What’s a giraffe got to do with philanthropy?
We believe that with access to good data, organizations can frame problems better and drive more strategic solutions. That’s why 2016 featured the launch of our new Enterprise Database Management System (EDMS), which will allow us to bring in at least 10 times more data into Foundation Center’s products and services. This will give users of our products and services unprecedented access to the largest amount of clean, coded, philanthropic data ever available before.
In 2016, we were able to code 2 million grants, up from about 300,000 in years prior. Most of these grants were processed through autocoding, which allowed staff time to refocus on summary analyses and data mining. It also enabled — for the first time — processing of grants of less than $10,000, which previously had only been included if they were directly reported by foundations.
“For small nonprofit organizations looking for local funding opportunities, this is a gamechanger,” notes lead data scientist Bereketab Lakew.
Now that we have access to machine-readable versions of many 990-PFs, as well as the ability to process these data files in bulk, we expect to dramatically increase the number of small grants in our database from smaller foundations. This will enable us to expand our reach into, and relevance among, smaller cities and rural areas where philanthropic activity was rarely tracked.
The data processing innovations didn’t stop at grants data. We also began connecting our other data sets (including news articles, blogs, research reports), coding them to our Philanthropy Classification System. This enables users to search all of Foundation Center’s knowledge, and also helps us spread the usage of our revised and updated taxonomy to other institutions, like GuideStar, that have expressed interest in coding their data consistently with our taxonomy.
This plays out a step further: our data scientists built a prototype that identifies relevant philanthropic news articles from large, online repositories before coding it into our Philanthropy Classification System, with the goal of making it publicly available online through our products and services. Prototypes like this have the potential to provide the sector with more easily accessible, comprehensive, real-time views of philanthropic giving trends. So far, the results have been promising, but many questions remain before we can make this a viable service in 2017. This is one example of the data and technology experimentation we’re committed to in order to create better, useful, and accessible knowledge for the social sector.
Believe it or not, evolving our technology systems has been fun. Berekatab shares, “With machine learning algorithm, you get some funny results while the computer is learning, and I like to ask why. For example, the Support Vector Machine (SVM) thought the word giraffe was more associated with ‘philanthropy’ than ‘wildlife conservation’. If you ask, ‘why would it think that?’, you see that there is actually a recipient called Giraffe Project, and all their grants are about philanthropy. There are interesting things like this when you turn words into data. I want to look deeper into these patterns of learning this year.”