Single Queue or Multiple Lines: Which Is Faster?
Picture a counter with two service points. Before opening, someone has to settle the layout: one queue feeding both tills, or a separate queue at each. It gets decided once and then lived with for years. It’s easy to assume one of the two is simply faster.
It isn’t. With the same two staff working at the same speed, both arrangements serve the same number of customers per hour. Throughput comes from how many people are serving and how quickly they work. Single queue vs multiple lines is a layout question, not a capacity one.
What the layout does change is everything else. Who waits longest, how uneven the waits are, and whether people are served in the order they arrived. Research published in 2024 adds one more: how fast the staff themselves work.
That counter is illustrative.
Key Takeaways
- With the same tills and staff, one shared queue and several separate queues serve the same customers per hour. The shape of the queue is a layout decision.
- What pooling changes is the waiting. A 2024 paper states the classic result plainly: pooling reduces the variability of workload between servers and improves average waiting times (Heliyon, 2024).
- One queue enforces arrival order, and several queues make being overtaken routine. Larson argued in 1987 that customers may become infuriated by exactly that (Operations Research, 1987).
- Several queues aren’t simply worse. In two online experiments in healthcare, a 2024 study found dedicated queues faster on processing speed (M&SOM, 2024). The counter-case dates to 1987.
- There’s no verdict. Pooling suits the customer’s wait, dedicated queues suit the server’s speed, so it depends on which one binds in your venue.
Queue shape is not capacity
Throughput is the number of customers a counter gets through in an hour. Two things set it: how many people are serving, and how long each transaction takes. Rearranging the people who are waiting changes neither. That’s the whole reason queue shape can’t add capacity.
A shared queue is often described as making a counter faster. Combining queues adds no server and shortens no transaction, so the rate at which customers leave the counter is unchanged. If people arrive faster than the counter can serve them, the queue grows either way, in one place or in several.
The classic comparison sets several separate single-server queues against one queue served by several servers, with the same staff on each side. A 2024 paper restates that comparison and calls its outcome well known (Heliyon, 2024). What moves between the two is how the work is spread across the servers.
Capacity is a separate lever, and it’s the one customers notice. In its 2023 supermarkets survey, Which? reported that 22% of shoppers said there weren’t enough staffed checkouts (Which?, 2023). That’s a complaint about how many tills were open, not about how the queues in front of them were arranged. No layout answers it.
So what does the shape of the queue change?
Four things, and throughput isn’t one of them. They move independently, and two of them move in opposite directions.
| What “faster” could mean | What one shared queue does to it | The evidence, and how far it goes |
|---|---|---|
| Customers served per hour | Nothing, with the same servers working at the same rate | Mechanism, not a measurement |
| The average customer’s wait | Improves it, by evening out the workload between servers | The classic queueing result, restated in a 2024 paper that calls it well known. Theory, not a field measurement |
| How much waits differ between customers | Reduces the spread. One slow transaction is absorbed by everyone rather than by one queue | The same classic result, restated in Heliyon, 2024. A standard model, not a shop |
| Being served in the order you arrived | Enforces it. One queue leaves nowhere to overtake | Definitional. Larson’s 1987 argument is about how customers react when that order is broken |
| How fast the servers themselves work | May reduce it. Dedicated queues beat pooled ones on processing speed | M&SOM, 2024. Two online experiments in healthcare delivery, set alongside earlier field studies in consumer services pointing the same way |
That third row is the one a small venue feels. With separate queues, one till can be stuck on a complicated order while the next stands idle, and the person behind that order absorbs the whole delay. A shared queue spreads it across everyone instead.
Over a full day, with the same customers arriving, the same work still gets done. What a shared queue prevents is a till sitting idle while someone waits. That isn’t extra capacity. It’s the capacity already on the floor not going unused.
The case for a single queue is about evenness, not speed
The orthodox result is about variability, and a 2024 paper puts it in one sentence. Pooling, it says, “reduces the variability of the workload between the servers, and thus improves all service performance measures, such as the average waiting times” (Heliyon, 2024). The paper calls this well known, which is the right status for it: a textbook result restated in 2024.
That’s a model rather than a measurement, and customers served per hour never enter it.
The second argument for one queue has nothing to do with time at all. Larson argued in 1987 that “customers may become infuriated if they experience social injustice, defined as violation of first in, first out” (Operations Research, 1987). One queue makes that violation almost impossible. Several queues make it routine, because the person who arrived after you can reach a till before you.
Larson called his own work a “speculative paper” in its abstract, so it’s fairer to him to read it as an argument than as a finding. Why that particular unfairness lands so hard is the subject of why queue jumping makes people angry.
