The connected stack by vertical

One attribute decides where a stack breaks

Most stack comparisons across verticals end up saying the same thing twice: the categories are similar, the words are different, buy a platform. That is true and useless, because it does not tell anyone where to look. The useful difference between verticals is not which systems are present — it is which attribute the buy is actually decided on, because that attribute has to survive every handoff between the system that authors it and the system that commits the order, and a stack is only as connected as that one path.

In apparel that attribute is the size curve at style-color grain. In footwear it is the width fitting crossed with the size run. In home and furniture it is the container the goods travel in. In beauty it is the dating on the stock. In toys it is an approval state. In jewelry and watches it is a cost base that moves on a market after the plan is set. Each of these is a dimension the buy is committed against, and each is routinely stored as text — a suffix on a code, a note on a line, a field in a system nobody plans from. Stored as text it still prints correctly on every document. It simply cannot be planned on, summed, or checked.

So each vertical below answers three questions rather than one. What is actually in the stack — not the reference architecture, but the systems a brand in that vertical really has, including the ones that are a portal, a rep’s spreadsheet or a compliance folder. Which handoff actually breaks — one seam, named, rather than a general observation that integration is hard. And what that costs in planning terms — the decision that gets made on a number that was true somewhere else.

Definition — Vertical-critical attribute

A vertical-critical attribute is the one dimension a vertical’s buy is actually committed against, which the generic stack tends to carry as a label rather than as a dimension. It is the size curve in apparel, the size run crossed with the width fitting in footwear, the committed material lot in accessories and bags, container consolidation in home and furniture, the dealer prebook book in outdoor, the model-year changeover in sporting goods, period-after-opening dating in health and beauty, licensor approval and test status in toys and games, regulatory and recall state in baby and juvenile, and the metal cost base and piece location in jewelry and watches. Stored as a code suffix, a note or a field nobody plans from, it prints correctly on every document and cannot be summed, grouped or checked — which is how a business can hold every fact in its stack and still be unable to answer the question it needs answered.

Used by: Merchandising, planning and buying leaders deciding which seam in a stack to instrument first

Related: Connected stack, planning grain, size curve, size run, width fitting, container consolidation, period after opening, model year, prebook

Two cautions before the list. First, a category is a job, not a licence: several of these ship bundled, and a brand with four products still has ten jobs to place. Second, none of what follows is a claim that one vertical is harder than another. They are differently shaped, and the shape decides where to look first. The structural argument that these are one problem rather than ten sits in the pattern beyond apparel; what follows is the operational version of it, vertical by vertical.

Apparel

Apparel buys a style-color against a single size scale, and that is the simplest shape in this list. The size curve is still the attribute the buy turns on. A style-color bought deep on the wrong curve is not a depth problem, it is a shape problem, and it will present itself as a depth problem in every report written afterwards — because a report that aggregates to style-color has already summed away the dimension the error lives in.

The stack has somewhere to put the curve. PLM holds the size scale as an attribute of the style, ERP holds a size-level quantity on the order line, and the allocation system holds a curve per door cluster. What no system holds is the relationship between them: that the curve the buy assumed and the curve the allocation applies are meant to be the same object, and that when they diverge, it is the buy that was wrong rather than the doors. Three systems each hold a size curve and none of them holds the same one, which is how a brand can have complete size data and no size answer.

Delivery phasing runs the same way. The plan phases receipts by month; the order carries an ex-factory date and an in-DC date; the factory moves the ex-factory date by email. Wholesale adds a second clock, because a retail partner’s delivery window is contractual and a chargeback follows a miss. The apparel treatment on the flagship is apparel brands, and the seam-by-seam version of this stack is the apparel systems landscape.

Systems actually in the stack

PLM for the style, colorway, bill of material and tech pack; PIM and DAM for the consumer-facing product record; ERP for the committed purchase order and the inventory value; a merchandise planning system for open-to-buy; allocation and replenishment; WMS; OMS; EDI for wholesale partners; a storefront; a reporting layer. Between them, a spreadsheet holding the buy at the grain the plan cannot reach.

The handoff that actually breaks

The buy into allocation. The size curve used to size the order and the size curve used to spread it across doors are two separate artifacts, built by two functions, reconciled by neither.

