Crowdfunding is one of the more curious emergent phenomena on the Internet. The idea is that somebody proposes a project, such as creating a smartwatch, an art installation or a video game, and then asks the Internet crowd for money.
Since 2001, crowdfunding sites have raised almost $3 billion and in 2012 alone, successfully funded more than 1 million projects. But while many projects succeed, far more fail. The reasons for failure are varied and many but one of the most commonly cited is the inability to match a project with suitable investors.
That raises an interesting question. What kind of projects attract the most reliable investment? And how do investors find projects that match their own goals?
Today, Jisun An at the University of Cambridge and a couple of pals say they have mined the data associated with a large number of crowdfunding projects to find answers to both of these questions. As a result, they have rained a machine learning algorithm to accurately match proposals with potential investors, like first dates on Valentines day.
The most famous crowdfunding site is Kickstarter. An and co begin by crawling data on projects and investors that appeared on Kickstarter between July and October 2013. That included information on over 1000 projects in the US that were funded by almost 80,000 investors.
During the same period of time, the team also collected tweets containing the word “Kickstarter” and then matched each tweet to a particular project if it contained the project title or a link. This produced a database of over 70,000 tweets. This allowed them to track the “buzz” around each project and the people who invested in them.
By studying the behaviour of investors, An and co divided them into two types: occasional investors who funded fewer than four projects, about half of the total number of investors, and frequent investors who funded more than 30 projects and who make up about 11 percent of the total.
They then mined the data to find out what kind of projects frequent investors were most likely to back. “We find that frequent investors are likely to fund projects that are well-managed; have high pledging goals; are global; grow quickly; and match their interests,” say An and co.
By contrast, occasional investors use a different set of criteria for their choices, acting more like donors mainly on art-related projects. “We suspect that occasional investors are lured into Kickstarter by their own friends and family members who might happen to be on Facebook,” say An and co.
Indeed, they backup this idea by counting the number of friends that founders have on Facebook. “[We] find that project whose founders have many Facebook friends tend to attract occasional investors, while finders with moderate numbers of Facebook friends attract frequent investors, partly confirming our expectation,” they say.
“We might thus infer that those who have supported a considerable number of projects act in ways similar to how investors would do, while occasional supporters appear to be behaving as charitable donors.”
An and co go on to ask whether it is possible to match investors with potential projects simply by analysing the characteristics of each. They use a portion of the Kickstarter dataset to train a machine learning algorithm to spot what type of investors prefer which kind of proposals.
They then use the algorithm to match investors to projects in the rest the dataset. For each proposal, the algorithm produces a list of investors ranked according the calculated likelihood of each one investing. “The best strategy achieves, on average, 84% of accuracy in predicting a list of potential investors’ Twitter accounts for any given project,” say An and co.
That is an interesting result that could provide a useful service for crowdfunding websites. “We have shown that it is possible to match new projects with willing investors, and that is extremely important, not least because the most common reason for failure in Kickstarter is the inability of founders to reach out to the right investors,” say An and co.
Indeed, these guys are currently developing a website that recommends a list of potential investors’ Twitter accounts given the input of a particular Kickstarter project’s URL. No word yet on when this will be available but potential proposes and investors alike should keep their eyes peeled.
Ref: arxiv.org/abs/1409.7489 : Recommending Investors for Crowdfunding Projects
The Physics arXiv Blog