This is less an argument against quantifying performance in general and more just an example of choosing the right metrics and interpreting them intelligently.

I have no doubt that more than a few moronic managers will try to make large decisions based on a trivial surface-level metric time per customer (which, at the very least, needs to be normalized against the complexity of the order), but still, I think the thing to be criticized is poor data analysis rather than the very concept itself. A smart manager would use this data to learn more about how their operations are going and have discussions with employees about what's going on and how various differences might be explained, while a stupid one would just fire the apparently slowest barista every few months, completely missing the fact that said barista has developed personal relationships with tons of customers and only performs slowly during off-peak hours when it makes no material difference at all.