When I was in college, no joke, we all used slide rules because there was no such thing as a usable pocket calculator until the early 70s when HP started the trend. A couple of well-heeled friends bought HP35s and started taking them to class (in those days, $400 was about a month's wages for a new engineer, let alone a starving college student). They were so much faster than the slide rules that all the departments had to institute new requirements to prevent the lucky ones from blowing the rest of us away during tests. Until calculators dropped enough that we could all afford one, the law was generally "no calculators on tests"; in fact, some of the professors set up the calculations so that there was little to no advantage in using one except for precision. The goal was teaching engineering, not number crunching.
But enough people were carrying the HP35/45 by 1973 that I had plenty of chance to use one. We had a couple of impromptu demos where a group would try to solve chained calculations (circuit meshes, chemical conversions, etc.) using AOS and RPN calculators; invariably the RPN users were able to outrun the others, and that decided me. After a little familiarization, I realized that the standard non-stack-based calculators actually wasted keystrokes, and if you needed an intermediate calculation's results, you lost a keystroke or two every time you had to access it. With the HP, you could just take numbers as they were given to you and keep calculating. Think you need another intermediate result level? Just hit Enter again, and if it wasn't necessary you didn't have to decrement the stack.
The early calculators which attempted to substitute parentheses for the stack were a real kludge - they expected you to know how many levels the calculation was going to require before you started - or write it down first.
Plus, different brands did different things with the "=" key that affected how you completed a calculation. Admittedly, arithmetic calculators have gotten better since 1972, but the stack-based HP architecture is still as natural today as it was then. Part of the reason I still have all my HP calculators is that I prefer RPN when I'm doing work even today. I still have my Post Versalog too, but more as a reminder of how far we've come.
Incidentally, slide rules (if you used them efficiently) behaved very much like a stack-based calculator. If you used the C and CI (or D and DI) scales alternately, it saved a step for every calculation in a chain...