If you go for a denomalized approach to scaling, you might end up wondering where all the duplicates actually are in the system. Do people / can we just store a list of all the places in another record? So that you don't have to make your code hard-coded to know.