A hotel runs the numbers on its doorman. The job, on paper, is opening a door. A door is a one-time cost with the odd service call. A person is a salary, every month, forever. The spreadsheet isn’t close, so the doorman goes.

What the spreadsheet had no column for: he remembered the regular’s name, waved off the guy who didn’t belong, made arriving somewhere feel like arriving somewhere. None of that was “opening the door,” so none of it was counted, so none of it was saved. It was just gone.

Rory Sutherland calls this the doorman fallacy. You define a role by its most measurable part, you optimize that part, and you quietly throw away everything you never put a number on. The saving is real and it sits right there on the slide. The loss is real but it sits nowhere.

Once you start looking for it, you see it everywhere, and there seems to be more of it every year. The new chatbot that handles most of the support tickets, where “most” is doing a lot of quiet work and the calls it can’t handle are usually the ones that were going to matter. The self-checkout that trims the cashier line. I do have to admit though that I like self-checkout. I use it nearly every time. But I do also notice the small human interactions that got lost with it.

So the question isn’t what a role obviously does, or even what gets lost when it goes. It’s what the role is quietly holding up that nobody will notice until it’s gone.

A doorman holds up trust, status, and a hundred small judgments about who belongs in the lobby, and you learn he was load-bearing the way you learn a wall was load-bearing: afterward, by what starts to sag.

Businesses love efficiency and people love comfort, and most of the time those point the same way. The doorman fallacy is just the reminder that most of the time is not all of the time, and that the moments they split are exactly the ones the spreadsheet can’t see.