Put a parcel on the kitchen scale. It weighs two pounds. The invoice says you were billed for nineteen. Nothing went wrong, nobody made a mistake, and arguing about it will not help, because the courier is not charging you for the weight of the thing. It is charging you for the space the thing occupied on a van, and it has a formula for turning one into the other.
- Dimensional weight is length times width times height divided by a number in your contract. For US domestic parcels UPS and FedEx both use 139, and USPS moved to 139 on 12 July 2026.
- Since 18 August 2025 UPS and FedEx round every fractional dimension up to the next whole inch before the formula runs, so an eleven and a tenth inch side is billed as twelve.
- That rounding rule means the shape of your box matters more than its volume. A tenth of an inch on each side can move the billed weight by a pound.
- Choosing the best box for an order is a known hard problem in computer science, and the research literature is explicit that real constraints like stability defeat simple heuristics.
- From 1 January 2030 the EU caps empty space in e-commerce transport packaging at 50 percent, and filler counts as empty space.
- So cost and regulation now push the same way. A packing decision made once pays twice.
What is dimensional weight, and why does it decide your bill?
It is a billed weight calculated from size rather than mass, and it applies whenever it comes out higher than what the scale says. A courier's real constraint is van space, not axle load. A box of pillows fills a van long before it strains it, so carriers price volume and use the greater of the two numbers.
The formula is arithmetic, and the only variable that is not your decision is the divisor. Packizon's comparison of the UPS and FedEx dimensional weight rules for 2026 gives both carriers the same US domestic formula: length times width times height divided by 139, with each dimension rounded up to the nearest whole inch before the division.
The divisors as they now stand are worth having in one place, because a shop comparing quotes is often comparing two different divisors without realising it:
| Carrier and rate type | Divisor, inches and pounds | Note |
|---|---|---|
| FedEx, US domestic | 139 | Applies with dimensions rounded up to whole inches |
| UPS, daily rates | 139 | The rate most regular shippers are on |
| UPS, retail rates | 166 | A larger divisor means a lower billed weight |
| USPS | 139 since 12 July 2026 | Was 166, and only applies above one cubic foot |
| DHL, metric | 5,000 for centimetres and kilograms | Different units, same idea |
Look at the UPS rows. Scale Blog's account of how the divisors changed through 2026 lists 139 for daily rates and 166 for retail, which is a counterintuitive piece of arithmetic worth pausing on. A bigger divisor produces a smaller billed weight. Somebody shipping occasionally at retail rates can be charged for less dimensional weight than a regular shipper on daily rates, before any negotiated discount is applied. If you are comparing a quote to your own invoice, check which divisor each was built on before concluding anything.
The rounding rule costs more than the divisor
This is the change most shops have not absorbed. Scale Blog dates the shift to 18 August 2025, when UPS and FedEx both began rounding every fractional inch of a package dimension up to the next whole inch before applying the formula. USPS adopted the same rounding on 12 July 2026 for the parcels where dimensional weight applies to it.
Run the arithmetic on what that does. A box measured at 11.1 by 9.1 by 5.1 inches has a true volume of about 515 cubic inches. Rounded up on every side it becomes 12 by 10 by 6, which is 720. Divided by 139, the honest measurement gives 3.7 and the rounded one gives 5.2, both rounding up again to a final billed weight. Four pounds becomes six. Nothing about the parcel changed. The measurement convention changed.
The consequence for a small shop is specific and actionable: stop thinking about box volume and start thinking about box dimensions, in whole inches. A box that is 12 by 10 by 6 costs you nothing extra over one that is 11.1 by 9.1 by 5.1, so if a supplier offers a carton whose stated size has awkward fractions, the fractions are pure loss. Buying boxes whose dimensions are already whole numbers is a free saving that requires no software at all.
Can software actually pick a better box?
Yes, and it is genuinely hard, which is why the results vary so much between vendors. Choosing how to fit a set of items into the fewest or smallest containers is the three dimensional bin packing problem, and a 2024 paper on online 3D bin packing with reinforcement learning states flatly that the problem is NP-hard and significant in logistics, warehousing and transportation.
