Also consider that if you have a 100M row table and the column has only 4 values, but those values are highly skewed (ie 3 values are 99% of the rows, 4th is 1%

Also consider that if you have a 100M row table and the column has only 4 values, but those values are highly skewed (ie 3 values are 99% of the rows, 4th is 1%), then a function index on the skewed column can be very useful: create fbi on table(case when skewed_col == 'small value' skewed_col else null end) The null values are never indexed, so now you have a nice small index that is good only for finding you 'needles in the haystack', so long as you put that case statement in your queries.