless, I would say that if your project has still not taken off by the middle of the campaign, your chances are quite low.”

The key difference between the Etter’s approach and the earlier paper, led by Michael Greenberg, is that the American team only looks at information available before the Kickstarter has been launched. It uses a large array of categories, including the project category, whether or not the campaign has a video uploaded, and how many rewards are offered, to calculate the likelihood of success.

Taking them together, Etter offers some advice to would-be crowdfunders. “As shown by Greenberg's work, it is really important to put a lot of effort into the campaign itself: description, introduction video, rewards, etc. But once it has launched, you should try to maintain some activity on social networks.

“This sounds trivial, but generating tweets about your project, by sharing updates, engaging backers, etc., is the best way to create attention and ultimately get people to back your project.

“Moreover, one thing that comes out from the many descriptive studies on Kickstarter campaigns is that campaigns that fail usually ask for more money, over a longer time period (with the exception of video games, for which successful campaigns have a higher goal on average). So you should be realistic about what you ask.”

But would-be inventors should also be realistic about their chances of success overall. Overall, only 44% of Kickstarter campaigns succeed, which means that simply assuming that every campaign will fail will still be correct more than half the time.