A line of five customers at a checkout counter can mean different things. Several shoppers may have arrived at once, one transaction may have taken longer than expected, or the cashier may be operating too close to available capacity. Hiring another employee without understanding the cause may not solve the problem.
Queueing theory helps small businesses examine the relationship between arrivals, service, waiting, and work piling up. In this practical guide, you will learn how to measure a waiting line in ten-minute intervals, interpret a real-world-style example, and decide what to investigate before changing staffing or processes.
We also explain how to use a free queue and wait-time Excel template to record observations and review the results. For related resources, visit our Processes & Operations hub.
On This Page
- What Is Queueing Theory?
- What Should You Measure?
- Waiting Line Example: One Checkout Counter
- How to Calculate Average Wait Time
- Arrival-to-Capacity Ratio vs. Utilization
- Why Averages Can Hide Waiting Problems
- Little’s Law: Connecting Flow and Time
- When Do You Need a Formal Queueing Model?
- How to Improve Waiting Lines
- Common Measurement Mistakes
- Download the Free Excel Template
- Frequently Asked Questions
What Is Queueing Theory, and Why Does It Matter to Small Businesses?
Queueing theory studies systems in which customers, orders, vehicles, or tasks arrive and may wait before receiving service. It is used in operations research to examine arrivals, available service capacity, congestion, and delays. A checkout line is an obvious example, but orders waiting to be packed or support tickets awaiting an agent also form queues.
The basic flow is Arrival → Waiting Queue → Service → Exit. Not everyone who arrives necessarily has to wait: when the server is available, service can begin immediately.
Formal queueing models can estimate waiting times under stated assumptions about arrivals and service. But a small business can begin with a more modest and often more useful task: observe what actually arrives, how many customers begin service, and how the waiting line changes. The example below is an observational analysis, not a probabilistic prediction.
Six Queueing Metrics Worth Measuring
| Metric | Meaning | Example |
|---|---|---|
| Arrival rate (λ) | Customers or tasks entering per unit of time | 27 customers per hour |
| Service time | Time from the start to the end of one service | 2 minutes per checkout |
| Service capacity | Approximate throughput a continuously available server could provide | 30 checkouts per hour |
| Wait time | Time from arrival until service begins | 4 minutes for one customer |
| Queue length | Customers waiting, excluding those already in service | 5 waiting at an observed cutoff |
| Time in system | Wait time plus time receiving service | 4 + 2 = 6 minutes |
The distinctions matter. Wait time is not service time, and queue length is not the same as the number of people in the entire system. If one shopper is paying and three others are in line, the queue length is three, while four shoppers are in the system.
Arrival rate = Arrivals ÷ Observation time. For example, 18 arrivals in 30 minutes correspond to 36 arrivals per hour. If a checkout takes about two minutes per customer, its simplified theoretical capacity is 60 ÷ 2 = 30 customers per hour when it remains available and works continuously. This capacity estimate is not proof that all 30 customers were actually served.
How to Measure a Waiting Line: A Small Retail Store Example
Consider Oak Street Market, a fictional U.S. convenience store with one checkout counter. The manager observes the line from 5:00 to 6:00 p.m., dividing the hour into six ten-minute intervals. The illustrative service-time benchmark is two minutes per customer, or approximately five service opportunities per ten-minute interval.
The observer records arrivals, customers who begin service, and the number still waiting at the end of each interval. These counts describe the waiting line; they do not by themselves prove how long any individual customer waited.
| Interval | Arrivals | Service starts | Waiting at end | Observed avg. wait (min)* |
|---|---|---|---|---|
| 5:00–5:10 | 3 | 3 | 0 | 0.5 |
| 5:10–5:20 | 4 | 4 | 0 | 0.8 |
| 5:20–5:30 | 8 | 5 | 3 | 2.2 |
| 5:30–5:40 | 7 | 5 | 5 | 4.5 |
| 5:40–5:50 | 3 | 5 | 3 | 5.2 |
| 5:50–6:00 | 2 | 5 | 0 | 3.3 |
| Total | 27 | 27 | 0 at final cutoff | See weighted calculation below |
*All figures are fictional. The average wait in each row represents the customers who started service in that interval; it must be measured separately, for example by recording arrival and service-start timestamps. It cannot be calculated from the arrival and service-start counts alone.
For each interval, use this waiting-line balance:
Ending queue = Starting queue + Arrivals − Service starts.
Between 5:20 and 5:30, the starting queue is zero, eight people arrive, and five begin service: 0 + 8 − 5 = 3 waiting. In the next interval, five people start from the earlier queue and new arrivals: 3 + 7 − 5 = 5 waiting. Later, the line shrinks as service starts exceed new arrivals.
By 6:00 p.m., nobody remains waiting at the final cutoff, although a customer may still be receiving service. This does not mean there was no congestion: the count reached five waiting customers at one ten-minute cutoff. It is the maximum among those six recorded endpoints—not necessarily the maximum at any instant during the hour.
