The box is chosen before the order exists, and that is the problem
Carton selection is usually treated as a packing decision made at the bench. In practice it is a cost decision made months earlier, when the box range was bought and the SKU dimensions were entered, and the bench only executes what that decision left possible.

Billable weight is the larger of two numbers
Every parcel is priced on the greater of its actual weight and its dimensional weight. That sentence is the whole subject. A light bulky item is priced as if it were heavy, and a dense small item is priced on its mass. Neither number can be improved at the bench once the box is closed.
Dimensional weight is length times width times height divided by a divisor set by the carrier. The divisor is not a constant across the industry, and it has been falling for years as carriers move more volume into the dimensional model. A box that was economical under one divisor is expensive under the next.
The practical consequence is that the levers are all upstream. Reduce the box, reduce the void fill, or increase the density. Reducing the box is the only one that changes both the dimensional weight and the material cost at the same time, which is why box range design deserves more attention than it usually gets.
The measurement itself has a tolerance. A box measured at thirty point four centimetres rounds into a higher band than one measured at twenty nine point nine, and the difference per parcel is small until it is multiplied by the annual parcel count, at which point it is a line item.
The divisor is a contract term, not a constant
Carriers publish divisors of one hundred thirty nine, one hundred sixty six, five thousand and six thousand depending on the service, the unit system and the account. The same parcel can be billable at eight kilos under one divisor and five under another, and neither carrier is being dishonest.
Which divisor applies to a given shipment depends on the service level and the negotiated agreement. A merchant comparing carriers on the headline rate without converting both to the same divisor is comparing two different products, and the comparison will be wrong in the direction of whichever divisor is more generous.
Negotiated divisors are a real commercial lever. A high volume account can sometimes move from a retail divisor to a more favourable one, and the effect on billable weight for bulky categories is larger than most rate discounts. It is worth asking for explicitly rather than hoping it appears.
The internal consequence is that the dimensional weight calculation has to be configurable rather than hard coded. A warehouse system that embeds one divisor will produce the right number for one carrier and a systematically wrong number for every other one it serves.
How many box sizes the operation can actually carry
A large box range fits more orders tightly and creates two costs that are easy to miss: the bench has more choices to make and the picker has more bin locations to search. Both increase the error rate, and the error rate increase often exceeds the freight saving.
The economic point is where the marginal box size stops paying for itself. If adding a size reduces the average dimensional weight by two percent on the orders it serves but increases mis-packing by a rate that generates one re-ship per thousand orders, the two effects are close and the decision becomes one of management capacity rather than arithmetic.
A practical range for a mid size operation is eight to twelve sizes with a clear rule about which one applies to which order profile. The rule matters more than the range, because a bench that chooses freely will converge on two favourite boxes regardless of how many are on the shelf.
The rule should be expressed as a mapping from order volume and the largest item's dimensions to a box, computed by the system rather than judged by the packer. When the computed box is wrong, that is a mapping error to be fixed once, rather than a judgement error to be repeated every shift.
Fill rate is the number that decides whether air is being shipped
A carton that is ninety percent full by volume is efficient. A carton that is sixty percent full is shipping air at the dimensional rate. The gap between those two states, across a catalogue, is usually the largest single freight saving available without renegotiating anything.
The target band for most operations is eighty five to ninety percent volumetric fill, with the constraint that the goods must be protected rather than compressed. Fill rate computed on the item volume alone understates the achievable number, because the packing material occupies volume too.
Void fill is a cost with two components: the material and the volume it adds. A box filled with loose paper reaches the right protective state and pays for a larger dimensional weight than the same protection delivered by a smaller box with a fitted insert. The insert is usually the cheaper of the two at volume.
Fill rate should be measured on a sample of shipped orders rather than assumed from the box range. Measure the box's internal volume, the item volume, and the fill material volume, and the residual is the air. A monthly sample of one hundred orders is enough to see the pattern.
Cushioning only works when it is in contact
A protective layer that does not touch the product does not protect it. Two centimetres of clearance above an item means the item accelerates across that gap before it meets the cushion, and the impact it experiences is the one the cushion was designed to prevent.
The drop test standard assumes the item is restrained rather than free. A box that passes a one point two metre drop with the product immobilised can fail the same drop with the product loose, because the failure mode changes from compression of the cushion to impact against the box wall.
Which is why the pack out instruction should specify contact rather than a material. Telling a packer to add two sheets of bubble wrap is a material instruction. Telling them that the item must not move when the closed box is shaken is a contact instruction, and the second one produces the intended result regardless of what material is used.
Suspension and retention packaging solve this structurally by holding the item at defined points rather than surrounding it. They cost more per unit and they are worth it for fragile high value items where the volume of cushioning would otherwise push the box into a higher dimensional band.
Where consolidation stops being worth it
Combining two orders into one parcel saves a first leg charge and adds dimensional volume. The saving is roughly fixed per parcel while the cost scales with the added volume, so the economics depend entirely on how dense the combined goods are.
