This chapter introduces different performance metrics to quantify the success of clinical tests and how well predictive machine learning models perform. This chapter shows how to compute metrics like sensitivity and specificity, which are especially important in domains like healthcare, where false negatives or false positives may have important consequences. Then, the receiver operating characteristic (ROC) curve is introduced, which provides a graphical representation to determine an appropriate cutoff value and can be used to assess the performance of a predictive model. This chapter ends by illustrating the concept of validation, which is an essential step to ensure that a clinical test, or a model, generalizes well to unseen data.