Picture yourself as a casual observer, watching an organization serve its customers. You might be in a contact center watching a queue status monitor. Or observing customers enter a busy store. Or watching people line up for ice cream on a beach boardwalk.
If you want to manage queues well, it's important to begin with an understanding of how customers will arrive. In customer service, broadly speaking, there are two main arrival patterns: random and peaked. Understanding the difference between them is essential because it determines how you estimate and manage the resources you’ll need (staff, workstations, and related).
The Two Common Arrival Patterns
Random arrival is by far the most common pattern. Customers arrive as they choose—randomly. In comes a customer. Then two, three, four more. Exactly when they arrive, moment by moment, is the result of countless individual decisions driven by needs and circumstances.
If you look at large enough blocks of time—say, 30 minutes or more—you can almost always spot patterns. The airport security line slows down in the early afternoon then builds again later in the day. Technical support centers tend to be busiest on Monday mornings. Ice cream stands see the biggest rush on summer weekend afternoons.
You can use those trends to forecast demand. You might know that you'll receive around 50 customer contacts between 10 a.m. and 10:30 ma.m. on Monday. What you can't predict is how many of those customers will arrive at 10:01, 10:02, or 10:03. That minute-by-minute variation is what we mean by random arrival.
The other very different pattern is peaked arrival. This occurs when customers truly arrive in a very concentrated period of time. A flight cancellation sends passengers rushing to phones or counters for help. A major network outage triggers a wave of service calls. The ballgame ends and a line instantly forms for parking exits. These are all examples of peaked traffic.
We often use the term peak loosely. For example, "March is our peak season," or "Mid-morning is our peak hour." But in queue management terms, peaked traffic refers specifically to a spike beyond normal random variation, a true surge within a short amount of time.
Here's the key distinction to remember: Random and peaked arrival must be staffed very differently. While random arrival is by far the most typical in contact centers, many organizations encounter both. Knowing which you are facing and when is critical to planning and making staffing decisions.
Staffing for Random Arrival
When customers arrive randomly and are served by a pooled group of agents, workforce planners typically rely on the widely used formula Erlang C, developed by Danish engineer A.K. Erlang in the early 1900s. Fortunately, no one needs to work the math manually. The formula is built into virtually every workforce management system and into countless online calculators. Many AI tools can also run Erlang C scenarios.
Erlang C will need to know how many interactions, the average handling time (average talk time plus average after-contact work time) and your service level objective. To illustrate, let's assume 100 customers, an average handling time of five minutes, and a service level objective of 80 percent answer within 20 seconds. Here's the output:
Erlang-C calculators can quickly show how service improves as staffing increases. Here, 17 agents equate to only 11 percent of calls answered within 20 seconds. Adding staff raises performance steadily—39 percent, 59 percent, 73 percent—until you reach your target. In this example, achieving at least 80 percent of contacts answered within 20 seconds will require 21 agents during that half hour. Other metrics follow from these choices, such as average speed of answer and agent occupancy (how much of the time agents are handling interactions versus waiting for them to arrive).
Today's AI-driven tools (stand-alone or built into workforce management systems) now make it much easier for managers to model staffing quickly. Instead of guessing or working through spreadsheets, leaders can run real-time what-if scenarios: What happens if volume runs 10 percent higher than forecast? If handle time drops by 30 seconds? If service level goals tighten? These capabilities can instantly calculate Erlang C outcomes or run computer simulations that show queue behavior under different conditions, giving planners an immediate view of staffing requirements and performance tradeoffs.
Staffing for Peaked Arrival
With peaked traffic, customers arrive in a very concentrated period of time. The staffing question becomes direct: What is the longest wait you are willing to allow customers to experience? Your answer determines staffing.
Here's a simple example to visualize this type of scenario. Let's say it takes three minutes to provide tickets for an event, and you have four ticket booths open. A shuttle bus just dropped off the first 25 customers. The first four go right to the counters, no wait. The next four wait an average of three minutes. The next four wait an average of six minutes. And so on. With four booths open, the longest wait could be 18 minutes. You can whittle that down to any time you'd like by opening more booths. Five booths would equate to a longest wait of 12 minutes.
You'll determine required staff around the wait is appropriate and tolerable. I once worked with a company that handled calls for a charity. The charity ran ads during NFL football playoffs. The peaked traffic that arrived was significant, but they were ready. and the campaign was a success.
Here's the thing to remember: The distinction between random and peaked traffic is important because that will determine both how you estimate and manage the staff and other resources you'll need. You might handle both random and peaked arrival. Knowing the difference will help you prepare for whatever comes your way.
Brad Cleveland is a customer service consultant and senior advisor to the International Customer Management Institute (ICMI).