I. The Ladder Essay 4

The Queue Is Not the People

System thinking

19 October 2026 12 min read Seen again: a queue at a school canteen

At twelve twenty, the canteen at a school becomes a corridor of people who have all decided, independently and reasonably, to be hungry at the same moment. A hundred and forty students, two serving hatches, and a lunch break of thirty-five minutes. Something has to give, and the thing that gives is a queue.

It is tempting to describe the queue in terms of the people in it. Some are slow. Some push in. Some are talking. And there is a student near the back who is in a bad mood about something and is taking it out on the person in front. If you watch the queue for ten minutes, all of this is visible and none of it is the explanation.

The explanation is that the canteen has one hatch serving sandwiches and one serving everything else, and everybody has to pass the same narrow point to reach both. The queue is not a crowd of badly behaved people. The queue is the shape of a system with a bottleneck in it, and the people in it are the most reasonable actors available.

Take the same hundred and forty students and change the structure. Put one extra person behind a second hatch, and split the queue in two at the point where the two hatches diverge, so that the queue for sandwiches goes left and the queue for everything else goes right. Nothing about the students changes. They are not more patient, they are not more polite, they are not more organised. They are the same hundred and forty people with the same appetites at the same moment in a building with one extra pair of hands on a Saturday.

What changes is the wait, and it changes a lot. In one queue serving one hatch, the last person waits for everybody ahead of them. In two queues, the last person waits for only the people in their own half. If serving one person takes twenty seconds, the last of a hundred and forty is looking at close to forty minutes of waiting, against a thirty-five minute break. The arrangement makes the lunch break mathematically impossible. The second arrangement does not. The students did not improve. The system did.

Structure before people

System thinking explains behaviour by structure: the parts, the relationships between the parts, the stocks, the flows, the constraints, and the boundary you have drawn around the thing. The claim is not that people do not matter. It is that when you want to predict behaviour in a situation you have not seen, structure predicts it better than character does.

This is the shift in blame. When the queue is long, the instinct is to look for the person causing it: the one pushing in, the one paying with coins, the one asking a question. System thinking redirects the question from who to how. It asks what the arrangement makes easy and what it makes hard. It asks where the constraint is, and whether the constraint is the thing you are actually working on.

The distinction matters because the two kinds of fix have very different reach. Telling the pushing student to be more polite fixes one student. Widening the hatch, splitting the queue, or moving the till fixes a hundred and forty of them, including all the ones who were never going to push in. The first fix depends on the second one behaving well. The second does not depend on anybody.

Stocks and flows

The simplest system is a bathtub with a tap and a drain. Water in, water out. The level rises when more comes in than goes out, and falls when less does. The level is a stock: a quantity that has accumulated. The tap and the drain are flows: rates of change. The whole model is the observation that the level depends on the difference between two rates, and that the difference can be small and constant, or large and temporary, and that the level looks the same either way while you are looking at it.

This is why system thinking feels strange at first. The level of a bathtub tells you almost nothing about the tap. A full bath could mean the tap has been off for an hour, or that it has been on for two minutes. If you only look at the level, you cannot tell whether to act. You have to look at the flows.

Fig. 1 A tank with a stored amount inside it. An inflow pipe enters at the left, an outflow pipe leaves at the right, and a line marks the level above the stored amount.

Queues are stocks of people waiting. The number waiting is a stock. The rate at which the hatch serves people is a flow out. The rate at which students arrive is a flow in. If the arrival rate is persistently above the service rate, the queue grows without bound, and no amount of goodwill in the queue will change that. If the arrival rate is below the service rate, the queue drains, and if the service rate is much higher, the queue empties and people wait idle at the hatch, which is waste in the other direction.

There is a stock hiding in almost every complaint about a service. A complaint about waiting is usually a complaint about a stock, and stocks respond badly to being told to hurry.

Bottlenecks and constraints

Through the hatch, one person every twenty seconds. That is the constraint. Everything else in the system is free to be fast, and it makes no difference to the output, because the output is set by the slowest part. Making the person who chooses the sandwich faster does nothing. Making the person who queues more orderly does nothing. Making the hatch faster changes everything, because everything downstream of the hatch is waiting on it.

What if the canteen kept one queue but added a second hatch, with the second hatch opening only when the queue was long? The structure is nearly the same, and the effect is much smaller. With a single hatch, the hundredth student waits for the ninety-ninth. With two hatches, the hundredth waits for the ninety-eighth, because the ninety-ninth is being served at the other hatch. The wait drops by one service time, not by half. Two hatches opened at the same time and fed from one queue would halve the wait. The same two hatches, arranged differently, do very different amounts of work.

Now change a different condition: what if the bottleneck were not the hatch at all, but the door at the end of the serving line, where students take a tray, find a space, and walk away? Nothing at the hatch changes, and the queue does not shrink. The constraint moved. A system thinking check asks where the constraint is now, not where it was last time.

"System thinking means ignoring people." It does not. It means asking what the arrangement is doing to the people before asking what the people are doing to the arrangement. A queue of people who are all in a hurry is not a moral failing. It is a measurement of how much capacity the system has, expressed in the only unit the system understands.