When using the companion Excel workbook, the starting queue for each interval must match the previous interval’s ending queue. Dates and times should also follow an uninterrupted sequence unless a new observation session is explicitly started. The English version will include these consistency checks.
How to Calculate Average Customer Wait Time
For a customer who arrives at 5:22 p.m. and begins checkout at 5:27 p.m., wait time = 5 minutes. If checkout itself takes two minutes, the total time from arrival to service completion is seven minutes.
For interval-based reporting, first calculate or observe the mean wait of customers whose service began during each interval. Then weight those interval averages by the number of service starts—not by arrivals:
Overall average wait (minutes) = Σ(Service starts × Average wait of those service starts) ÷ Σ(Service starts).
For Oak Street Market, the weighted total is (3 × 0.5) + (4 × 0.8) + (5 × 2.2) + (5 × 4.5) + (5 × 5.2) + (5 × 3.3) = 80.7 customer-minutes. Divide by the 27 customers who began service: 80.7 ÷ 27 = 2.99 minutes.
This average describes the 27 service starts observed in the hour, not necessarily every arrival cohort under a different measurement method. If your data refer instead to customers arriving within each interval, you must group and weight their waits by arrivals. Do not mix these populations. The adapted English workbook will follow the service-start definition used here.
Practical limit: a spreadsheet containing only arrivals and service counts cannot reconstruct actual customer wait times. Someone must observe or record the waits, such as with timestamps, and enter the interval averages. If wait observations are unavailable, report not measured rather than presenting a calculated estimate as an observed fact.
Arrival-to-Capacity Ratio: What Does 90% Mean?
The store recorded 27 arrivals per hour and an estimated checkout capacity of 30 customers per hour.
Arrival-to-capacity ratio = Arrival rate ÷ Theoretical service capacity × 100 = 27 ÷ 30 × 100 = 90%.
For the 5:20–5:30 interval, arrivals occur at an equivalent hourly rate of 8 ÷ 10 × 60 = 48 customers per hour. Relative to the estimated capacity of 30, that interval’s ratio is 160%. The next interval produces 140%. These figures indicate offered arrival load above the simplified service capacity for those intervals; they do not mean the cashier was physically occupied 160% or 140% of the time.
Do not label this observational calculation as measured server utilization. Actual utilization requires an appropriate measure of busy time divided by available time, or a justified formal model. Under restrictive steady-state assumptions, queueing models may relate utilization to arrival and service rates, but our one-hour observation does not establish those assumptions.
The useful management question is whether the store has enough capacity at the moment demand arrives. A manageable hourly average can hide short bursts that generate waiting. For a broader method of sizing available resources, see Small Business Capacity Planning.
Why Average Arrivals Can Hide Long Waiting Lines
Two stores can each receive 30 customers in one hour and have very different waiting experiences. One may receive about five customers every ten minutes; another may receive two, three, four, twelve, six, and three. Their hourly average is identical, but the second pattern concentrates demand in a short period.
Service also varies. A simple purchase may take under a minute, while a return, price lookup, or payment issue may take several minutes. Variation in both arrivals and service can create waits even when average capacity seems adequate. Queueing research therefore treats variability as an important driver of congestion, particularly when resources operate close to capacity.
Before changing permanent staffing, compare the same measurements across representative days and peak periods. A sixty-minute sample may reveal a problem worth investigating, but it does not establish what happens during every shift.
Little’s Law: Connecting Customers, Flow, and Time
Little’s Law relates the average number of customers in a stable system to the average arrival or throughput rate and the average time each customer spends in that same system:
L = λ × W, where L is the long-run average number in the system, λ is the corresponding average flow rate, and W is the average time in the system.
For example, suppose a stable service process handles an average of 12 customers per hour, and each customer spends 10 minutes from arrival to completion. Ten minutes is one-sixth of an hour, so L = 12 × (10 ÷ 60) = 2 customers in the system on average.
The system includes both people waiting and anyone being served. To analyze the waiting line alone, use Lq = λ × Wq, with queue-only length and queue-only waiting time. Keep the population and time units consistent.
Little’s Law is a long-run relationship under suitable flow and stability conditions. It should not be used to infer precise individual waits from six queue-endpoint observations. The Excel template measures a specific observation session; it is not a Little’s Law simulator.
When Do You Need a Formal Queueing Model?
Start with observational measurement when your question is when did the line grow, how many were waiting, and what was the observed delay? Move to a formal model or simulation if you need defensible predictions about staffing levels, expected waiting time under new demand, multiple checkout counters, appointment rules, or service guarantees.
A familiar academic model, M/M/1, assumes one server, a Poisson arrival process, exponentially distributed service times, and other specified conditions. Its equations should not be applied automatically to a real store merely because you know an hourly arrival average and a nominal two-minute service time. Real businesses may have several servers, interruptions, customers who leave, or demand that changes throughout the day.
For the first analysis at Oak Street Market, the interval table and measured waits support practical investigation. They do not justify predicting a precise average wait under a different staffing arrangement.