Planning consequence

Runs break in the middle sizes first. The sizes that remain on the floor then read as slow demand rather than as the tail of a broken run, and that read is what sizes the next buy.

Footwear

Footwear buys pairs against a matrix, not units against a curve. A style-color is committed across a size run crossed with a width fitting, and both dimensions have to hold for the buy to be sellable — a full run in the wrong fitting sells no better than a broken run in the right one. That second dimension is the whole difference between the footwear stack and the apparel stack, and it is the dimension most likely to be absent from the plan.

The reason is structural rather than anyone’s fault. A product system organised around a garment models one size scale per style, because a garment has one. Asked to carry a width, it offers a text field, a code suffix or a separate style record — all three of which print correctly on a purchase order and none of which the plan can roll up by. The width exists everywhere as a label and nowhere as a dimension, so the question “how did the D fitting perform against the EE” can only be answered by a person who knows the naming convention and is willing to rebuild it in a file.

The model-year calendar adds the second complication. Footwear changes a style when the model changes, not when the season turns, so carryover is a plan decision and a product decision at the same time: a carryover style returning with a revised outsole is a new record to the development system and the same line to the plan. Where prebooks commit pairs before the season is readable, the buy locks early and the correction window closes with it. The footwear treatment sits on footwear brands; the planning argument for a committed-once buy is in when you cannot chase.

Systems actually in the stack

The apparel stack plus a development record that has to carry the last, the tooling and the outsole spec, and a model-year calendar deciding when a style changes rather than when a season does. Width is frequently carried as a code suffix or as a second style rather than as a dimension.

The handoff that actually breaks

PLM into planning. The size run crossed with the width fitting is a matrix; a product system built around a garment’s single size scale has a place for the run and no place for the width, so the second dimension survives as naming convention.

Planning consequence

Run integrity gets measured per code instead of per run and fitting. A run short in one width reads as two partly healthy codes rather than one broken run, so the shortfall never reaches the next prebook.

Accessories and bags

Accessories and bags remove a dimension and add a constraint. There is no size curve at all on bags and small leather goods, so the entire bet moves onto colorway, material and hardware. That sounds like a simplification and is not one, because the dimension that replaces size is committed earlier and is far less reversible.

Leather is bought by the hide, hardware and trim in lots, and both are typically committed ahead of the finished-goods order to protect lead time and price. The finished-goods plan therefore nets against a commitment the brand has already made and rarely holds: the material sits on a sourcing spreadsheet, a vendor’s system, or an email confirming a lot reservation. The plan can see the finished-goods buy and cannot see what constrains it, which is exactly backwards for the decision being made.

The commercial pattern underneath makes this sharper. One structure carries two rhythms: fashion colorways carry the seasonal bet while hero colors and the evergreen core replenish, and an attach rate read in season changes how deep the core needs to be. Moving core depth is the right commercial call and can be the one call the committed material does not allow. The accessories treatment sits on accessories brands.

Systems actually in the stack

PLM carrying a bill of material heavy in leather, hardware and trim; a component or raw-material commitment that frequently lives in a sourcing spreadsheet or at the vendor rather than in the brand’s own ERP; the rest of the apparel stack with the size dimension removed.

The handoff that actually breaks

Sourcing into planning. Materials are committed in lots ahead of the finished-goods buy, and the constraint that commitment creates is not held anywhere the plan can read.

Planning consequence

A colorway or material mix change is agreed in season that the committed material cannot support, and the discovery happens at the vendor rather than in the plan.

Home and furniture

Home and furniture replaces the size dimension with an option structure — fabric, finish and configuration against one item — and replaces the season clock with an ocean clock. Lead times stretch across months, orders are placed against container minimums rather than against demand alone, and the goods are received as containers rather than as order lines. That last point is the whole seam.

An ERP purchase order carries a line, a quantity and a date, and it carries them honestly. What it has no way to express is that eleven of its lines and four of another order’s lines will travel in one container, that the container will be consolidated by a forwarder after the order is confirmed, and that a consolidation decision changes which options arrive together. The order’s dates stay exactly as written while the goods behind them move, so there is no discrepancy for anyone to notice — which is why this seam is quiet rather than loud.

