Hi Juan, this looks like a totally reasonable way to do it. I’m curious about the size of the dataset you ran this on? How long did it take?