How to Improve Waiting Lines Without Guessing
| What you observe | What to investigate | Potential next action |
|---|---|---|
| Queue grows mainly during short peaks | Timing of arrivals and available staff | Move breaks, adjust coverage, or trial a peak-hour backup cashier |
| Service takes longer than expected | Payment steps, returns, price checks, and task design | Remove unnecessary steps and standardize routine transactions |
| Arrivals routinely exceed capacity | Resource capacity and longer-term demand | Recalculate capacity and consider additional service resources |
| Backlog repeats at one handoff | Transfers, approvals, or dependencies | Map the underlying process and identify the constraint |
| Workload varies widely between periods | Appointment patterns or demand scheduling | Test appointment slots, pickup windows, or demand smoothing |
| Recorded waits remain long even with manageable load | Service-time variation, interruptions, and measurement quality | Observe transactions in more detail before making staffing changes |
For Oak Street Market, the first action might be a limited peak-hour trial rather than a permanent second cashier. Measure comparable days, adjust coverage during the 5:20–5:40 p.m. peak, and compare actual waits, queue endpoints, labor hours, and customer service outcomes.
If the problem lies in the sequence of checkout tasks or handoffs, use How to Create a Process Map. If you need complementary measures for service demand, pending cases, and response speed, see Customer Service KPIs. Those guides answer related but distinct questions.
Common Queue Measurement Mistakes
- Reporting service starts as completed services: a customer who begins checkout may still be in service at the cutoff. Define exactly what you count.
- Confusing the queue with the whole system: waiting customers and customers in service are different groups.
- Calling a high arrival-to-capacity ratio actual utilization: offered workload is not a direct record of server busy time.
- Calculating wait time from counts alone: arrival and service-start totals do not reveal how many minutes each customer waited.
- Weighting waits by the wrong population: average waits observed for customers beginning service must be weighted by service starts.
- Ignoring continuity: the next interval’s starting queue must agree with the previous interval’s ending queue for one uninterrupted observation.
- Calling the largest recorded endpoint the absolute maximum: the queue may have been longer between observations.
- Making permanent staffing decisions from one hour: repeat measurements during representative demand patterns.
Download the Free Queue and Wait-Time Excel Template
The free Queue and Wait-Time Excel template helps you observe a service point in consistent intervals. The planned English version retains three worksheets: Instructions, Glossary, and Queue Measurements. It is designed for a basic-to-intermediate user measuring one checkout counter, service desk, or comparable resource at a time.
Record arrivals, service starts, starting and ending queues, observed average wait by service-start interval, and average service time. The adapted workbook will use the same Oak Street Market example and distinguish arrival-to-capacity ratio from measured utilization. It will also check interval continuity, identify invalid entries, and calculate the weighted average wait using service starts.
Limitations: this spreadsheet records observational measurements; it does not automatically reconstruct individual waiting times from arrival counts, measure true busy-time utilization, or forecast M/M/1 performance. Reliable results depend on consistent observation and accurate input data.
Frequently Asked Questions About Queueing Theory
What is the difference between queueing and queuing theory?
Both spellings refer to the mathematical study of waiting lines. Queueing theory is common in operations research, while queuing theory also appears in textbooks and business resources. They describe the same field.
Can a small business analyze a waiting line without advanced mathematics?
Yes. Start by counting arrivals and service starts, recording queue lengths at consistent times, and measuring actual waits. These observations can reveal peak periods and mismatches between demand and available service even before you use a statistical model.
How long should I observe a waiting line?
A 30- to 60-minute observation can be a useful first diagnosis when it includes a relevant operating period. Repeat it across different days or time slots when demand varies before making permanent changes to staffing or capacity.
Does a 90% arrival-to-capacity ratio mean the cashier was busy 90% of the time?
No. It means observed arrivals were 90% of a simplified theoretical capacity estimate for the hour. Measuring the cashier’s actual busy-time utilization requires different data.
Does a waiting line mean I need another employee?
Not necessarily. A line can result from temporary arrival peaks, variable service duration, interruptions, process handoffs, or insufficient capacity. Determine which explanation is supported by your measurements, then test the least disruptive reasonable improvement.
Sources and Further Reading
- INFORMS — Stochastic Modeling and Queueing for Operations Research Analysts. Queueing models, variability, and the conditions behind predictions.
- IEEE Technology Navigator — Queueing Analysis. Model assumptions and performance measures.
- MIT OpenCourseWare — Queueing Systems Lecture Notes. Flow, Little’s Law, and queue modeling.
- MIT Sloan — Queues: Theory and Applications. Academic readings on single- and multiserver systems.
- Victoria University of Wellington — Queueing Systems M/M/1 and M/M/c. Queueing formulas and their assumptions.
- TU Delft OpenCourseWare — Little’s Law. Relationship between throughput, time, and work in process.
- IBM — Performance Cookbook: Queueing Theory. Operational meaning of arrival rates, service time, queue depth, and delay.
Discover more from
Subscribe to get the latest posts sent to your email.
