All of China’s top tech firms today offer consumer credit through their apps and websites, but back in 2015 the idea was new. That was when JD.com (NASDAQ: JD), China’s second-biggest online retailer, launched a new personal credit offering to increase online merchandise sales. The idea was to extend small personal loans of about $1,000 on average for their customers. Problem was, the company was unable to underwrite unbanked consumers composing the vast majority of its customer base. At the time, the Chinese state-run credit bureau only covered about 20% of Chinese citizens
There was another way to solve the problem—machine learning and lots more data—and JD.com and ZestFinance were two of the first companies to apply it at scale to underwriting credit risk. JD is the world’s third largest Internet company by revenue and invests much of its profit in wringing out inefficiency in every step of the consumer purchase.
Although JD is a leader in deploying AI and robotics, it turned to Zest in 2015 to partner on building a new machine learning underwriting model to accurately score consumers that lack a traditional credit profile. Zest helped JD.com’s data team identify and select the most predictive credit signals from among tens of thousands of data points gathered from shopping behavior and website data.
For example, applicants that play games on their phones during the day were found to be a higher credit risk (presumably because they were unemployed and had a lower income). The teams used ZAML to generate new features for the model, such as building submodels to verify an applicant’s information based on the applicant’s web browsing or order history. Testing and model validation showed the new ML underwriting could responsibly score segments of the unbanked population that looked substantially similar to the banked population for which they had credit bureau data available.
JD.com’s lending profits dramatically increased as a result of the model’s success. It has since renamed its JD Finance group as JD Digits to bring the power of ML and more data to improve efficiencies throughout the company and in other industries, as well.