Queueing Theory and the Waiting-Cost Trade-off

Definition

A waiting line (queue) forms whenever arrivals to a service system are random and can outpace the system's capacity to serve them at that moment, even if average capacity is sufficient. Queueing models describe the arrival process and the service process as probability distributions — commonly the Poisson distribution for the number of arrivals in a period — and from those distributions derive "operating characteristics" like the average number waiting and the average time in line. Because adding service capacity reduces waiting time but costs more, the design question is an explicit trade-off: the total cost is "the sum of the waiting cost and the service cost," and the goal is to choose the number of channels that minimizes that sum, not to eliminate waiting altogether.

In the Book

Chapter 11 develops the model through the Burger Dome fast-food restaurant, a "single-channel waiting line" where every customer passes through one order-taking station. The book models customer arrivals with the Poisson probability function, working out that at an arrival rate of 45 customers per hour there is a 0.4724 probability of zero arrivals, 0.3543 of one arrival, and 0.1329 of two arrivals in any given minute — and uses this distribution, paired with a service-time distribution, to derive Burger Dome's average wait and queue length. The chapter then performs an economic analysis: it assigns a waiting cost per customer per hour and a service cost per channel per hour (estimated at $7/hour at Burger Dome), sums them across different numbers of service channels, and shows that the total-cost curve is minimized at a specific number of channels — more channels than that overspends on idle service capacity, fewer starves customers with excessive waits.

Why It Matters

Queueing theory replaces "add more capacity when things feel slow" with a quantified trade-off between two costs that pull in opposite directions — under-provisioning and letting the people or things waiting bear the cost, or over-provisioning and paying for idle capacity. That same balance recurs anywhere requests arrive unpredictably against a shared, limited server: support agents, hospital beds, checkout lanes, or compute capacity, wherever the honest question is not "how do we eliminate the wait" but "how much capacity is the wait actually costing us to avoid."