Accounting for human nature, parallel queues beat single on service speed
Death, taxes… and waiting in lines.
Yes, it might be time for the everyday act of queuing to join the short list of Most Inevitable Human Experiences. In our hurry-up-and-wait world, who among us has not engaged in the folly of calculating the fastest line at the grocery store, bank, café, post office, DMV or TSA security screening?
Fact is, there is a science to queue design. And its consensus has long favored a single-line model serviced by multiple cashiers—the system of choice in many retail environments.
But new research from the University of Washington Foster School of Business indicates that a system of “parallel” queues, each serviced by a dedicated cashier, can provide a more expedient delivery of customer service.
The reason for this reversal of the conventional wisdom? Human nature. People, the study shows, work harder when they feel accountable.
“Past research in operations management has compared single and parallel queues using mathematical models that assume service speed is static,” says co-author Masha Shunko, an assistant professor of operations management at Foster. “But when we add a behavioral component to the equation, we find that the servers adjust their performance based on the environment they’re working in.”
Humanity at the register
This insight came from a pilot field study that Dr. Shunko and collaborators Julie Niederhoff and Yaroslav Rosokha performed at Purdue University. To begin observing the behavioral implications of queue design and visibility, the researchers manipulated the ticket lines at Boilermaker volleyball matches. By shifting between single and parallel queues, they were able to measure the relative pace of ticket transactions. Cashiers in the parallel queues processed transactions faster. And cashiers who had better visibility of either kind of queue worked faster.
To test the hypothesis in a more controlled environment, they designed a lab study using a virtual grocery store checkout line. Study participants acted as cashiers tasked with ringing up the grocery purchases of customer avatars.
By manipulating queue design, visibility and incentives, Dr. Shunko and her colleagues were able to test the work rate of study participants—and the resulting queue speeds—under different conditions and combinations of conditions. There were parallel versus single lines. Visible versus obscured lines. And per-unit-processed incentive pay versus flat-rate compensation.
What’s my motivation?
The variations revealed a number of insights about queue design and its effect on the productivity of service workers.
Across nearly every variation and combination of variations, parallel queues moved faster than single queues. Single queues slowed down service workers.
“Servers work faster in parallel systems because they have a sense of accountability for their own queue instead of relying on others to work through a lengthy single line,” says Dr. Shunko. “They also get more immediate feedback from seeing their efforts reduce the length of their dedicated line.”
The researchers also discovered that a partially obscured queue (by a wall or shelf, for example) tends to slow down servers in most scenarios. The one exception they found was when servers working a single line are incentivized for their productivity rather than paid a flat wage. In this case, the slow-down effect of an obscured queue is mitigated or even reversed.
“Visibility is always good for efficiency in parallel queues. And it’s important in both queueing system when you don’t have incentives for performance,” Dr. Shunko says. “When you do provide compensation incentives for faster work, however, obscuring some of a single line can actually enhance competition between servers.”
Not necessarily faster
In the brutally competitive $2.6 trillion American retail industry, every seller is searching for ways to optimize the customer experience. And until now, much of the existing theory has been missing an important behavioral element.
Of course, Dr. Shunko admits that her study only begins to explore the human side of queueing.
“We find that servers work faster in parallel lines. And as a result this may make lines move faster,” she says. “But parallel lines are not always the most efficient way to process customers. Other factors can contribute to the speed at which lines move. And the single line system offers benefits in some contexts.”
For instance, one slow customer—who has a lot of items, a product return or some other complication—is less likely to derail a single-queue system.
In most contexts, however, Dr. Shunko says that single-queue systems have the potential to slow down the performance of servers. “The magnitude depends on the context,” she says. “And there are strategies to compensate for this slowdown using incentives and increasing or, in some cases, decreasing visibility of the line.”
“Humans Are Not Machines: The Behavioral Impact of Queueing Design on Service Time” is forthcoming in the journal Management Science.