Two consequences follow. Floor sets and dealer prebooks are calendar commitments made to third parties: a showroom floor set and a dealer’s delivery window are dates the brand has promised, and half a floor set is not half a result. And special orders sit beside stocked items in the same system while belonging to a different economics, so a plan that does not separate them lets a customer-specific special order consume stocked open-to-buy and understate what the floor still needs. The home and furniture treatment sits on home and furniture brands.

Systems actually in the stack

An item master carrying option structure — fabric, finish, configuration — against a landed cost; ERP purchase orders with per-line dates; a freight forwarder’s system holding the container plan; a floor-set calendar living in a merchandising document; special-order flow running beside stocked flow.

The handoff that actually breaks

The purchase order into receipt. A purchase order line expresses an item, a quantity and a date; the real unit of receipt is a container, consolidated after the order was placed.

Planning consequence

A floor set built against per-line dates arrives in two halves, and the plan holds no record that the change happened — every per-line date is still exactly as written.

Outdoor

Outdoor runs two kinds of product through one line plan. Soft goods plan like apparel, by style-color and size; hard goods plan by model and specification on a model year, changing when the engineering changes rather than when the season turns. One plan has to hold both, and most systems in the stack are built to hold one or the other well.

The dealer prebook is what makes the timing unforgiving. Units are committed by dealers ahead of the season, through a B2B portal for some accounts and as spreadsheets from reps for others, and that book is the demand signal the buy is placed against. The order book therefore exists before the merchandise plan does, and converting a book of prebooks into a plan — by model, by delivery, by channel, net of the direct business — is a conversion performed in a file by whoever is fastest with it.

Two further constraints sit on the same plan. Counter-seasonal categories keep two seasons open at once — snow committed while summer is selling — so a single open-to-buy view has to carry two cycles without netting them against each other. And minimum advertised price policy bounds how far a markdown decision can move price in the dealer channel, which means the in-season lever the direct business would reach for is partly unavailable. The outdoor treatment sits on outdoor brands.

Systems actually in the stack

A mixed development record — soft goods by style-color and size, hard goods by model and spec on a model-year calendar; a B2B portal or spreadsheet book holding dealer prebooks; technical-material lead times tracked in sourcing; MAP policy held in a commercial agreement rather than in a system.

The handoff that actually breaks

The prebook book into planning. The order book is committed through a dealer channel before the merchandise plan is built, so turning a book of prebooks into a plan is a manual conversion.

Planning consequence

The buy is fixed while the season is still unreadable, and the correction window closes before the first read. Markdown response is then bounded by MAP in the dealer channel.

Sporting goods

Sporting goods runs the model-year cycle of outdoor and adds two things to it. Seasonality is by sport rather than by calendar season, so a single plan carries several demand shapes that peak in different months and do not share a markdown moment. And team and league programmes place orders on their own windows, sized to a roster rather than to a forecast, arriving as spreadsheets from reps and as commitments a brand cannot politely re-phase.

The seam is at the purchase order layer, at model-year changeover. The date the new model year lands is the date prior-year stock stops being carryover and becomes closeout, and that is a planning decision disguised as a product date. Ahead of it, remaining stock is inventory with a future; behind it, the same units are a liability competing with the model that replaced them. The date moves — a spec refresh slips, a component is re-sourced — and the plan is typically the last artifact to hear.

The compounding problem is that every amendment has to do two things at once. A revised team order or a refreshed spec changes a quantity the plan has already netted into open-to-buy and changes when receipts land. Handling only the quantity leaves the receipt phasing stale, and handling only the date leaves the envelope wrong — and because each is individually defensible, the pair is rarely caught in the same review. The sporting goods treatment sits on sporting goods brands.

Systems actually in the stack

The outdoor stack plus team and league programmes arriving as rep spreadsheets on their own delivery windows, and a model-year changeover calendar that usually lives in a product document rather than in the plan.

The handoff that actually breaks

The purchase order layer at model-year changeover. Spec refreshes, team-order revisions and the timing of the cut all amend orders the plan has already netted, and each amendment has to re-phase receipts in the same change.

Planning consequence

The changeover date decides whether prior-year stock is carryover or closeout. When the plan does not hold that date, the decision is made late and the stock is marked rather than transitioned.

