Two SKUs share a container and the freight bill is 4,800 dollars. Assign it by weight and the heavy one carries the cost. Assign it by value and the expensive one does. The products are the same, the bill is the same, and the margin report will tell two different stories depending on a choice most teams never write down.
By weight is the honest choice when freight is genuinely priced on mass: dense, heavy products where the carrier's charge tracks kilograms. Atomix Logistics' landed cost guidance uses this kind of example, showing total cost components allocated against the units they actually cover. By value works when freight follows insurance and liability more than mass, typically for high value, light goods where a percentage of value mirrors the cost drivers. By units is fair only when the units are genuinely comparable in size, weight, and value.
Most catalogues are mixed, which is where the method breaks. A container holding both cast iron pans and phone cases should not be split by units, because the pans drive the freight bill and the cases ride along. Split by weight instead and the allocation follows the cost. The test is simple: does the driver you chose cause the cost. If the method assigns cost to items that did not create it, it is a distribution habit, not an allocation.
Write the method per freight lane and per product family, in the model itself, with the date it was set and who set it. Then test it once against an alternative method on your own data. If the choice flips a product from profitable to loss making, that product's margin is an artefact of the method and the team should know before it orders more.
Keep the allocation tied to the invoice. Freight, brokerage, handling, and insurance arrive as line items; each should have an owner, a basis, and a source document. A landed cost model built on a single lump sum labelled logistics, divided by an unwritten rule, cannot be audited by anyone, including the person who built it, three months later.
First is silent method drift. One analyst allocates a new product line by units because it is fast, another allocates the same line by weight next quarter, and the margin trend for that line is now measuring the analyst change. Second is allocation by sales value when value is really price: a discounted SKU looks cheaper to land because its price fell, which is backwards. Use cost or weight, not a price the market sets.
Third is allocating fixed costs like a per shipment fee across units that change. A 200 dollar brokerage fee spread over 20 units is 10 dollars a unit. Spread over 2,000 units it is 10 cents. Volume changes then look like margin improvements, and the purchasing team gets praised for a shipment size decision that changed nothing about the supplier.
Every quarter, take the five highest volume SKUs and the five thinnest margin SKUs and reallocate their shared costs under a different method. If the ranking of those ten barely moves, the method is a footnote and the business can trust the margins. If it flips, the method is the business, and the next pricing decision should wait until the team agrees which allocation reflects how costs actually arrive.
That check takes an afternoon and it is the only test most landed cost models ever need. Models do not fail because the arithmetic is hard. They fail because a method chosen in January for one product mix is still dividing costs in October for a different one, and nobody wrote down which method it was.
Keep reading: A worked landed cost with a stacked duty: one container, every line shown, MPF and HMF: the two small fees that quietly wreck cost models.
When freight cost genuinely follows mass: heavy, dense products where the carrier charges by kilograms. Allocating by units in that case loads cost onto light items that did not cause it and makes heavy items look cheaper than they are. Use the driver that causes the cost, weight, value, or units, and write the choice into the model.
Because sales value is a price the market sets, and discounts move it. A discounted SKU would look cheaper to land simply because its price fell, which is backwards. Allocate on cost, weight, or physical drivers, never on a price that changes for reasons unrelated to freight.
Reallocate the top volume and thinnest margin SKUs under a different method. If the profitability ranking flips, the method is doing the work, not the products. That is the signal to agree on the correct basis for each lane and product family and to document it with a date.
Brokerage fees, handling, insurance, merchandise processing and harbour maintenance fees where they apply, and the returns allowance where a return cannot recover duty. Each item needs a source document and an allocation basis. A model resting on one lump sum labelled logistics cannot be audited.