The Snaking Line, Solved: Queue Arithmetic for Parcel Season

On the second Saturday in December, the post office on the precinct stops being a shop and becomes a holding pen. Somewhere near the door, a laminated sign apologises for the wait. A child counts the people ahead. A man in a raincoat sighs at his ticket, number 47, while the display insists it is serving number 12.

If you have ever stood there wondering whether to stay or bail, you already have the raw material for a bit of algebra. Queueing theory sounds like a postgraduate module, but the version you need at the door is a single equation with three moving parts. It is called Little's Law, and it is the reason you can make a decent guess about your wait before you commit.

A queue is a rate with a tail

Start with units, because units are where most queue arguments go wrong. Arrival rate — call it λ, the Greek letter lambda — is measured in customers per minute. Service rate, μ, is also customers per minute, per counter. Queue length, L, is people. Waiting time, W, is minutes.

Keep those units in view and the algebra almost writes itself. People divided by people-per-minute gives minutes. That is the whole trick.

Little's Law without the lecture

Little's Law says that, on average, the number of people in a system equals the rate at which they arrive multiplied by the time each one spends inside it:

L = λW

In words: if people pile in faster, or linger longer, the line gets longer. Obvious, in a way, but the equation earns its keep when you rearrange it. Divide both sides by λ and you get:

W = L / λ

Now you can estimate your wait from two things you can actually count: how many people are ahead of you, and how quickly new people are joining.

There is a second version, more useful at a counter. If the queue is stable, the rate at which people leave equals the rate at which they arrive. The leaving rate is set by the service rate μ. So the time you spend waiting is roughly the number ahead of you divided by the combined service rate of the open counters:

W ≈ L / μ

That is the one to use when you can see how many tills are staffed.

How to time a branch in two minutes

You do not need a clipboard. Stand near the ticket machine and count how many people take a ticket in a ten-minute window. Divide by ten. If 14 people joined in ten minutes, λ is 1.4 per minute. Round to 1.5 and you are close enough.

Next, service rate. Watch one counter and time three transactions end to end — the rummage for the parcel, the label, the card reader, the receipt. If they average four minutes, that counter serves 0.25 people per minute. Three counters open gives μ = 0.75 per minute.

Then count the line. Be honest: include the people at the counters as well as those in the snake. If 14 are waiting and three are being served, L = 17. Your estimate is:

W ≈ 17 / 0.75 ≈ 23 minutes

Twenty-three minutes is not a disaster. It is also not nothing on a Saturday in December.

The bit people forget: is the queue growing?

The estimate above assumes the line is roughly steady. It is not, if arrivals outpace service. Compare λ with μ. If λ is 0.5 and μ is 0.75, the branch is coping and the wait you calculated will hold. If λ is 0.9 and μ is 0.75, the queue is filling faster than it drains. Your 23 minutes is a lower bound, not a forecast.

Parcel season is exactly when λ spikes. Last posting dates create a cliff edge: the week before them, arrival rates at counters jump, and the same branch that felt fine in November becomes a different equation in December.

Rearranging to answer the actual question

The useful move is not the arithmetic. It is the rearrangement. Ask: what would make this wait shorter?

Lower L: come back when fewer people are in the room. Mid-morning lulls are real, and the ticket machine data usually shows them.

Raise μ: find a branch with more counters open, or a sub-post office tucked inside a corner shop where the parcel traffic is lighter.

Lower λ: do not add yourself to a queue that is already over capacity. If λ exceeds μ, joining makes it worse for everyone behind you, and you will wait longer than the person who walked away.

Try the comparison between two high streets. Branch A has 20 people and two counters at four minutes each: μ = 0.5, so W ≈ 40 minutes. Branch B has 12 people and three counters at five minutes each: μ = 0.6, so W = 20 minutes. Branch B looks busier per counter but is the better bet. Raw line length is a red herring; the ratio is what matters.

What the algebra cannot tell you

Queue maths is honest about averages and quiet about everything else. It does not know that the person at the front has four international parcels and a passport renewal. It does not know that the ticket system hides the true line, so your count of 17 is really 24. It does not know about the counter closing for a break, or the customer who needs extra time and deserves it.

Treat the number as a starting point, not a verdict. A wait of 23 minutes with a book is a different experience from 23 minutes without one. And if you are popping in just quickly on 20 December, the maths would like a word.

A small habit worth keeping

The next time you hesitate at the door, do the two-minute count. Take the tickets, time the counter, divide. You will not always be right, but you will be less wrong than the person guessing by feel. And you will have used a rate, a unit and a rearrangement to make a better decision about your own afternoon — which is, more or less, what algebra is for.