A personal care brand sells one 8oz bottle four ways. On their own site it ships in a branded box with tissue and a card. On Amazon it needs a polybag with a suffocation warning and a scannable FNSKU. To a regional grocery chain it goes in a 12-count case with a GS1-128 label and a specific pallet tie. To a subscription box partner it ships loose, unlabelled, in bulk cartons of 48.

Same liquid. Same bottle. Four packaging specifications, four sets of consequences for getting it wrong, and — until recently — one shift supervisor who knew all four by heart.

When she left, the brand took $14,000 in chargebacks in a quarter.

The problem is that packaging is a channel attribute, not a product attribute

Most systems model packaging on the product. One product, one set of dimensions, one weight, maybe a case quantity. That model is fine right up until the same product goes out through more than one door, at which point it quietly stops describing reality.

What is actually true is that packaging is a property of the product and the destination together. The bottle has no single answer to “how is it packed” — it has four, selected by where it is going.

Systems that cannot express that force the knowledge somewhere else, and “somewhere else” is always one of three places: a person’s memory, a laminated sheet at the pack station, or a spreadsheet nobody has opened since the person who wrote it left. All three fail the same way, which is silently and at the worst possible moment.

What each channel actually demands

The specifics vary by partner, but the categories are consistent.

Channel	Typical requirements	Failure mode
DTC	Branded outer, insert, protective fill, gift options	Bad reviews, higher return rate
Marketplace (FBA-style)	Polybag with warning above threshold, item labelling, prep by category, shipment-level labelling	Prep fees, receiving delays, unfulfillable inventory
Retail / wholesale	Case pack quantity, inner pack, case labelling to spec, pallet configuration, ASN matching the physical pallet	Chargebacks, refused deliveries, scorecard damage
Subscription / B2B bulk	Bulk cartons, minimal packaging, count accuracy over presentation	Partner rework billed back to you

The retail column is where the money is, because retail is the only one that fines you. A marketplace charges a prep fee and moves on. A grocery DC that receives a pallet whose ASN does not match the physical configuration issues a chargeback, and does it again next time, and the scorecard damage affects the next buying conversation.

Model it as data or lose it to turnover

The fix is conceptually simple and organisationally hard: packaging specification has to be a record in a system, keyed on the combination of product and destination, and it has to be what the warehouse sees at the moment of packing.

The record needs to carry at least:

• Pack configuration — units per inner, inners per case, cases per pallet, pallet tie and height
• Labelling — which label types, applied where, generated from what data
• Materials — the specific outer, fill, and inserts, as consumable SKUs so they can be forecast and reordered
• Dimensional and weight profile for that configuration, since it differs per config and drives freight
• Compliance flags — suffocation warnings, lot or expiry marking, country of origin, anything the partner’s routing guide mandates

That last point is the one that turns packaging from a warehouse concern into a systems concern. The routing guide is a compliance document. Encoding it as data is the only way it gets applied consistently, because the alternative is asking a picker to remember a PDF.

The three costs of getting it wrong

Chargebacks, which are visible. Retail packaging and labelling violations are among the most common deduction categories, and they are self-perpetuating: the same misconfiguration produces the same deduction on every shipment until someone notices the pattern in the remittance detail.

Rework, which is semi-visible. Product that arrives wrong gets reworked, either by the partner and billed back, or by you on return. Either way you paid for the packaging twice plus the freight on a round trip.

Material waste, which is invisible. When packaging is not modelled, packaging materials are not forecast. Brands over-order the branded boxes and run out of the polybags, or the reverse, and the emergency purchase of the wrong thing at the wrong price never gets attributed to the planning gap that caused it.

Packaging cost per unit, same product, four channels:

  DTC branded outer + tissue + card         = $2.41
  Marketplace polybag + FNSKU label         = $0.38
  Retail case pack (per unit, 12-count)     = $0.29
  Bulk carton (per unit, 48-count)          = $0.11

  Contribution margin difference between
  the DTC and retail configuration:           $2.12/unit

That spread is why the packaging spec has to reach the margin model. A brand costing every unit at the DTC packaging rate understates retail margin badly enough to distort channel strategy, and one costing everything at the bulk rate does the opposite and finds DTC less profitable than the spreadsheet promised.

