Nothing is more common, or more useless, than a precisely accurate answer to the wrong question. Too much of what passes for policy analysis starts with finding available datasets and running them through a regression program, rather than asking “What is it we actually need to know to make this decision?”

A back-of-the-envelope calculation (BOTEC) starts with the question and then uses the best information available – even if that information is only approximate – to figure out the relevant answer. Sometimes that sort of rough calculation leaves the right course of action in doubt. That’s the time to roll up your sleeves and go collect some new data or do fancier analysis.

But, as often as not, the answer is clear once the key factors have been written down. Once you know that Option A is roughly 10 times as good as Option B, figuring out whether it’s actually 10.13926 times as good or only 9.81352 times as good is pretty much a waste of time compared to figuring out how to put Option A into practice.

(source)