Test results are the ultimate tool for benchmarking and improving a custom-trained NLU model. This course introduces the various kinds of issues we can find here and how to resolve them, improving prediction accuracy.