Whether it’s ordering food or making a doctor’s appointment, “as customers wait in line, there comes a point where they eventually get so bored or frustrated that they leave,” he says Ahal BasabuKellogg business professor.
This problem of customers leaving without ever being served is one the service industry has studied for decades, he says, and it happens in two main ways: “referral,” which is when someone sees a long line and decides not to join it at all, and “abandonment” (often called “abandonment”), which is when someone joins the line but leaves before service begins to stop waiting.
Solving the problem isn’t just about adding staff or speeding things up. Customer behavior and expectations also shape whether people join a line in the first place — and whether they stay.
Imagine, for example, a modern ice cream parlor with lines out the door. Each customer who enters the queue has a certain amount of time to wait before giving up. Those who succeed, however, may feel entitled to celebrate their successful endurance.
“When it’s your turn to get ice cream after a long time, you feel like you’ve earned the right to try more flavors,” says Bassamboo. “As I make you wait, you change your behavior.”
A series of studies from Bassamboo with Ohad Perry of Southern Methodist University, Chengguang Wu of the Hong Kong University of Science and Technology and Kellogg PhD student Deniz Şimşek shed light on this phenomenon and how businesses can use it to optimize wait times.
One of their key findings is that the amount of time customers have to wait for a service – their “patience clock” – is linked to how much time they end up spending on that service – their “service clock”.
“What we were looking for in our research is, ‘Why might this be happening?’ he says. “And the simplest explanation is that these two clocks—your patience clock and your service clock—are not actually independent.”
To wait or not to wait
Customer service systems are extremely complex, so researchers like Bassamboo often turn real-world scenarios into simplified models with key assumptions so they can analyze them and extract insights.
One such common assumption is that the maximum time a customer is willing to wait before they give up and leave is independent of how long they spend with the business when their turn comes.
But when Bassamboo and his colleagues took a closer look at modeling service systems, they suspected that this long-held assumption failed to capture how customers react in the real world.
Bassamboo and colleagues categorized the relationship between a customer’s maximum wait time and the time they spend on a service in two settings.
First, there is “extrinsic” dependency, where customers enter a service location with a predetermined maximum waiting time based on their individual needs.
A simple example might include someone who is on a tight schedule and can spend up to 10 minutes on a service call with their Internet provider. Whether they have to wait 1 minute or 3 minutes before being transferred to customer service won’t directly change the time they spend talking to an agent, as they have a specific question they want answered as quickly as possible. However, although the wait time may not change the service time, they are linked in the sense that the customer will simply stop the service call if the wait time exceeds 10 minutes.
With “endogenous” dependence, the waiting time does Change the amount of time a person is willing to spend receiving a service, as in the ice cream parlor example. The more time people spend waiting in line, the more time they can spend sampling and choosing flavors.
When the researchers entered data from various service systems into “exogenous” and “endogenous” dependency models, they turned out to be much more closely related than the researchers had imagined.
“What we noticed is that for every exogenous model, there is an equivalent endogenous model,” says Bassamboo. “So the endogenous model is actually a much larger family of models, and the exogenous is a smaller subset.”
Discovering this link between the two types of dependence helped them confirm that customers’ wait times and service times are indeed linked.
Congestion collapse
The researchers then developed a unified model that captured these two types of dependencies simultaneously. This unified model was a “vehicle for analysis” that allowed the team to extract insights from service systems by entering data such as the rate at which customers enter a system and the time it takes to receive service.
The unified model helped Bassamboo and his colleagues identify critical differences between exogenous and endogenous dependencies and how they affect what businesses care about: customer wait times.
In the case of exogenous dependence—in settings where customers with a predetermined time limit seek a service—there is a unique point of stability when there are no volatile changes in customer waiting and abandonment times. For businesses, these arrangements would allow for predictable wait times and, in turn, consistent staffing.
But in the case of endogenous dependence—in environments where waiting motivates customers to change their service time—waiting times can vary dramatically and frequently. This scenario is particularly relevant to businesses where the customer has discretion over how long a service takes.
“You can go from very small queues and a very well-behaved system to suddenly very long queues,” explains Bassamboo. When the wait at, say, an ice cream shop fills up in this sudden, erratic way, it’s likely due to a “congestion collapse, where, even though my system has enough staff and capacity for the number of customers that should have been in line, because I got unlucky with two or three picky customers who decided to try 10 flavors.”
One of the practical takeaways from this finding is that, in this latter scenario, a business should be able to address long wait times with a relatively simple, temporary fix rather than a major change to the waiting system.
“You can actually get maybe someone from the back of the store to come and help get the line down really quickly so the wait is zero,” says Bassamboo. “And when the wait comes back, everyone goes back to trying one or two flavors instead of five.”
