https://en.wikipedia.org/wiki/Fermi_problem
It's also about knowing what facts need to be true for a BI to work. Then we know what facts we need to verify to determine if it's true, in contrast to all the meaningless verbiage in this thread.
Also, Scarmig's comment was quite valuable. It showed exactly what the driving effect is - that's precisely why you should do a back of the envelope calculation. And as all the non-quantitative comments show, it's also why any comments suggesting an effect but not doing such a calculation are worse than useless.