Case Study: Zidisha achieves higher loan repayments and scales micro-lending with DataRobot

Zidisha is Transforming Lives with DataRobot

Zidisha is a nonprofit online microlending community that connects lenders directly with entrepreneurs in low‑income countries. Faced with high cost and time barriers to traditional credit assessment, limited data science resources, and more than 10,000 loan applications per month, Zidisha needed a scalable way to predict default and fraud without hiring loan officers. They partnered with DataRobot to apply machine learning models and real‑time deployment to their loan screening process.

Using DataRobot, two Zidisha executives with no formal data science training built and deployed fraud and default propensity models in under two weeks and put them into real‑time use as applications arrive. In an A/B test the models reduced defaults by 5% (measured in dollars repaid), lowered declines of worthy applicants, and allowed Zidisha to direct more capital to trustworthy borrowers — increasing repayments and enabling the organization to fund more loans. DataRobot’s platform and support also positioned Zidisha to continue improving accuracy as data grows and to scale its AI efforts internally.

Zidisha

Julia Kurnia

Founder and Director