Currently the star assignment logic is a little bit clunky. A clean 4-quartile cut means that depending on the distribution, two features with a 4 star value could vary significantly under the hood. We should look at the actual distribution and see if there are clearer cuts to make.