Kidswear pack configurations vs single-unit picking: warehouse tradeoffs

Why does the kidswear pack vs single unit warehouse question keep breaking operations?

It is 7:40 AM at a 3PL in Kentucky that ships for a $22M kidswear brand. The wholesale wave printed overnight includes 640 prepacks bound for a regional department store, each pack a 2T through 6T size run in one polybag with one UPC. The DTC wave printed at 6 AM includes 2,900 eaches, most of them the same styles in the same sizes. The problem: the WMS is showing 118 units of style 4471 in size 4T available, and both waves want them. The pack pick has to break down to size, or the eaches have to be cut. Someone on the floor is going to make that call in the next fifteen minutes, and whichever way it goes, a report downstream is going to be wrong.

This is the daily shape of the kidswear pack vs single unit warehouse tradeoff, and it is almost never solved cleanly at the WMS layer. It is solved, or not, in how the brand models its inventory before the pick ticket ever prints.

What is a prepack in kidswear, and why does it exist?

A prepack (sometimes called a size run, assortment, or bulk pack) is a fixed bundle of the same style across multiple sizes, treated as one sellable unit with one SKU and one UPC. In kidswear, the most common shapes are 2T-3T-4T-5T (a four-piece toddler run), 4-5-6-6X (a girls’ run), and 8-10-12-14 (a big kids’ run), usually one of each size, sometimes with doubles on the middle sizes based on sell-through history. The pack ships in a polybag or inner carton and is broken to floor by the retailer.

Prepacks exist because department stores and mid-tier chains do not want to write purchase orders at the size level for 40 styles across 90 doors. They want to buy 12 packs of style 4471, 8 packs of style 4472, and let the pack composition handle the size curve. The warehouse benefits too: one pick, one scan, one carton line on the ASN, one line on the invoice. For a wholesale order of 300 units, that is 75 picks instead of 300 if the pack is a four-piece.

Single-unit picking (eaches) is the DTC and specialty-store shape. The customer bought one 4T dress, the picker finds one 4T dress, it goes in a polybag. There is no bundling, no assortment logic, no size curve baked into the SKU.

Why does this become a warehouse execution problem at BP5?

BP5 in the 6 Breakpoints framework is where warehouse execution gets less predictable, and the 3PL blind spot lives here. Kidswear brands hit BP5 hard because they run both modes in the same building against the same physical inventory. A single 4T dress in a bin might be committed to a wholesale pack, a DTC order, or an Amazon Vendor Central PO, and the commitment logic often lives in three different systems that do not talk to each other in real time.

When I started Uphance, the shape I kept seeing in kidswear specifically was this: the brand’s ERP or spreadsheet thought inventory was fine, the 3PL’s WMS thought inventory was fine, and then the warehouse manager would walk the floor on a Thursday and realize half the 4Ts had been picked into packs on Tuesday for a wholesale wave that had not yet shipped, and the DTC oversells were about to start. The tools were each internally consistent. They were just not consistent with each other, and no one had modeled the pack-to-each relationship as a first-class thing.

For a $15M brand running wholesale plus DTC plus a 3PL, we consistently see 6 to 9 hours a week burned just reconciling inventory across Shopify, the 3PL portal, and the wholesale order book, with a 2 to 3 percent oversell rate at peak. In kidswear with prepacks in the mix, that oversell rate runs higher because one broken pack cascades across four or five size SKUs at once.

How should prepacks be modeled in inventory?

There are three defensible models, and most kidswear brands are running the wrong one for their volume.

Model one: pack as its own SKU with its own physical inventory. The pack is assembled at receipt (or at the factory) and sits on the shelf as a discrete unit. The eaches inside are not available to any other channel because they physically do not exist as eaches. This is clean but expensive: you commit inventory to a channel at the point of assembly, and if wholesale demand softens, you break packs down manually and take a labor hit.

Model two: pack as a virtual SKU backed by component eaches. The pack SKU exists in the item master with a bill of materials pointing to the underlying size SKUs. Physical inventory sits as eaches. When a wholesale order for 100 packs releases, the system reserves 100 of each size and prints a pick ticket that assembles at pack-out. This is the flexible model, and it is the right one for most brands doing under $50M in wholesale, but it only works if your inventory system understands that the same 4T unit cannot be simultaneously available as a pack component and as a DTC each.

Model three: dual-track with channel-aware ATS. Physical inventory is eaches, but the available-to-sell calculation is split by channel with committed pools. Wholesale sees the wholesale-committed pool, DTC sees the DTC-committed pool, and a rebalancing job runs on a cadence (daily is usually enough) to move inventory between pools based on open orders and forecast. This is the correct model above $30M or when wholesale is more than 40 percent of revenue, and it is the model most brands should be moving toward. It requires that your inventory system, order management, and warehouse execution share one truth about what is committed where.

The warehouse execution scorecard walks through the diagnostics that flag which model a brand has actually implemented, which is usually not the model the ops team thinks they are running.

When does pack picking actually pay off versus eaches?