Two dense items consolidate well because the added volume is small relative to the added weight. Two bulky light items consolidate badly because the combined dimensional weight can exceed the sum of the two separate dimensional weights, since the new box has to be sized for both at once.
The break even point is where the incremental dimensional weight cost equals the saved first leg charge. That calculation is simple and it should be run per order profile rather than as a policy, because a policy that always consolidates will lose money on exactly the orders that look most attractive to consolidate.
Consolidation also has a service cost. Holding one order to wait for another delays the first, and the delay may cross a cutoff. The consolidation decision belongs after the cutoff logic rather than before it, otherwise the freight saving is paid for with a late delivery.
Board grade and stacking are the same decision
Box strength is specified as an edge crush test value, which describes how much force a wall can take before it buckles. Stacking a pallet six high multiplies the load on the bottom carton by five, and a board grade chosen for single parcel delivery will fail under that load.
The relationship is not linear in practice because humidity degrades board strength substantially. A specification that passes in a dry warehouse can fail in a humid one, and the failure appears as collapsed cartons at the bottom of a pallet rather than as a material defect.
The practical specification is to state the load the carton must bear rather than the board grade, and let the supplier propose the grade that meets it at the ambient humidity of the destination warehouse. Stating the grade transfers the engineering decision to someone who does not know the stack height.
For parcel networks the relevant test is different again, because parcels are sorted on belts and chutes rather than stacked. The failure mode there is corner crush and burst, which responds to the corner design and the tape pattern more than to the board grade alone.
Measurement is a process, and a centimetre moves money
Carton dimensions enter the system from three places: the supplier's specification, a manual measurement, and an automated dimensioner. The three disagree, and the disagreement is the reason a box can be billed at one band in one warehouse and another band in a second.
Manual measurement with a tape has a realistic repeatability of about half a centimetre, which is enough to cross a band boundary on a box that sits near one. Automated dimensioners are more repeatable and they are consistent, which matters more than absolute accuracy for billing disputes.
The rule should be that the dimension on the record is the one the carrier will measure, including any bulge from overfilling. A box specified at thirty centimetres that bulges to thirty two is a thirty two centimetre box for billing, and the record should say so.
Round in the carrier's favour when in doubt. It is cheaper to price the parcel at the higher band consistently than to win a dispute and lose the account relationship over a systematic under-declaration.
Every packaging change is a re-measurement event
A change of box supplier, a change of board grade, a change of tape pattern or a change in the void fill all change the measured dimensions or the billable weight. Most operations change all four over a year and re-measure none of them.
The re-measurement trigger should live in the packaging specification rather than in a person's memory. When a specification version changes, the dimension and weight records for every SKU packed in that box are flagged for re-measurement before the next shipment.
The cost of skipping the re-measurement is asymmetric. An over-declared dimension is a small ongoing overcharge. An under-declared dimension is back-billed adjustments, which arrive months later, in bulk, and are very hard to attribute to a specific change.
Treat the packaging specification as a controlled document with versions and effective dates, and the re-measurement becomes a workflow rather than a project. Treat it as tribal knowledge and the dimensions in the system will describe boxes that no longer exist.
The cost model that belongs in the SKU record
The dimensional weight calculation should be available at the SKU level, not only at the order level. A merchant deciding whether to add a colour variant needs to know what that variant will cost to ship, and that number depends on the box it will be packed in.
Three fields make this possible: the packed dimensions of a single unit, the packed weight, and the box family the unit is assigned to. From those, the system can compute a per unit ship cost at each carrier's divisor and expose it in the buying decision.
That number changes the sourcing conversation. Two suppliers quoting the same landed cost for the same product can have different ship costs if one packs the item in a retail box that adds two centimetres on each dimension. The retail box is a freight cost, and it should be compared as one.
The model should also carry the fill material assumption, because a change from loose fill to a fitted insert changes the achievable box for a whole product family. Updating the assumption is a one line change in the model and a significant change in the answer.
Where the calculation is carried through to box selection per order profile, the working method is the dimensional weight and carton right sizing for parcel fulfilment, and the divisor per carrier is the first thing it has to get right.
- 01Compute billable weight as the greater of actual and dimensional, and show both.
- 02Store the divisor per carrier and per service rather than as a global constant.
- 03Ask for a negotiated divisor explicitly on bulky categories.
- 04Cap the box range at a size the bench can choose from reliably, and compute the choice in the system.
- 05Measure volumetric fill on a monthly sample of shipped orders.
- 06Specify contact rather than material in the pack out instruction.
- 07Shake test the closed box rather than trusting the cushion thickness.
- 08Run the consolidation break even per order profile instead of as a policy.
- 09Specify the load a carton must bear rather than the board grade.
- 10Record dimensions as the carrier will measure them, including bulge.
- 11Flag every SKU for re-measurement when the packaging specification version changes.
- 12Expose per unit ship cost in the SKU record and use it in sourcing decisions.