Let’s say you’d like to withdraw cash from an automated teller machine. As you approach the bank, you see three terminals, with a customer using each one. A bunch of other people wait in a single queue, centred midway along the row of ATMs. You observe that one of the customers is very fast, zipping through the transaction; the other two seem to have trouble using their keypads and reading their screen prompts. Do you
- Commit to the “fastest” terminal, waiting directly behind the customer at that ATM
- Stick with the single, central queue, allowing the person at its head to use the first available ATM, whichever it is
If you chose option 2 – thinking only it fulfills the strict first-in, first-out discipline behind this sort of queueing, or maximises average throughput – you’re probably a very frustrated person. Because most of us have other things on our mind than fairness or efficiency in queueing. We may need to get the next train, or be rushing to an appointment, or just not really feel like wasting time in a boring ATM line. Choosing option 1 can – if we cleverly pick the right line – allow us to skip ahead of others who arrived beforehand. And knowing that, even if we don’t, some bastard is sure to try this, most of us won’t bother with option 2, the single first-come first-served queue. For whatever reason, when faced with parallel servers (a row of ATMs, a counter full of cashiers at McDonald’s) people tend spontaneously to form multiple queues, one for each service point.
Nor is this really the customers’ fault. Multi-server, single-queue systems are becoming more familiar: they’re used for airport check-in, at most government offices (for e.g license renewal and registration), inside banks, and for the new automated checkouts at supermarkets. But each of these examples shows the importance of queue-area design (cordons, signs, buffer space) for engineering cooperation amongst those waiting in line. Customers are instructed to queue in a particular way, not left individually to choose between two options, then forced to argue it out when they disagree. It is impossible to be served in any order besides that in which you arrived, whatever the relative service speeds of the staff. (My ATM example is a little misleading, because the queue space usually backs onto a narrow footpath, making multiple parallel queues more sensible than a single long line. Where ATMs are located in an internal vestibule, however, single queues are more common.)
But look at this recent example from a Melbourne supermarket, which gives customers a huge, empty queueing space to play in, and says: sort yourselves out. Madness!