When preparing data for supervised classification ML$,$ which attributes should the data have?

Data should be generated randomly by using mathematical algorithms

Data Should be labeled

Data should contain only instances of the target

Data should be representative of production