Applying AI to Automate Credit Decisions at Facebook
Managing credit limits for Facebook’s advertisers requires thousands of credit decisions annually. We built a machine learning solution to help automate credit decisions. It uses gradient boosted decision trees to predict whether an account’s invoices will become bad debt with 92.4% ROC-AUC. This enabled Facebook to reduce manual credit decision volume by 62% while avoiding revenue loss resulting from accounts reaching their credit limits.
Julie Ward Drew is an Engineering Manager in Facebook’s Enterprise Engineering organization, where she builds AI & machine learning solutions for finance and compliance. She has over 20 years of experience applying optimization and machine learning to enable step-level improvements across a wide range of enterprise applications. Julie was awarded the 2009 Franz Edelman Prize for Outstanding Achievement in the Practice of Operations Research and Management Science by INFORMS. She received a Ph.D. in Operations Research from Stanford University and a Sc.B. in Applied Mathematics from Brown University.