Looking at a measure at centrality implies you understand the distribution is represents. I find it good practice to either observe the distribution before any

Looking at a measure at centrality implies you understand the distribution is represents. I find it good practice to either observe the distribution before any derived measure, and if that's difficult for some reason, get a few measures of centrality, mean and median if you can. The further they are apart, the unlikelier it is you're dealing with a normal distributed quantity, an assumption people make far too often to my taste.