The scenario runs the assistant through a set of test cases and shows in numbers how often it picked the right action.