Choosing how many bins to include in a histogram can be a tricky design decision. There are many articles out there that recommend algorithms or rules of thumb for calculating the “optimal” number of bins, however, I don’t think that any calculation can do this reliably. In this post, I argue that the “optimal” number of bins depends mostly on the specific insight that needs to be communicated about the data, and not on the nature of the data (number of values, standard deviation of the values, etc.)