Only by revenue? If so, that's more classification than segmentation as segmentation usually has multiple variables. If you're looking at just one variable and specific percentiles start with PROC RANK and GROUP=100. That will give you the percentile for each value and you can then create your groups as desired. proc rank data=sashelp.cars out=want groups=100; var mpg_city; rank mpg_city_rank; run; *see distribution of ranks; proc freq data=want; table mpg_city_rank; run; This will create the ranking variables and you can then filter - 0 is lowest and 99 is the highest usually. ... View more