The second misconception is that drawing a diagram of a system improves it. Diagrams are cheap and seductive. A box-and-arrow picture of a school looks impressive and makes it look as if the box for "communication" is a thing you can go and improve. What actually changes behaviour is a change in capacity, a change in a rule, a change in a sequence, or a change in what gets measured. If the diagram does not lead to one of those, it is a picture of a problem, and the problem is exactly where it was before the picture was drawn.

Fig. 2 A wide channel on the left carries square items. It narrows to a thin opening, the items are backed up in a queue, and a single line leaves to the right as output.

Where you draw the boundary

This is the part of system thinking that changes the most about how you look at things, and it is the easiest to get wrong. Every system you examine has a boundary, and the boundary is a choice. The boundary decides what is inside, what is outside, what can be influenced, and what is simply given.

Take a teacher and a student. Where does the system end? If the boundary is drawn around the classroom, the teacher is a person with a technique, and the problem is that the technique is not good enough. If the boundary is drawn around the school day, the teacher is part of a timetable, and the problem might be that the timetable gives the teacher four consecutive periods of the hardest class. If the boundary is drawn around the family, the teacher is one of several influences on whether the student does their reading. Each boundary gives a different answer to the same question, and each answer is true inside its own boundary and incomplete outside it.

Fig. 3 The same counters and the same people in two panels. In the first, a dashed boundary encloses only the counters. In the second, it encloses the counters and the people together.

Boundaries also have a direction of effect. Whatever is outside the boundary arrives from somewhere and goes somewhere, and the flows across the boundary are what you can actually influence. If you draw the boundary too tightly around the thing you control, you will spend all your effort on the part you control and ignore the part that actually determines the outcome. This is the mistake of the shop that optimises its packing table and wonders why customers still queue at the door.

The most common way to get the boundary wrong is to draw it around the people rather than around the work. When the canteen manager draws a boundary around the serving hatch, the students are outside, and the students become the source of the problem. When the boundary is drawn around the lunch break, from the moment students leave a classroom to the moment they sit down, the students are inside, and the queue becomes a thing the system does to them rather than a thing they do. Nothing about the students changed. The boundary moved, and the blame moved with it. That is the practical payoff of getting the boundary right, and it is available at almost no cost.

Why fixing one part breaks another

A school timetable is a constraint system with no slack. Every period is used, every room is occupied, every teacher is placed exactly once, and there is no spare capacity anywhere because a spare period is a class that does not happen. In that system, the only free moves are exchanges: one arrangement can be better than another, but not by adding anything.

Now the school notices that one class is too large, and moves a student into another form. The timetabler goes to work. The student was in a subject taught in room 14, and the new form is taught in room 3 at a different time, so the student now has a gap. The gap is filled by moving another student, who is now in a different form, and the chain runs for two hours. By the end of the day the timetabler has balanced the class sizes and created a timetable with a gap in it, a teacher with a free period, and a student who now has a free period too, which they will spend, you can bet, in the corridor.

Nothing was done wrong. Each change was locally correct. The system responded the way a tightly coupled system responds, and the improvement in one place became a new problem somewhere else. This is the ordinary fate of structural changes in systems with no slack, and it is why the question "what does this break?" belongs next to the question "what does this fix?"

Fig. 4 The same four people. On the left, one counter and one queue, so the last person has three ahead. On the right, two counters and two queues, so the last person has one ahead.

Limits of models

A model is a simplification, and every simplification is wrong somewhere. A diagram of a school is not the school. The danger is not that models are wrong. The danger is that a model that works well enough for one purpose gets used for another purpose it was never fitted for, and the mismatch is invisible from inside.

Every model has a purpose built into its shape. A queueing model with one server is fitted to a canteen with one hatch. Hand it a canteen with two and it will tell you confidently that the wait is the same, which is false. The model did not lie. It answered a question about a different system, and you did not notice because the answer came out in the same units.

Second limit: a model cannot show you a solution you did not think to draw. If the diagram has no box for information flow, or no box for the queue itself, then no amount of studying the diagram will reveal an information problem or a queue. The model constrains the search. That is what makes it useful and what makes it dangerous.

Limits

System thinking explains behaviour after the fact better than it predicts before. Once you know the structure, the behaviour is easy to read. But you often only learn the structure by watching the behaviour fail, and by then you have already paid for it. System thinking is a way of paying less often, not a way of never paying.

And structure is not the whole of what happens. Two systems with identical structures and identical parts can behave differently if the parts themselves differ. The queue theory says the average wait depends on the variability of arrival, not only the mean. A school where everyone arrives at once and a school where arrivals are spread out have the same average, the same service rate, and very different waits for the unlucky. Structure sets the frame. The parts still speak.

Stuck

You are stuck in system thinking when you can describe the structure perfectly and still cannot get anything done. The symptom is the diagram that keeps being redrawn. Another box, a different arrow, a nicer layout. The system has become the object of attention, and the people inside it are now a detail.

The way out is to ask what a change would do, and then make a small one. System thinking without a change in the world is a hobby. Pick the constraint. Ask what would have to be different for the constraint to move. Change that one thing, even slightly, and watch what happens for a week. A system that will not answer that question is a drawing, not a system.

Sources

The stock and flow vocabulary, the idea of a constraint, and the idea of system boundaries are all standard across the systems thinking tradition.
Donella Meadows wrote about leverage points, the places in a system where a small change makes a large difference. That is the origin of the phrase, and it is used here in that sense.
All numbers in the canteen example are illustrative, not measurements from a real school.