This question of whether a beam-type powder scale (really a balance, as pwc correctly pointed out) needs to rest on a level platform comes up from time to time. The last time it did, I took my 50-yr-old RCBS 510 and ran a little test. First, (again as pwc correctly pointed out), the side-to-side leveling of the bench is completely unnecessary - as long as you can zero the beam - because the beam operates independently of the frame. I knew that would be the case but tested it anyway, placing the scale at a number of obviously odd angles, using one of my check weights, and finding no difference in either accuracy (the ability to get a known weight correct) or precision (the ability to reproduce the same weight multiple times). (If there are any other scientists out there, yes, I know strictly speaking a balance measures mass, not weight.)
I next checked the front-to-back leveling question, propping up the frame of the scale at a few angles that anyone could see were clearly not level - I didn't attempt to measure, but I would say up to around 15 degrees or more off level - and was somewhat surprised to find that there wasn't any difference there either. Now, I'm sure that at some point the front-to-back angle would become so severe as to impact performance, but there's apparently no need to get the level out and agonize over it. And, of course, other types/brands of scales may perform differently.