Back of the envelope calculations are not meant to be very accurate - but they are meant to be better than random guesses. The process of making one usually helps remove bias from your answer, and should converge to a general ballpark which is close to the answer you're looking for.
Several PMs I've spoken with so far tell me that, when interviewing, while they don't expect a very accurate answer from a back of the envelope calculation for the interviewee, they do expect that the number won't be off by too much - that is, not more than one order of magnitude in either direction.
I was surprised to hear an answer so consistent - multiple PMs from multiple industries, citing exactly the one order of magnitude margin of error.
So, I'm wondering where that one order of magnitude number came from - is it some sort of industry-wide standard? Is there a methodology book or course that introduced it? Or is it just accidental anecdotal evidence from my social circle, and not spread much farther than that?