There’s recent evidence that the feeling is a separate variable from the clock. A 2025 experimental paper found customers more satisfied with one queue-skipping arrangement than another “even when the wait time is held constant” (M&SOM, 2025). Three experiments with subject-pool participants, measuring satisfaction rather than speed. Felt time behaves like that in general, which is the subject of why waiting feels longer than it is.
Four reasons combining queues can backfire
There’s a formal counter-case, and it’s nearly forty years old. In December 1987, Michael Rothkopf and Paul Rech argued in Operations Research that “combining queues, especially queues of people, may at times be counterproductive” (Operations Research, 1987). They named four reasons.
Customer reaction. A shared queue is longer than any of the queues it replaces, and their first objection was simply that: customer reaction. A queue that looks long from the door is a different proposition from four short ones, whatever the maths says. Whether a longer queue turns people away is taken up in why customers leave a queue.
Jockeying. With separate queues, people move between them, which flattens some of the imbalance a shared queue is meant to remove. Notice the direction: jockeying is listed as a reason pooling may disappoint, not as a flaw in separate queues.
Cost and service time. A shared queue needs floor space, often barriers, and some way of calling the next person forward. There’s a speed cost too. A 2024 study reports that “following a change in queue configuration”, dedicated queues beat pooled ones “with respect to processing speed without sacrificing quality” (M&SOM, 2024).
The 2024 paper’s mechanism has two halves that pull apart. On one side, “queue length awareness motivates servers to work faster”. On the other, “customer ownership of those in queue may distract them and result in a slowdown” (M&SOM, 2024). Two online experiments in a healthcare delivery context, and the paper scopes its advice to settings where the server controls how long a job takes.
The absence of before-and-after studies. Their fourth complaint has been partly answered since. A 2015 study followed an emergency department that switched from a pooled system to dedicated ones, and compared patient records from before and after (Management Science, 2015). For a shop switching from several queues to one, no published before-and-after test was found.
Larson’s fairness paper and this counter-case ran on consecutive pages of the same December 1987 issue of Operations Research, pages 895 to 905 and 906 to 909. They shared the title stem “Perspectives on Queues”. The case for pooling and the case against it were published side by side.
The study everyone quotes did not replicate
Much of the popular coverage claiming that parallel queues beat a single queue traces back to one paper. It was published in Management Science in January 2018 by Masha Shunko, Julie Niederhoff and Yaroslav Rosokha (Management Science, 2018). Its title opens “Humans Are Not Machines”.
It was then retested formally. The paper was one of ten chosen for the Management Science Replication Project, whose results were published in Management Science in 2023 (Management Science, 2023). The project’s page for it records the outcome as “Non-Replication” (project page).
The scope matters, because the whole paper wasn’t overturned. Only Hypothesis 1 was tested: “service times are shorter when customers are aligned into multiple parallel queues instead of a single pooled queue”. The stated condition was that queues are visible and pay is flat. The measure is service time, the server’s speed, not the customer’s wait.
The 2023 project page records the rest. It ran on Amazon Mechanical Turk, because the original did, targeting 244 people per site. The sites were the University of Wisconsin-Madison and the University of South Carolina, preregistered as AsPredicted #71741 (project page). Neither the original nor the replication measured a real shop.
So the study behind that coverage is an online experiment that didn’t hold up when it was run again. Other evidence points the same way, though. A 2018 study used field data from a real supermarket and found servers in dedicated queues worked faster (Management Science, 2018). Its authors put that down mainly to social loafing, the tendency to ease off when the work is shared.
Which queue should you join?
This part is for the person standing there rather than the person designing the counter. In several parallel queues you’re making a choice, and it’s a guess. With three or more queues moving at different rates, most people are standing in one of the slower ones. That’s arithmetic.
That’s why the queue beside you so often seems to be moving better. Sometimes it is. In a single queue there’s no choice to get wrong, which is a genuine difference in the experience and not a difference in how fast the counter works.
If you are choosing, the useful thing to look at isn’t only how full the trolleys are. It’s how many separate transactions sit in front of you. Every customer carries a fixed cost: greeting, payment, bagging, the loyalty card question. A queue of two big shops can clear before a queue of five small ones.
If you only have two tills
Most venues arguing about this have two or three service points, not twelve. At that size the question is less about theory and more about the room you have. Both columns describe things you can check on your own floor.
| One shared queue suits you when | Separate queues suit you when |
|---|---|
| Any member of staff can serve any customer | Staff are doing different jobs, such as collections or returns |
| Customers don’t mind who serves them | Customers are pre-committed to a service type or a named person |
| There’s floor space for one line | The floor plan can’t hold one line without blocking a door |
| Arrival order matters to your customers | Staff work better seeing their own queue |
When rows on both sides describe your counter, ask which constraint actually binds. If people give up before they’re served, the waiting is the problem, and how long customers wait before leaving takes that further. If they’re served slowly, it isn’t.
