5 Our methodology

At ideas42 we have developed a five-stage behavioral design process. The first step is to define the problem and try to remove any embedded assumptions about why certain behaviors may be occurring. Using quantitative and qualitative methods, we next diagnose what behavioral bottlenecks may be driving the problem. We then design interventions that directly address the key bottlenecks we’ve diagnosed. To test whether our design successfully addresses the problem, we conduct a randomized controlled trial. Finally, if an intervention proves effective, we look for opportunities to scale the solution to a larger population or adapt it to other contexts.

Both of the following case studies use this five-stage process to address a behavioral problem. Below is a short introduction to each. From there, we will demonstrate how adding machine learning for heterogeneous treatment effects and adaptive experimentation changes both the size of the impact we can achieve and our approach to design altogether.