Another way to start classifying random metrics, and if they're flat, is to look into Petrov classifications if you have not OP.