Also, data cleaning is the wrong image, I think. Often it's not just about cleaning but figuring out if it is the right kind of data, has the information you need, etc. and then actively looking for better data. It's not like you get a dirty slump of data and that's it.