top of page

AI Case Study increases customer loan approval rate by 150% using Zest Finance's machine learning platform

Working with Zest Financial's ZAML platform, was able to apply machine learning to analyse customer data including online browsing and order history in order to offer loans to shoppers who previously may not have been approved.


Financial Services


Project Overview

"Zest Automated Machine Learning (ZAML)'s data assimilation tools allow lenders to acquire, onboard, and prepare massive
amounts of disparate data for modeling. This data can come from external sources. However, it often starts with additional internal
data the lender has but can’t use in underwriting. ZAML’s modeling environment makes it easy for data scientists to train,
ensemble and productionalize models extremely efficiently. Together these tools drastically lower the time and financial cost of ML adoption. And ZAML’s explainability tools solve black box concerns, providing model insights to executives and tools to support analyses needed for compliance. ZAML facilitates the inclusion of hundreds or thousands of variables not used in traditional underwriting. With this additional data, the ML models built with ZAML can produce accurate credit decisions for previously hard-to-score borrowers, allowing lenders to safely approve additional borrowers without increasing risk... used ZAML to increase its portfolio approval rate markedly. Key to this improvement was the ability to incorporate additional data into the new ML-based underwriting process."

Reported Results

Zest Finance claims it "increased’s approval rate by 150%."




General Operations

Background is China's second largest e-commerce company. "ZAML facilitates the inclusion of hundreds or thousands of variables not used in traditional underwriting. The China credit bureau only covers about 20% of Chinese citizens. As a result, traditional underwriting can only approve a small fraction of applicants. This limitation on approval rate increases the per-loan marketing cost and makes it exceptionally difficult to grow a portfolio". The goal of this is to increase lending rates safely.



"The joint JD-ZestFinance modeling team incorporated
browsing data from applicants into the underwriting model. Additionally, 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."

bottom of page