I banged my head against the wall figuring out how to sort a simple bar chart of a category (which graph builder can automatically count and display the proper bar heights for, but gives no sorting abillity by that count...). All the articles out there talk about sorting by some other continuous variable dropped into the merge zone of the graph builder http://www.jmp.com/support/notes/42/770.html But I just want it sorted by the count of each category (you know, the ones that Graph Builder already counted! This is something matplotlib in python does automatically, btw, when you give it categories and counts). One solution out there is to make a summary table of that category...ugh, why make more tables to do such a simple thing!?
So the easiest and most intuitive solution for me is to create a new column of 1's in your data table. Then whenever you chart a category, you drop the 1's (i called it count) column into the merge zone of the axis and choose sum as the statistic for sorting. Now all your categories can be bar charted and sorted by count.
Yes, you could add ANY continuous column to the merge zone and choose N as your sort statistic, but I think this is odd because the chart then says "category sorted by 'unrelated continuous variable'" when it's really sorted by the counts (hence calling my column of 1's count). Yes, you could edit that text but holy cow, we are just applying band-aid after band-aid here. Also you would want to make sure your column you are sorting by has data in every row or the count's will be wrong. So again, I think the most intuitive solution is a column of 1's. It's elegant and ridiculous, just like JMP!
Why graph buidler can't just do this on it's own is another wtf moment for JMP but hey I'm on 12, so maybe 13 can do this?
Hope this helps, and please do enlighten me if I am missing something that makes this all easier/irrelevant.