Since our inception over 10 years ago, ideas42 has implemented over a hundred projects in 45 countries. Because of our focus on testing, this means we have also worked on a similar number of impact evaluations, including dozens of RCTs. Much of this effort is showcased in our blog, where we tell the story of our work, highlight our project findings, and provide tips on how to approach challenging behavioral problems.
But as time goes by, we’ve realized that there’s another facet of our work that is slightly different and not as visible: call it the backstage of our behavioral science work and our testing efforts, if you will. How does our analytical work get done? What insights do we uncover in the process? What tools do we use, and how do we apply them?
We think this data carpentry aspect of our work is interesting and worth sharing, so we’re creating this blog as an outlet for it. You can think of it as a behind the scenes look at our analytical work before it’s polished and published.
Of course, this idea is not new. Many other organizations have created their own data blogs. We actually draw inspiration from a few of these, including Urban Institute’s wonderful Data@Urban, and the World Bank’s Development Impact blog, that recently turned 10.
Our ultimate goals are simple: we hope to help make our behavioral science work more transparent and open, to share some of our analytical practices so that whoever is interested can use them, and also to learn from what others are doing and contribute to the bsci community. Of course, we also expect to have fun in the process; you could say we’re intrinsically motivated to do this.
What can you expect from this space? Over the next few months, you may find pieces on how we apply natural language processing to complement our behavioral diagnosis tools, tutorials on using web scraping libraries in Python and R to collect policy data, or even how we used adaptive experimentation to find the best messaging treatment and reduce vehicle booting in New York City.
-the data42 folks