The role of Twitter in our crowdfunding success
Here’s our super secret recipe for crowdfunding a science project. Keep in mind that crowdfunding is more like cooking than baking. Our proportions may not be to your liking, so feel free to substitute creatively. However, there are some essential ingredients, beginning with a pinch of angels — family and enthusiastic early pledges (2%). Next, add some Facebook friends (25%) from the various phases of your life, along with some Twitter followers (25%). Don’t forget those evangelists on both Facebook and Twitter who tirelessly promote your cause on their news feeds (8%). Finally, round it out with a healthy measure of strangers (40%) who learn about your crowdfunding campaign from print, online and social media coverage.
In a follow up post, I will speak to that 40% of Crowd4Discovery (C4D) supporters who are “out of network.” In this post, I’ll finish the preliminary analysis of the in-network 60%. During the C4D campaign, I blogged about the distribution of donors in my Facebook network. Two results jumped out at me. First, my “science” friends turned out strongly, and by science friends I mean current or former scientists whom I met over a decade of academic training. Second, donation size was more a measure of individual cash flow or personal resonance with the project, and less a reflection of the strength of our friendship per se, or where donors are located in the network.
Thanks to Tony Hirst, who kindly generated my Twitter network data files on December 3, 2012, I was able to generate the above graphical snapshot of my dynamic follower network, at the time numbering 1,315. Keep in mind that unlike my Facebook network, the above is a directed graph, which means the connections, or edges, are one-way — A follows B, but B doesn’t necessarily follow A. I’m the biggest node (in blue). Like all the network graphs I’ve visualized using Gephi, the size of each node is directly proportional its degree, or connectivity.
98% of my followers — the other 2% are orphans — comprise 6 modularity classes, or clusters, as shown below:
These clusters represents distinct online scientific communities to which I belong, with the exception of an orphans cluster (blue), which contains mostly non-scientists. I first ascertained whether C4D donors are spread out randomly across my follower network or concentrated in specific clusters:
The total turnout of my Twitter network was 10%, which is lower than the 17% turnout of my Facebook network. The distribution of donors in my follower network (right; donors in yellow and non-donors in blue) appears to be random with respect to cluster membership (left), with some clusters performing slightly above or below the network average. Interestingly, there is no correlation between the amount donated and donor degree, just as we observed no correlation in the Facebook network data:
One regard in which donors in my follower network behaved differently from donors in my Facebook network is the timing of their donations. It’s fair to say that our closing surge was stoked by a fusillade of retweets and mentions. 38 of my “tweeps” turned out on the last day of the campaign, and almost half of the 124 donors in the final 26 hours came from Twitter:
Our closing surge was more robust on Twitter than on Facebook. It appears that I had “tapped out” my Facebook network over the course of the 52-day campaign. I can think of several reasons why Twitter proved to be so pivotal in the home stretch. First, the ephemeral nature of tweets plays well with a looming deadline, and it didn’t hurt that I tweeted hourly funding updates in the final 8 hours of the campaign, adding to the sense of urgency. Second, the amplification potential on Twitter due to retweeting is higher than that of Facebook. Even though one can technically share content on Facebook, we never got much traction with Facebook sharing.
What do you think? Did we use the Twitter megaphone effectively? Is 10% Twitter turnout an underperformance, and overperformance, or right on target?