Frequently Asked Questions
Does a single queue serve more customers per hour than several queues?
No. With the same staff serving at the same rate, both arrangements clear the same number. Even the classic result that favours pooling reports gains in average waiting times rather than in customers served (Heliyon, 2024). Throughput follows staffing, so a queue that keeps growing all day is telling you about capacity, not layout.
Why do banks use one queue and supermarkets use several?
Mostly because the work differs. Bank counters are largely interchangeable, so any cashier can take any customer, which is what a shared queue needs. Supermarket checkouts sit at the ends of aisles with trolleys behind them, so one line often has nowhere to stand. Rothkopf and Rech named the cost of combined queues among their objections in 1987 (Operations Research, 1987).
Do express lanes make a queue faster?
Not demonstrably. What has been tested is how being overtaken feels. A 2025 experiment compared two kinds of queue-skipper with the wait held constant. Customers were more satisfied meeting a line-sitter, someone paid to hold a place, than an express-line customer (M&SOM, 2025). The authors warn service providers about customer backlash against express lines. Where to draw the cut-off remains open.
Does it help to switch to a queue that’s moving faster?
Sometimes it pays, and it always costs you the place you already had. You’re swapping a position you know for a guess about the transactions ahead of you, which is what really sets the pace. Switching queues has a name, jockeying, and Rothkopf and Rech listed it in 1987 among the reasons combining queues may disappoint (Operations Research, 1987).
Sources
- Heliyon, Hindy Ling, Etgar and Bar-Gera, “Server pooling models for separate and bounded queues”, published 1 February 2024, retrieved 2026-08-28, https://pmc.ncbi.nlm.nih.gov/articles/PMC10907676/
- Management Science, Davis, Flicker, Hyndman, Katok, Keppler, Leider, Long and Tong, “A Replication Study of Operations Management Experiments in Management Science”, volume 69 issue 9, pages 4977 to 4991, published September 2023, retrieved 2026-08-28, https://doi.org/10.1287/mnsc.2023.4866
- Management Science, Shunko, Niederhoff and Rosokha, “Humans Are Not Machines: The Behavioral Impact of Queueing Design on Service Time”, volume 64 issue 1, pages 453 to 473, published January 2018, retrieved 2026-08-28, https://doi.org/10.1287/mnsc.2016.2610
- Management Science, Song, Tucker and Murrell, “The Diseconomies of Queue Pooling: An Empirical Investigation of Emergency Department Length of Stay”, volume 61 issue 12, pages 3032 to 3053, published December 2015, retrieved 2026-08-28, https://doi.org/10.1287/mnsc.2014.2118
- Management Science, Wang and Zhou, “Impact of Queue Configuration on Service Time: Evidence from a Supermarket”, volume 64 issue 7, pages 3055 to 3075, published July 2018, retrieved 2026-08-28, https://doi.org/10.1287/mnsc.2017.2781
- Management Science Replication Project, “Shunko et al. (2018)” project page, University of Texas at Dallas, retrieved 2026-08-28, https://msreplication.utdallas.edu/shunko-et-al-2018/
- Manufacturing & Service Operations Management, Althenayyan, Ülkü, Yang and Cui, “Not All Lines Are Skipped Equally: An Experimental Investigation of Line-Sitting and Express Lines”, volume 27 issue 1, pages 287 to 304, published January 2025, retrieved 2026-08-28, https://doi.org/10.1287/msom.2022.0338
- Manufacturing & Service Operations Management, Song, Armony and Roels, “Queue Configurations and Operational Performance: An Interplay Between Customer Ownership and Queue Length Awareness”, volume 26 issue 6, pages 2284 to 2304, published November 2024, retrieved 2026-08-28, https://doi.org/10.1287/msom.2023.0202
- Operations Research, Larson, “OR Forum-Perspectives on Queues: Social Justice and the Psychology of Queueing”, volume 35 issue 6, pages 895 to 905, published December 1987, retrieved 2026-08-28, https://doi.org/10.1287/opre.35.6.895
- Operations Research, Rothkopf and Rech, “Perspectives on Queues: Combining Queues is Not Always Beneficial”, volume 35 issue 6, pages 906 to 909, published December 1987, retrieved 2026-08-28, https://doi.org/10.1287/opre.35.6.906
- Which?, Aikman, “Unexpected item in bagging area: are self-checkouts on the way out?”, published 15 December 2023, retrieved 2026-08-28, https://www.which.co.uk/news/article/future-of-supermarket-checkouts-aAjrv9l0iwUu