Prepacks pay off when three conditions hold together. First, the retailer requires or prefers pack-level ordering, which most department stores and mid-tier chains do for basics and core styles. Second, the size curve inside the pack matches actual sell-through at the door, so the retailer is not stuck with 5T deadstock while 3Ts sold out in week two. Third, your warehouse can either receive packs pre-assembled or assemble them in a dedicated zone with dedicated labor, not by pulling a picker off the DTC line.

Eaches win when the style is fashion (short lifecycle, uncertain size curve), when the retailer is a specialty boutique that buys at the size level anyway, when DTC is more than 60 percent of the style’s expected sell-through, or when the size run itself does not match the pack shape (a style that only comes in 4T-5T should not be forced into a 2T-6T pack).

The operational anti-pattern I see most often: brands prepack everything because the biggest wholesale account asked for it three years ago, and now 70 percent of the item master is packs, DTC allocation is a nightmare, and the warehouse is breaking packs down manually every week to fulfill Amazon orders that came in at the each level. The rule I would push: prepack the top 20 percent of styles by wholesale volume, keep everything else in eaches, and only build packs for the accounts that actually require them.

What does this cost when it goes wrong?

The direct costs are visible. Broken pack labor runs 3 to 5 minutes per pack at a fully-loaded warehouse labor rate, which on 500 broken packs a week is real money. Chargebacks from department stores when a pack ships short, or when the ASN says 12 packs and the retailer receives 11 plus a loose 4T, run 3 to 8 percent of the invoice line depending on the retailer’s compliance manual. If your retailer chargebacks exceed 1 percent of wholesale revenue on packed styles, the pack model is broken, not the warehouse.

The indirect costs are worse. Oversells on DTC when eaches were quietly consumed by pack assembly force refunds, which for kidswear also often means the customer buys the size elsewhere and does not come back for the next drop. Inventory writedowns at end of season are higher on packs than eaches because you cannot markdown a pack the way you can markdown a size that did not sell. And the reporting layer at BP6 becomes unreliable because sell-through by size is obscured whenever the sale happened at the pack level.

How should the warehouse be physically laid out?

One rule: never mix pack picking and each picking in the same zone with the same labor pool. The motion is different, the scan pattern is different, the QC step is different, and the error rate on both goes up when a picker context-switches. In practice this means either a dedicated pack assembly and pick area (best for brands where packs are more than 30 percent of outbound units) or a batched pack window (packs picked in the morning wave, eaches in the afternoon, no overlap).

The pack zone should hold pre-assembled packs where possible, or component eaches organized by pack composition (all 2T-6T girls’ dresses in one bay, not scattered by style across the whole warehouse). The each zone should be organized for pick path efficiency across DTC orders, which usually means style-then-size, not the pack-composition layout.

For brands on a 3PL, this layout question is often the source of the biggest hidden cost. The 3PL bills per pick, per pack-out, and per SKU location, and a warehouse that has not been laid out for the brand’s actual pack-to-each ratio ends up paying for motion that a 30-minute layout conversation would have eliminated. The multi-warehouse and 3PL operating model diagnosis is usually where this conversation starts, because the same problem replicates across every node once the brand adds a second warehouse.

What has to be true in the inventory system for any of this to work?

Four things, and if any one of them is missing, the model collapses back into spreadsheets and Thursday-afternoon fire drills.

One, the item master must model the pack-to-each relationship as a structural relationship, not a note in a description field. The system has to know that pack SKU 4471-PK-TODD contains one 4471-2T, one 4471-3T, one 4471-4T, and one 4471-5T.

Two, available-to-sell has to be channel-aware. When a wholesale order commits 100 packs, the DTC ATS for the four component sizes has to drop by 100 each in real time, not overnight, not after a batch job.

Three, the warehouse execution layer has to know which mode it is picking in for a given wave, and the pick tickets have to reflect the model (pack pick from a pack location, or component pick with an assembly step at pack-out).

Four, the reporting layer at BP6 has to be able to reconstruct size-level sell-through from pack-level sales, which requires that every pack transaction gets exploded back to its component sizes in the data warehouse. Without this, merchandising is planning next season’s size curve blind.

This is a lot to ask of a stack of point tools stitched together with CSV exports, which is why the brands that solve it usually solve it by connecting warehouse execution to the same inventory model their order management and PLM already use. The tools do not have to be one product, but the model of a pack has to be one model.

The decision that actually gets made on Thursday afternoon

When the warehouse manager walks the floor and sees 118 units of 4T against two waves that both want them, the decision they make in that moment is downstream of decisions the brand made months earlier: which styles to prepack, how to model the pack in inventory, how to lay out the pick zones, and how to calculate channel-aware availability. If those decisions were made well, the Thursday-afternoon question does not come up because the system already knows which units are committed where. If they were made poorly, no amount of WMS configuration will fix it, and the brand will keep burning an FTE on inventory reconciliation and taking chargebacks that never should have shipped.

The kidswear brands that get past this cleanly are the ones that stopped treating prepacks as a warehouse problem and started treating them as an inventory architecture problem. The warehouse just executes the model. The model is where the money is made or lost.