NP-hard is not marketing language. It means that as the number of items grows, finding the provably best arrangement becomes computationally impractical, so every working system is using an approximation. That is fine. An approximation that beats a human guessing is worth money. But it does mean nobody can sell you optimal packing, and a vendor claiming it does not understand its own product.
The same paper is useful about why the easy approaches fail in a real warehouse. It notes that traditional heuristic algorithms often fail to address dynamic and physical constraints in real time, and that its own contribution is integrating object rearrangement and stable placement. Stability is the word doing the work. A mathematically perfect arrangement that collapses when the van brakes is not a solution, and a system that has never modelled physics will confidently propose one.
So the honest description of what packing software does for a small shop is narrower and more useful than the pitch. It picks a good box from the boxes you actually stock, given dimensions you actually recorded. Both of those inputs are yours, and both are usually the reason the output disappoints. This is the same pattern as what route optimisation really turns out to be once you look at it, where the algorithm is the easy part and the data underneath it is the job.
Why optimising for volume is the wrong target
Most packing tools minimise wasted volume, because that is the natural objective and the one the research measures as space utilisation. Your invoice does not price volume. It prices the rounded product of three whole numbers divided by a divisor.
Those two targets usually agree and sometimes diverge sharply. A tool choosing between a 10 by 10 by 4 box and a 13 by 8 by 4 box sees 400 cubic inches against 416 and picks the first. On a divisor of 139 the first bills at 2.9 and the second at 3.0, so in that case the tool is right. Change it to a 16 by 6 by 4 box, 384 cubic inches, apparently the best of the three, and you may have created a parcel that trips a carrier's length or girth surcharge and costs more than any of them. Volume is a proxy. The bill is the thing.
For a shop with a dozen products, the fix is not better software. It is a one page table: for each product, the smallest whole inch box you actually stock that holds it safely, and the resulting billed weight at your divisor. That table takes an afternoon, never needs rebuilding until your range changes, and will beat a generic tool fed bad dimensions.
What changes on 1 January 2030?
Empty space becomes a compliance question rather than only a cost one. Reporting on the EU's Packaging and Packaging Waste Regulation describes a cap of 50 percent empty space in e-commerce transport packaging, with the regulation becoming applicable on 12 August 2026, full enforcement of the empty space limit from 1 January 2030, and harmonised calculation methods expected from the European Commission in February 2028.
Two details in that account matter more than the headline number. The first is what counts as empty space: any unoccupied volume inside the package, explicitly including air cushions, bubble wrap, foam and paper padding. Filler is not a solution to empty space, it is the definition of it. A shop that responds to a void by adding more paper is making the compliance position worse while spending money.
The second is that the rule also restricts packaging designed to create a misleading impression of volume, naming double walls and false bottoms. That is aimed at presentation packaging rather than a maker shipping from a spare room, but it is a clear signal about direction.
There are exemptions, and they are sensible rather than generous. Packaging may exceed the threshold where the extra volume is genuinely necessary to protect the goods, and certain designs tied to registered trademarks or product specific requirements may be excluded. A fragile ceramic in a well packed box is not the target. A small item rattling in a large carton is.
The reported threshold was nearly stricter. Early discussions of the regulation included a 40 percent cap before the final rule settled on 50 percent. If you are designing packaging now for a range you expect to sell for years, the margin between those two numbers is a reasonable amount of headroom to build in.
Where the cost case and the rules point the same way
Rarely do a cost saving and a compliance deadline ask for the same change, and here they do. Both want less air in the box.
| Change | Effect on your bill | Effect on the 2030 rule | Cost to do it |
|---|---|---|---|
| Buy boxes sized in whole inches | Removes loss created purely by rounding up | Neutral | Nothing, at the next reorder |
| Add two intermediate box sizes | Fewer orders shipped in a box too big | Directly reduces empty space | Some cash tied up in stock |
| Measure every product once, accurately | Makes any quote or tool trustworthy | Needed to prove a ratio later | An afternoon |
| Replace filler with a smaller box | Lower billed weight and less filler bought | Filler counts as empty space, so this is the fix | Nothing, often a saving |
| Check your divisor and rate type | Can change every parcel you ship | Neutral | One conversation |
The row worth acting on first is the third. Nothing else on that list works without it, quotes cannot be compared without it, and from February 2028 there will be a published method for calculating a ratio you will need product dimensions to satisfy. Recording length, width and height for every product you sell is the least glamorous data entry in commerce and it unlocks the rest.
