DevOps for Data Science – Part 9 - Application Performance Monitoring

Because we are talking data science, how should we think about measuring the DS model performance? For instance if the model was a text classifier, is there a method to measure how it is performing in terms of classifying text, I have read that model performance can degrade over time as the date being sent them deviates from the original training data?