Anatomy of a crowdfund: the homestretch

November 19, 2012

We are in the final week of campaigning for Crowdsourcing Discovery! As of this writing, we’ve raised $18,246 from 258 donors, which is 74% of our stated goal of $25,000.

 

With 48 hours left the $6,754 question is – can we go all the way?

 

Here’s a plot of money raised over time, which I introduced several updates ago:

The black dashed diagonal line represents an idealized constant rate of donations over time, such that a quarter of the goal is raised in a quarter of the time, and so on. The ideal rate is $481/day. The orange line charts our actual progress. The observed average daily rate now stands at $372/day.

 

For the first two weeks, we over-performed relative to the diagonal, averaging close to $1,000/day post-launch. We were outstripped by utopia in the third week, and thereafter we steadily grew, with a few spurts along the way (see day 40) to even out the really slow patches, but a rate too slow to reach our goal unless it were to accelerate. Basically, we need a hockey stick finish.

 

How often do campaigns asking for $25,000 fund all the way when they need to raise 25% of their goal in the final 2 days?

 

One unfortunate fact I’ve learned during our campaign is that fundraising stats on specific crowdfunding projects are in short supply, making comparisons let alone projections difficult. I’m aware of at least one thorough post-game analysis. However, analyses based on averaging many fully funded projects are available on the blogs of crowdfunding portals like Kickstarter and IndieGoGo. This past summer, the folks at IndieGoGo blogged about a well-known feature of successful crowdfunding campaigns that seems to hold across project topic areas: 50% of contributions are made in the first and last 10% of a campaign.

 

Here’s IndieGoGo’s aggregate data showing funds raised as a % of goal (bottom axis) by % of campaign length, broken up into 10% chunks:

 

 

This is what is meant when people say crowdfunding campaigns are “U-shaped.” In the above rendering the U has fallen over on its right side, forming a C: an initial wave of contributions is followed by a long trough that gathers into a closing upswell that exceeds the opening burst.

 

Here’s that same plot with our data as of the start of this week. Keep in mind that our campaign is 52 days long. We’ll need a strong surge (yellow) in the homestretch, and with 2 days left we actually only need to raise 25% of our goal:

 

 

If I can indulge, Nate Silver-style, in numerically informed prognostication, our challenge boils down to finding 100 new donors, half of whom need to be strangers. I came up with these projections based on the following two pieces of data. First, the average contribution to our campaign has hovered around $70 since the outset: the remaining $7k divided by $70 equals 100.  Second, approximately ½ of CSD’s 258 donors are people I know IRL and ½ are people I don’t know, including online-only connections.

 

So can I realistically hustle up 50 friends? Using the open source graph program Gephi, I created a graph representation of my Facebook network. The size of each node is proportional to its degree, or the number of connections to other nodes. In other words, the largest nodes are hubs and the smallest nodes are orphans. The network on the left has been color-coded to reflect modularity classes, essentially friendship clusters. For example, my graduate school friends comprise the red cluster. The network on the right has been color-coded to reflect donors (yellow) vs non-donors (blue).

 

Two things jump out at me. First, the distribution of donors appears to be random with respect to degree, which means both hubs and orphans and everyone in between are donating. Second, the distribution of donors also appears to be random with respect to modularity class, which means there are donors from different phases of my life. I might have hypothesized before we started that all my science friends would donate but not my non-science friends. Interestingly, that doesn’t seem to be true, suggesting that the content of our project matters more than membership in a specific clique. In the remaining hours of our campaign, I will be curious to see how late-donating friends distribute across my network, and whether there will be evidence of a contagion effect.

 

And for my fellow data nerds, here are the daily box scores through the end of week 6 (note that I will be posting these data on figshare as soon as our campaign ends):

 

 

As a parting thought, everyone I’ve talked to in these last few days has wished me good luck, which is incredibly uplifting after a 50+ day campaign, I can promise you!

 

The thing about getting lucky is that it actually happens from time to time. Take tweeting, for example. My two best tweets of all time, judged by the number of times they were retweeted, are two random remarks that at the time seemed just like everything else I ever tweeted, but they get incredibly loud on the human mic:

 

 

The number of retweets (x-axis) and the number of impressions (y-axis) are plotted. Impressions are what the Twitter analytics tool called Crowdbooster calculates, in analogy to a Klout score, I suppose.

 

In yellow, my all-time best tweet is:

 

 

The title of second-best tweet is currently held by:

 

Getting a tweet with a link to our donation page to be retweeted over 100 times in the homestretch would constitute getting lucky.

Chance favors the prepared mind, I’m told.

 

For a summary of week 5, please go here.

For a summary of week 4, please go here.

For the summary of week 3, please go here.

For the summary of week 2, please go here.

For the summary of week 1, please go here.

And for the summary of the first 96 hours, please go here.

 

Related Posts

LucidNYC talk
Fall crowdfund drive
Indie science