Health and beauty

Beauty plans shades within a franchise the way apparel plans sizes within a style — the shade is the dimension the buy is decided on, and a shade range with a gap performs like a size run with a hole in it. Around that sit testers, gift-with-purchase units and launch quantities, all drawing on the same budget without selling in the same way, which means a single depth number covers three different economics.

What makes the seam distinct is dating. Period-after-opening and shelf life put a date on depth, so the same quantity is a sensible buy or a write-off depending on when it will actually sell. That dating lives where lots and batches live — in the warehouse and transaction systems — and the plan is held at shade and period grain with no date attached. Both sides are internally correct; the buy is still sized without the constraint that decides whether it can be recovered.

Two more pressures land on the same stack. A gondola reset is a retailer’s calendar, not the brand’s: the reset date fixes when a launch can reach the shelf and when the outgoing assortment comes off it, so launch quantities are sized to a window someone else controls. And reformulation forces a product identity question in the consumer record — the reformulated product has to be distinguishable from its predecessor in the catalog, or the sell-through read spans two different products under one name. The health and beauty treatment sits on health and beauty brands.

Systems actually in the stack

A formulation and regulatory record outside PLM; lot and batch tracking in ERP and WMS; a shade range and franchise structure living in the commerce catalog and PIM; retailer POS as the in-season signal where the brand sells wholesale; gondola reset dates set by the retailer.

The handoff that actually breaks

Lot and dating data in the warehouse and transaction systems into the plan. Depth has an expiry on it; the planning grain is dateless.

Planning consequence

A buy too deep for its dating can end as a write-off rather than a markdown, and where a reformulated product is not given its own identity in the consumer record, the sell-through read spans two different products under one name.

Toys and games

Toys and games plan the item against a retailer commitment, and the calendar is dominated by a gifting peak that cannot be moved. Everything upstream is measured against that fixed point, which changes the character of every delay: a week lost anywhere before production is a week taken out of manufacturing, not a week taken out of selling.

Two gates sit in front of the buy, and neither lives in a commercial system. A licensed property can only be sold inside its licensed window and only in the form the licensor has approved — artwork, packaging, territory and term — and that approval state typically lives in an email thread or a licensing tool the plan has no relationship with. Then safety testing and certification gate shipment: an item that has not passed applicable testing does not ship whatever the order says, and the standards involved are public and checkable — CPSIA and ASTM F963 in the United States, EN 71 in the European Union.

The result is a plan that has to carry three states on one item at once: the retailer commitment, the order placed against it, and whether the item is approved and tested. Retailer commitments themselves move — revised quantities and revised dates against a peak that does not move — so receipts have to be re-planned before the retailer’s window closes, and that re-plan is only possible if all three states are visible together. The toys and games treatment sits on toys and games brands.

Systems actually in the stack

An item-level product record; licensor approval state held in email or a licensing tool; safety testing and certification in a compliance system or a folder of reports; retailer commitments arriving by EDI and as buyer spreadsheets; a gifting peak that fixes the whole calendar.

The handoff that actually breaks

Approval and certification into the buy. An item cannot be committed until the licensor approves it and cannot ship until it passes testing, and the plan holds neither state.

Planning consequence

Time lost waiting on an approval comes out of production rather than out of the selling window, because the gifting peak cannot move to accommodate it.

Baby and juvenile

Baby and juvenile plans hardgoods models the way outdoor plans hard goods — by model and model year rather than by season — with soft goods on an age-based size scale beside them, and a stage progression running underneath both, because a product is bought for a stage the child is about to enter rather than the one they are in. Registry demand makes some of that visible before the purchase happens, which is a genuinely useful early signal and one that lives in a channel system rather than in the plan.

The defining feature of this stack is that regulation can stop a model mid-season. Standards apply directly to the hardgoods the plan is built on — FMVSS 213 in the United States and ECE R129 in Europe are the public examples for child restraints — and a standards change or a recall is not a markdown decision or a phasing decision. It is a stop condition, and it has to reach every open order, every in-transit unit and every allocation simultaneously, because partial compliance is not a state that exists.

