Spread Over What Arrived
When part of an order never turns up, the freight you already paid has to be carried by the goods that did. Ours spread it across everything ordered instead — so a shipment that arrived forty per cent short left every unit you actually had costed ten per cent under what it cost you, and left the rest of the money attached to goods that do not exist.
Suppose you order a hundred cases, and the freight to get them to you is a thousand dollars, and sixty cases arrive. The other forty are not on the next sailing. They are not coming at all — the supplier is out, or the balance was damaged, or shipping forty cases to a small island on their own costs more than the cases are worth.
The verdict
Every importer knows the shipment that arrives short and stays short. What is less obvious is that it is also a costing event, and one that most systems get wrong in the direction nobody checks — because a cost that comes out low is not a problem anybody reports. Ours got it wrong until August, and it got it wrong specifically for the businesses least likely to see a correcting delivery. The question to put to any vendor is small enough to ask in a demo and precise enough that they cannot talk around it: receive sixty of a hundred with freight attached, and tell me what one unit cost.
The thousand dollars has been spent. It brought sixty cases. So each of those sixty is carrying sixteen dollars and change of freight, not ten. That is not an accounting opinion; it is what happened to the money.
Ours divided by a hundred.
Where the four hundred dollars went
Nowhere, which is the interesting part. It was not lost, or double-counted, or posted to the wrong account. It was allocated to forty cases that were never received, and since those cases do not exist as stock, the cost attached to them is attached to nothing at all. It simply stops being visible.
What is visible is sixty cases, each of which looks ten per cent cheaper to have acquired than it was. And a unit cost that comes out low does not raise questions. It raises margins.
An overstated cost gets investigated. An understated one gets celebrated, and then it gets used to set a price.
That is the practical damage, and it is worse than a valuation error on the balance sheet. Somebody prices the line off a cost that is short. The margin looks healthy. It is healthy on paper for exactly as long as nobody reconciles the freight invoice against what the goods are carrying, which — as we found out about our own software — can be a very long time.
Why this is the Caribbean's version of the problem
Because whether the error corrects itself depends entirely on whether the missing goods ever turn up, and that is a question about geography and shipping economics rather than about software.
Where your supplier is up the road, a short delivery is a nuisance with a fixed ending. The balance arrives on Thursday, it gets received against the same order, and the allocation is recomputed across everything that came. The number was briefly wrong and then it was right. Nobody notices, and nobody needs to.
Across a region served by transshipment — where goods reach a hub port and then move island by island on smaller vessels, and where any given market may be a few pallets on a mixed sailing — a short shipment is frequently not a delay. It is the end of the matter. Nobody is going to move forty cases across a second leg on their own for a market this size, and everybody involved knows it. So the order is closed short, the balance is written off or re-sourced, and there is never a later delivery to trigger the recalculation.
The distinction that makes this a separate post
Our sister post on this defect, The Bill and the Sum of Its Parts, covers the opposite failure — one freight bill counted once per delivery, so an order arriving in two parts booked double. That one is about an error you cannot see because it looks like an expensive quarter. This one is about an error that never gets a chance to correct itself, because the event that would fix it does not happen. Same audit, opposite directions, different reasons they survive.
What the numbers actually did
| Ordered 100, freight 1,000, only 60 arrive | Before | After | The truth |
|---|---|---|---|
| Freight carried per unit | 10.00 | 16.67 | 16.67 |
| Landed unit cost at a 50 purchase price | 60.00 | 66.67 | 66.67 |
| Freight attached to goods that exist | 600 | 1,000 | 1,000 |
| Freight attached to goods that never came | 400 | nil | nil |
| Effect on a margin calculated from that cost | overstated | correct | — |
The fix is to divide by what arrived rather than by what was asked for, and to recompute across every delivery on the order as each one lands so that the total always comes back to the bill. Both halves were wrong; both are now correct and tested. An order that arrives complete, in parts, or permanently short now produces the same answer as the money.
The receiving habit this actually argues for
There is a practice worth taking from this that has nothing to do with which system you run, and it is about closing orders rather than leaving them open.
An order that is short and will stay short sits in two states depending on what you do with it. Left open, it is a promise of goods that are not coming, it distorts every commitment and expected-arrivals figure you read, and it invites somebody to receive against it a year later. Closed short, it is a completed transaction that cost what it cost — and it is also the moment at which the arithmetic in this post gets its final answer, because the quantity received stops being provisional.
