🧪
Knowledge Challenge
A friend thinks you can answer this question about ML Engineering Practice
Your data science team trained a fraud detection model that achieves 0.94 AUC offline. Three months after deployment, fraud losses are UP, not down. Most likely cause?