We love to build data science and machine learning models by diligently cutting out test data sets, then training and creating models, and finally testing them on the test data to measure performance.