Why First Come, First Served Isn’t Always the Best Approach to Customer Service

As in most places, queues at call centers are typically handled on a first-come-first-served basis. But is this the most efficient process? A new study suggests there might be a better way.


“First-come-first-served is not always a great approach,” says Achal Bassamboo, professor of operations at the Kellogg School. “Companies need to have a more dynamic policy that’s based on how much time a customer has already spent waiting in line.”


A better policy, he thought, would center on the idea that not all customers are equally patient—some are more willing to wait in line than others, particularly if they know that the service they require is extensive or time-consuming. And people’s patience also changes over time as they are kept on hold.

Using data on customer wait times and the time required for the services they need, Bassamboo and a colleague devised a model that can help call centers schedule responses dynamically.

The key is to divide customers into groups based on how much patience they are likely to have, and then prioritize each group differently. This method turns out to be more efficient at reducing queue lengths and keeping customers satisfied than a first-come-first-served policy.

Patience Has Its Limits

A first-come-first-served policy is the most obvious way to handle a queue.

“The moment you say you’re not doing that, the first question is, ‘Why not?’” Bassamboo says.

Previous research on optimizing queue service typically assumed that a customers’ patience is “memory-less,” meaning that the amount of time they have already waited does not impact their decisions. For example, an extra 5 minutes of waiting is treated the same regardless of whether you have already waited 10 minutes or 30 minutes.

“But that is not reflective of real life,” Bassamboo says. “It is not a great assumption to work with when thinking of situations like call centers.”

In other words, we all remember how long we have been waiting and how annoyed we are getting.

But we are not all uniformly annoyed by the wait. If we know we have a complex question that will likely take a lot of time to address, we are likely to be more patient in that line.

For example, a person may expect a short task like withdrawing money at the bank to require little time and get easily annoyed with a long wait. But that same person may be more willing to wait a while in the same line to apply for a mortgage.

Defining Call-Center Success

Most models do not account for this connection between service time and patience. So Bassamboo and his colleague Ramandeep Singh of the University of Southern California constructed a model that links the amount of time a customer expects they will need with a representative and the amount of patience they will have while waiting on hold.

They assumed that customers entered the queue at a fixed rate, and also that requests were processed at a fixed rate. Then, the team studied how handling customer requests in different orders should affect three common measures of call-center success: queue length, abandonment (the rate at which customers got frustrated and hung up), and how accurate the call center could be with predicted wait times.

Each of these metrics requires a different strategy, Bassamboo explains.

“If you want to minimize abandonment—that is, you don’t want anyone to leave—then you would try to first serve the most impatient customers,” he says. “But if your goal is to keep queues short, then you are better off doing the opposite, because the impatient people will leave anyway, and shorten the queue.”

In reality, however, most companies are functioning between these two extremes. “You do not want people to leave, but you also don’t want to make them wait,” he says.

Classifying Customers

Using their model, the researchers found that the best way to handle calls was by creating thresholds of customer patience based on the complexity of the customer’s issue. This divided customers into multiple groups, which the call center could then handle differently.

A call center would first serve everyone in the most patient group since their issues take longest to resolve and thus the line (and wait times) would get much shorter once they are helped, and then everyone in the most impatient group, which has quick-to-resolve needs. Then representatives would serve people in the third group (those with a middle level of patience) in a last-come-first-served manner.

It sounds counterintuitive, but the model found that within this group, it made the most sense to first serve customers who have waited the least amount of time, because those on the line the longest have already shown themselves to be very patient.

In general, whether the goal was shorter queues or less abandonment, the researchers found that a system only needed up to two thresholds, which divided people into at most three categories.

“If you choose the thresholds appropriately, you’ll get much better performance than doing it in the usual first-come-first-served way,” Bassamboo says.

These thresholds proved to be effective at improving queue lengths, abandonment rates, and predicted wait times.

“Across all the metrics of success, the principle was the same,” he says. “The thresholds may change based on who you want to prioritize, but it remains simple—there are just two thresholds, and everything revolves around that.”

To implement such scheduling, call centers would need to analyze their own customer data on typical wait times for different tasks using the researchers’ model, Bassamboo says. Although a first-come-first-served policy may continue to be the best for certain call centers, others may find that categorizing customers by patience improves efficiency.

“Based on our results, this system pays off a lot in certain settings in terms of providing better service or becoming more lean,” he adds. “Implementing such policies can help companies’ bottom line, because you do not need to hire more people to get the same performance.”