Getting the spec to the pack station

A packaging record nobody sees at the moment of packing is documentation, not a control. The gap between the two is where the chargebacks happen.

The pattern that works is that the pick or pack document is generated from the destination, not from the product. The packer scans the order, and the screen or slip tells them the configuration for that order specifically — this outer, these labels, this count per case. They are not asked to know which channel this is or to remember what that channel wants. The system already resolved it.

Three properties make that reliable:

The packer never chooses. If the interface presents options, the interface has moved the decision back to memory. Resolve it upstream and present one answer.

Materials are scannable. Confirming the correct outer by scanning it catches the substitution that happens when someone reaches for the wrong box because the right one is out of easy reach. This single control removes a large share of packaging errors in most operations.

Labels are generated, never selected. A GS1-128 built from the order and pallet data cannot carry last week’s numbers. A label picked from a folder of PDFs regularly does.

None of this requires a sophisticated WMS. It requires the packaging specification to be data that the pick document can read, which is the same requirement as everything else in this article.

Onboarding a new retail partner

The moment a new retail partner is signed is the moment the packaging model gets tested, and the timeline is usually compressed because the commercial conversation took longer than expected.

A workable sequence:

1. Extract the packaging and labelling requirements from the routing guide before the first PO. The routing guide is usually a long PDF whose packaging section is a few pages. Those pages are the specification.
2. Model the configuration as a record against the products that partner will buy, including case quantity, inner pack, labelling and pallet configuration.
3. Price it. The case configuration has a per-unit packaging cost that differs from your DTC cost, and it belongs in the margin model before you agree to the price list, not after the first invoice.
4. Order the materials. New case sizes and label stock have lead times that routinely exceed the gap between signing and first order.
5. Run one physical test pallet and check it against the guide before the first real shipment. The first shipment is the most expensive place to discover a misread requirement.

Step three is the one that gets skipped, and it is the one that determines whether the account is profitable. A retail price list agreed against DTC packaging assumptions can be wrong by a point or two of margin before a single unit ships.

Who owns packaging

Part of why this stays messy is that packaging has no natural owner.

Marketing owns the DTC unboxing. Operations owns the pack station. Finance owns the material spend. Sales owns the retail relationship that came with the routing guide. Each owns a slice, none owns the specification, and the specification is the thing that has to be correct.

The brands that handle this well give one person the packaging configuration as an explicit responsibility — usually in operations, sometimes in a supply chain role — with authority over the records and accountability for compliance. That person does not design the unboxing or negotiate the retail terms. They own the answer to “how does this product get packed for this destination”, and they own it in the system rather than in their head.

That is a small organisational change that prevents the failure mode in the opening paragraph, where the answer walked out of the building.

Where this usually breaks in the stack

Packaging requirements tend to live in whichever system was closest to the problem when it first appeared. The retail specs end up in the EDI tool because that is where the routing guide was implemented. The marketplace prep rules end up in the marketplace integration. The DTC rules end up in the WMS or in a pack station SOP. Nothing holds all four, so nothing can answer “how do we pack this for that customer” in one query.

That fragmentation is also why the material forecast is wrong. Demand planning sees units of product. It does not see that 30% of those units need a $2.41 outer and 40% need a $0.29 case, because the channel mix that determines the split lives in a different system from the forecast.

Endless Commerce holds packaging configuration as part of the product and channel model rather than as a per-integration setting, so the retail case pack, the marketplace prep rule and the DTC presentation are attributes of the same record, applied at pick and reflected in both the freight calculation and the material forecast. EndlessEDI reads the retail configuration directly, so the ASN describes the pallet that was actually built.

Where to start

Do not start with the system. Start with the document nobody has written.

For each active channel, write down the packaging specification exactly as your best packer would describe it, and then check it against the partner’s current routing guide. The gaps you find in that exercise are, in most cases, already costing you money in deductions that have been coded as something else in the remittance detail.

Then check who else knows. If the answer for any channel is one person, that channel is one resignation away from a bad quarter, and the fix is not documentation for its own sake — it is getting the specification into a system that shows it to the packer at the moment of packing, whoever is standing there.