We record short deliveries as themselves — on a goods received note that counts what arrived rather than copying what was ordered — and leave the order open for the balance, which is the right default, since refusing goods that are physically on the floor is how controls get bypassed. But the decision to stop waiting is a human one, and it is worth having somebody whose job it is to make it. That is also true of supplier documents that expire while an order is still open, which is argued separately in An Expiry That Stops an Order.
Receiving that records what arrived
A short delivery is recorded as the quantity actually counted, with the order left open for the balance rather than closed as though it were complete.
Freight spread over what arrived, not over what was ordered
Fixed August 2026 and held by a test. Goods that never came absorb none of the cost of getting the goods that did.
Allocation recomputed as each delivery lands
So a balance that does arrive later redistributes the bill correctly, and one that never arrives leaves the cost where it belongs.
Over-receipt refused, short receipt reported
More than you ordered is blocked with a configurable tolerance; less than you ordered is flagged and recorded, never refused.
Any prompt to close an order that will stay short
Not built. Nothing notices that an order has been part-received and untouched for months, and nothing suggests closing it.
Re-costing goods already sold when a balance is abandoned
Not built. If units sold before the order was closed short, they were costed on the information available then.
Two, and the first is a nudge rather than a feature
Both come straight out of the audit that produced this post.
A stale part-received order prompt
An order part-received and untouched past a threshold you set, surfaced for somebody to decide about — close it short, chase the balance, or re-source it. The decision stays human; only the reminder is automated. Nothing today distinguishes an order waiting a week from one waiting since March.
Costs charged against costs allocated
One line per order: what was billed, what the goods are carrying, and the difference. On a complete order it is zero. On a short-closed one it shows you exactly what the shortfall cost you, which is a number most importers would like to have and almost none can produce.
Neither of these is large. Both exist because a defect survived in our own product for want of somebody comparing two numbers that should have been equal, and we would rather ship the comparison than rely on having been careful.
Talk to us about receiving and costingWhat AWRA OpsHub does today
- Freight divided by what arrived, so the units you hold carry the whole cost of getting them to you.
- Allocation recomputed across every delivery on the order, so the total always reconciles to the bill however the goods turned up.
- Short deliveries recorded rather than refused, with the shortfall visible and the order left open for a balance that may still come.
- Over-receipts blocked by default with a tolerance you set, counted against everything already received rather than against each delivery on its own.
What it does not do
- No prompt to close a stale part-received order. An order short since March looks exactly like one short since Friday.
- No re-costing of units already sold when the rest of an order is later abandoned.
- No allocation by weight or volume, which is the wrong basis for freight billed by measurement.
- No reconciliation screen showing costs charged against costs allocated — the number whose absence is why the original defect survived.
Not ours, by choice
- We are not going to tell you when to give up on a balance. That is a commercial judgement about your supplier, your sailing schedule and what the goods are worth to you.
- We publish no tariff, duty or port charge figures for any market in this region. Where our copy carries a rate anywhere on this site it is dated and sourced, and these are not.
Four questions for a system that has to cost short deliveries
Receive 60 of 100 with freight attached. What is the landed unit cost?
What a straight answer sounds like
A number you can check against the freight divided by sixty.
Why it matters
This is the entire test and it takes five minutes. If the answer divides by a hundred, every margin on that line is overstated.
If the balance never arrives, does anything ever correct the cost?
What a straight answer sounds like
A clear yes or no about what triggers a recalculation.
Why it matters
Where balances routinely never arrive, an error that only self-corrects on the next delivery is permanent.
How do I close an order short, and does the system ever suggest it?
What a straight answer sounds like
A demonstration of the first. Probably a no to the second — ours is a no.
Why it matters
Open orders that will never complete distort every commitment figure you read.
Can I see freight charged against freight allocated, on one screen?
What a straight answer sounds like
A screen, or an admission that the two numbers are never compared.
Why it matters
A figure that should always be zero is the cheapest control there is, and almost nobody builds it.
Bring us the order that arrived short
A real one, with its freight and clearing charges and what actually turned up. We will land it in front of you and show you the unit cost, the shortfall, and where our allocation is still blunter than your lane deserves.
Talk to us about import costing