Ah, I didn't consider density, that's a good point! Taking a look at the study you mention (linked in other comments), it seems like the escalators they studie

Ah, I didn't consider density, that's a good point! Taking a look at the study you mention (linked in other comments), it seems like the escalators they studied were well to the "too few people want to walk" side of the curve to be efficient. I wonder how close the break-even point is to the simple metric, "do you have to wait to enter on the right?" Or thinking of the queue length for standers, clearly when the length is 0 allowing walking is more efficient (you've taken nothing away from the standers, and the walkers are more efficient). But there is some point where the stand-queue length, let's call it N, is big enough that the corresonding wait-time means overall throughput is lowered by allowing walking. I think it would be quite reasonable to require standing in those situations. So the rule could be something like, "stand on the right, walk on the left, unless there is a queue of more than N people waiting to stand on the right, in which case you may stand on the left." Easy right? ;-) (I wonder if N would vary w/ escalator length? Intuitively, almost certainly. So we'd have to post signs next to each escalator w/ its individual stand-queue limit. :-)