Why the yard goes from fine to broken so suddenly
Waiting at a dock does not rise in proportion to the number of trucks. It rises slowly while the doors have slack and then very steeply once they do not, and the transition is narrow enough that an operation can cross it in a single quarter without anything visibly changing.
The mechanism is randomness rather than volume. If forty trucks arrived at exactly evenly spaced intervals and every one took exactly fifty-five minutes, five doors would handle them with no queue at all up to the point of literal saturation. Trucks do not arrive that way. They cluster, they are late, and one container takes three hours while the next takes twenty minutes. When the doors have spare capacity, a cluster gets absorbed and forgotten. When they do not, the cluster has nowhere to go and the delay it causes is still there an hour later, on top of the next cluster.
That is why the door count table on this page matters more than the minutes in it. Take a door away at seventy percent utilization and you add a few minutes. Take one away at eighty-five and you add a lot. The curve is the finding; the specific numbers come from a model that does not know your yard.
What is actually being modelled
A standard multi server queue. Trucks arrive at random at some average rate, each takes an average time at a door, and a truck that arrives when every door is busy waits for whichever frees up first. From those three things the model gives the chance that an arriving truck has to wait at all, the average length of that wait, and how many trucks are sitting in the yard at any moment.
The assumptions are visible and each one is wrong in a useful way. Random arrivals overstate clustering if you run tight appointment windows and understate it if half your inbound turns up unannounced before nine. A single average service time flattens the real distribution, where the long tail of hard-to-unload loads does most of the damage. Every door being interchangeable ignores carrier assignments, temperature-controlled doors, and the one door that has the good dock leveller.
Corrections to all of that exist and are not on this page, because the shape does not change and the shape is the whole message: below a threshold, extra volume is absorbed; above it, extra volume becomes waiting, quickly.
The peak is the number that bites
Average utilization across the whole receiving window is a comforting figure and mostly a fiction. Inbound freight arrives in the morning because drivers start early, so a dock at sixty-five percent for the day may be at ninety-five percent until noon and nearly idle after three. Every hour of that queue happens in the morning and every idle crew hour happens in the afternoon, and the daily average shows neither.
The peak field on this page is a crude handle on that. Set it to fifty for genuinely even arrivals and watch what happens when you set it to seventy. Appointment scheduling is the intervention that targets this specifically, and it works by moving arrivals rather than by adding capacity — which is why it is usually the cheapest thing on the list and the hardest to get carriers to accept.
Both sides of the tradeoff cost money
It is tempting to treat waiting as the problem and doors as the solution, and then the arithmetic on the staffing side arrives. A door is only a door when a crew, equipment and somewhere to stage the freight are behind it. Size the dock so that no truck ever waits and you are paying crews to stand at idle doors for a large fraction of the day. Size it tight and you pay in detention, in drivers who will not book with you, and in receiving backlogs that turn into inventory nobody can find.
Neither cost disappears. What the model does is put the two on the same page in the same units so the choice can be made deliberately instead of by whichever one complained most recently. And there is a third lever that neither column shows: reducing the time at the door itself. Cutting fifty-five minutes to forty-five raises effective capacity by more than a fifth without a single dollar of construction, which is why palletisation, pre-advice and unloading method are usually worth attacking before door count. What arrives on those pallets is covered by the pallet load planner, and where it goes once it is off the truck by the pallet rack layout calculator.
Questions people ask
How many dock doors do I need?
Work it from door hours rather than from a ratio. Trucks per day multiplied by average time at the door gives the door hours you demand, and doors multiplied by the receiving window gives what you have. The ratio between those two is utilization, and you want it comfortably below the point where waiting starts climbing steeply — the table on this page shows where that is for your own numbers. Anything quoted as doors per thousand square feet is describing somebody else's freight profile, since a building receiving floor loaded containers and one receiving palletised freight need completely different door counts for identical square footage.
Why does the wait explode above about 85 percent utilization?
Because a busy system has no slack left to absorb the randomness. Any queue with variable arrivals or variable service times has an average waiting time that includes a term roughly proportional to utilization divided by one minus utilization, and that term climbs without limit as utilization approaches one. At seventy percent the divisor is 0.3; at ninety percent it is 0.1, three times smaller. Every cluster that arrives at high utilization has to wait for capacity that is already committed, so the delay it creates is still being worked off when the next cluster arrives. There is no threshold in the equipment or the crew, and nothing changes on the dock at that point. It is a property of the arithmetic.
Does the model handle appointment scheduling?
Not directly, and appointments are one of the largest available improvements. Scheduling replaces random arrivals with something much more regular, which cuts waiting substantially at the same volume and the same door count, because most of the waiting was caused by clustering rather than by volume. The peak share field is the nearest handle available here: a scheduled dock behaves like a low peak share and an unscheduled one like a high peak share. Compare the two settings and the gap approximates what scheduling is worth before spending anything on doors.
Is the detention figure a forecast of what I will be billed?
No, and it will usually be low. Detention is charged on individual trucks that exceed free time, so it depends on the tail of the waiting distribution while this page uses the average. An operation whose average truck clears well inside free time can still pay detention every week, on the handful of trucks that arrived during the morning cluster behind a difficult unload. Treat the figure as an order of magnitude and a way to compare scenarios. Your actual exposure comes from your carrier agreements and your own detention invoices, which are the only reliable source for what your terms are.
Should I add a door or speed up unloading?
Speed first, almost always, because it is cheaper and it improves both columns at once. Cutting time at the door raises effective capacity without construction, without another crew, and without another dock leveller, and it reduces detention and idle time together. The usual sources are on the shipper's side rather than yours: floor loaded trailers instead of palletised, no pre-advice so nothing is staged, paperwork settled at the door instead of beforehand. Doors are the right answer when volume has genuinely outgrown the building, and by then the number is easy to justify because the queue is visible from the road.