Kickstarter certainly seems interesting; I’ve been browsing their site (and, in doing so, found two excellent little projects that seemed worthwhile and put some cash towards them) and taking some notes on what seems to succeed and what doesn’t. With all the interest in drones and wildlife survey, I was thinking that it might be fun to get a drone project funded through Kickstarter.
In a later moment of creative procrastination (avoiding all the other projects I have to do), I found a database of all the Kickstarter projects from about a year ago – 45,000 of them or so. Data on category, subcategory, goals, and the funded percentages. Did a little exploring in R, hence the not-very-pretty graphs below.
How does a project’s category affect the likelihood of funding?
In the boxplot below, 1.0 indicates funding reached 100% (and project received money) – so the median bar at 1.0 indicates that the projects in that category had a 50:50 chance of being funded. I’ve coloured the ones that reached 1.0 in green, and their categories are written below (tough to see I know); I’ve truncated the Y axis to just 300% of the goal (Kickstarter projects famously raise far more than initially asked):
- Dance, comics, food, theater, art, music & film are at or near 50:50.
- Technology, fashion, photography, games, design and publishing get nowhere near this parity!
How much money gets asked for, and how does that go?
Quantiles for funding goals, all categories:
0% 25% 50% 75% 100%
0 1,650 4,000 8,800 21,474,836
How do those quantiles succeed? Plotting the quantile levels (plus an arbitrary > 20,000) as groups:
Asking for smaller amounts of money looks successful … and the probability of success declines precipitously over $8,800.
You’re either gonna succeed, or not, and no halfway!
That same data is better plotted as two continuous variables against each other … which is where the story gets really interesting. Here I plot the goal size (up to $10,000) on the X and percentage funded (up to 500%) on the Y:
Have you ever seen a distribution like that? Zooming in:
So – there are two sets of projects! Either you’re going to get funded – and then you tend to do pretty damn well (> 100%) or you’re NOT going to succeed, and struggle to break about 10%.
Looking at that graph again, this time with a linear model fitted to the two separate groups:
The ‘funding succeeded’ group (green line) starts at around 400% funding and then tapers off (reaching 100% at around $55,000) – this is not really representative, as the boxplot shows, but instructive. remember that the top of the graph (up to hundreds of times the goal being raised) is not shown here.
The unsuccessful group (red line) is almost perfectly flat! 10.5% of funding gets raised across all levels.
So, we’re interested in technology.
We’ve already seen that it doesn’t do so well (back to the first graph), but what are the subcategories like?
So – better than even chance of success overall for open hardware projects – i.e. those that will build something and then share the plans freely. “open software’ and general “technology” projects don’t do very well.
How much money were they asking for within these categories?
Open Hardware Open Software Technology
$8350 $5910 $10000
So – looks like I could get some Kickstarter help if I ask for relatively small amounts of money, for building something that will get shared widely. If I ask for more than $10,000 I’m not gonna get anything unless it REALLY sells itself – which is where that fund / not fund weird distribution comes in.