Global Giving just upgraded its website and alongside that I came across Features of Great Organizations , a blog post by friend Marc Maxson. I’ve had contact with Marc over the years and does some pretty neat, creative data analysis that GlobalGiving uses to drive its work to try to help more money get to the best grassroots organizations.
Today I took 1,325 evaluations from hundreds of in-the-field travelers to GlobalGiving projects over the past 5 years and analyzed them. BigML, the DIY machine learning site, allowed me to understand what defines a great organization in about 15 minutes. That alone is cool, as this type of analysis would’ve taken weeks just 5 years ago! Given one column in the data that represents the outcome you want (or don’t want) to achieve, BigML organizes the rest of the data into a branched contingency tree, like this:
Reading the tree reveals which other questions in the evaluation are the most reliable predictors of answers in that primary outcome column. Statisticians run something similar called a principal component analysis. The labels at the top of each branch of the tree define what makes a great organization, apart form an average or a poor one.