Why Accuracy Keeps Lying to You in Imbalanced Classification

My fraud detection model had 99% accuracy but caught zero frauds. Here's why accuracy is the wrong metric and what to use instead.