Which surcharges hide behind the same numbers?
Several, and they are triggered by single dimensions rather than by volume, which is why a clever packing choice can cost more than a dull one. A parcel that is long and flat can stay well under any weight threshold while crossing a length limit that moves it out of standard handling entirely.
The pattern to watch for on your own invoices is a line item that appears on some parcels and not others of similar weight. That is almost always a dimension based charge rather than a weight based one. Once you know which of your products sit near a threshold, the fix is usually to change the box rather than the carrier, and it is worth knowing before you redesign packaging around volume alone.
There is a second reason to care about the longest side specifically. Because every dimension is rounded up before the division, error is not symmetric. Under measuring a short side by a tenth of an inch is harmless once rounding is applied. Getting the longest side wrong can move the parcel into a different handling category. If you are going to measure carefully in one place, measure the longest dimension of each product, and measure it with the packaging on rather than the product bare.
What about the boxes you did not choose?
Suppliers and fulfilment partners make packing decisions on your behalf, and their incentives are not identical to yours. A supplier shipping you stock cares about damage claims, not your onward parcel costs. A fulfilment partner packing your orders may be optimising for picking speed, since labour is their cost and postage is often yours.
That is worth one specific question when you sign or renew with a third party: who chooses the box, and on what rule. If the answer is that their system picks it, ask which objective it minimises and whether you can see the box assignment per order. If the answer is that a person picks it, ask what the box list is and whether you can add a size. Neither question is adversarial and both are hard to ask later.
The same logic applies to dropshipped and print on demand items, where the box is entirely outside your control and the dimensional weight is baked into a price you are quoted rather than a bill you receive. In that arrangement there is nothing to optimise, which is worth knowing so you do not spend an afternoon trying.
Does any of this need AI at all?
Less than the vendors suggest, and more than a spreadsheet at a certain size. The useful distinction is between measuring, choosing and predicting.
Measuring is where machine vision genuinely earns a place, because measuring a few hundred irregular products by hand is miserable and error prone, and a system that photographs an item and returns dimensions removes exactly the blocker described above. Choosing a box from a small set is arithmetic that a lookup table does perfectly, and an approximation algorithm only starts to pay when you have many products, many box sizes and multi item orders where the combinations get out of hand. Predicting, meaning forecasting which box sizes to stock for next season's order mix, is the part where a model reading your own order history beats intuition.
A one person shop with twenty products needs the table and nothing else. A shop shipping mixed baskets of fifty products needs real packing logic, and should ask any vendor two questions: what objective are you minimising, and does it use my carrier's divisor and rounding rule or a generic volume measure. A tool that cannot answer the second is optimising for something that is not on your invoice.
Worth remembering too that the parcel is the last step of a chain where the same data keeps mattering. Dimensions decide the billed weight, they decide whether a customs declaration is plausible when you ship abroad, and they decide whether the delivery promise on your product page was ever realistic. We looked at the wider version of that in the decisions that sit between an order and the doorstep, and at the paperwork side in what a customs code on your parcel is actually declaring.
What to do this week
Pull your last twenty invoices and compare the billed weight against the actual weight on each. If the billed figure is higher on more than a handful, dimensional weight is setting your shipping cost and every box decision is a pricing decision.
Then measure your five best sellers with a ruler, round every dimension up to the next whole inch yourself, divide by the divisor on your contract, and see whether the box you currently use is the smallest one in your cupboard that would hold the item safely. Most shops find one product shipping in a box two sizes too large because that is the box that was to hand the first time somebody packed it, and nobody revisited the decision. A shop that holds its own product and order records rather than renting them can run that comparison across its whole range from data it already has.
None of this is interesting work. It is arithmetic, a ruler and a reorder. It is also one of the few remaining places where a small shop can take a real cost out without raising a price or selling more, and the rule arriving in 2030 means the work is not optional for much longer.