Our first step is to frame a good predictive question, one that can be solved by the data at hand and has measurable, real-world impact. We work with nonprofits to understand their needs and identify productive partnerships.
The next phase is to host an online, open-innovation competition where freelance developers and data scientists submit statistical models. Using our competition platform and evaluation engine, the models are ranked based on how well they predict data withheld from the competitors.
Finally, we want to close the loop. We work with the organization to leverage the top model—as insight, a fresh statistical approach, or a tool for analyzing new data—enabling them to more effectively and sustainably carry out their mission.
Consider a nonprofit microlender. Using data on loans and outcomes, DrivenData would run a competition to predict default. A good model predicts which loans involve the most risk. A better model might determine the loan amounts that minimize the probability of default. Using the winning solution, the lender can decrease negative outcomes for recipients and improve its long-term impact. You can find more examples on our information for nonprofits page.