Certification, lot and serial traceability correctly live in the ERP and quality systems, and should stay there. What the plan needs is not the traceability data but the state it implies — whether this model is shippable right now, and what is committed against it. Alongside that runs the ordinary model-year problem: the new model’s landing date decides how much outgoing stock the plan still needs, and moving that date without re-planning the outgoing model is how a transition becomes a clearance. The baby and juvenile treatment sits on baby and juvenile brands.

Systems actually in the stack

Hardgoods models on a model-year calendar — car seats, strollers, nursery furniture — beside soft goods on an age-based size scale; certification, lot and serial traceability held in ERP and quality systems; registry demand visible in a retailer or direct channel ahead of purchase.

The handoff that actually breaks

Regulatory and quality state into the whole plan at once. A recall or a standards change is a stop condition that must reach open orders, in-transit units and allocation in the same moment.

Planning consequence

The model-year changeover and the stop condition both act on the plan, and a plan holding neither learns about each from the function that happened to notice.

Jewelry and watches

Jewelry and watches break the usual assumption that cost is a number set once. Gold and silver trade on a daily market, so a precious metal cost base moves after the plan is built and keeps moving while orders are open. The plan is not wrong when it is written; it stops being right while nothing about it changes, which is a failure mode no reconciliation catches because there is no discrepancy to find — only two correct numbers from two moments.

The second structural feature is where the inventory physically is. Memo and consignment place stock with retailers who do not own it, which means the brand’s position includes units it cannot sell from where they are sitting and cannot count as a retail sale until they sell through. Where velocity is low and value is high, this is not a rounding issue: one piece in the wrong store is a material share of a reference’s position, and a position read without the memo split reads as depth that does not exist.

Underneath both sits the unit of inventory itself. In higher-value categories inventory is tracked piece by piece rather than as a quantity, with a serial number and, in markets that mandate assay, hallmarking on the piece. That is the right control and it changes what planning means: a plan that sizes a buy in units is planning an aggregate that the business does not actually hold. Model year and reference changes then add the sporting-goods problem on top — the changeover decides when a reference becomes prior season. The jewelry and watches treatment sits on jewelry and watch brands.

Systems actually in the stack

A product record carrying metal, stone and reference specification; piece-level serialized inventory in higher-value categories; a standard cost in ERP set at a point in time against a metal cost that moves on a daily market; memo and consignment placing stock in locations the brand does not own; hallmarking and assay requirements in the markets that mandate them.

The handoff that actually breaks

Cost into the margin plan, and location into the inventory position. A metal move re-prices open orders after the plan is set; memo and consignment placements change the position without a sale.

Planning consequence

Planned margin is computed against a cost base that has since moved, and the position includes pieces that cannot be sold from where they are sitting.

The attribute, where it lives, and the seam that breaks

Read down the middle column and the pattern is hard to miss. In nine of the ten verticals, the attribute the buy turns on is authored outside the planning stack altogether — a development record, a dealer’s order book, a forwarder, a production line, a licensor, a market. Apparel is the single exception, and it fails differently for it: the size curve is a planning artifact, so the apparel problem is not an absent author but three systems each holding their own copy.

Four things that are true in every vertical

The attribute the buy turns on is authored outside the system that commits the buy. A width fitting is decided in development; a container is consolidated by a forwarder; dating is created on a production line; an approval is granted by a licensor; a metal cost is set by a market. None of those are planning decisions, and all of them decide whether a plan is executable. That separation is not a defect in any one product — it is what happens when a commercial decision depends on an operational fact.

The attribute is carried as a label rather than as a dimension. Stored as a suffix, a note or a field in a system nobody plans from, it prints correctly on every document and cannot be summed, grouped or checked. This is the specific reason a brand can have every fact in its stack and still be unable to answer the question it needs answered — the data is present, and the shape is missing.

The failure is silent while it happens. Each system stays internally consistent throughout: the order’s dates are as written, the plan’s numbers add up, the report reconciles to its source. Nothing throws an error, because nothing is in error. Two correct records simply describe different worlds, and the disagreement only surfaces when goods arrive, a shade expires or a piece cannot be sold from where it is.

The read-back inherits the error. The most expensive consequence is never the season in which the break happens. It is the next one, because the sell-through read produced by a broken run, a half-arrived floor set, an expired shade or a stalled licensed item looks exactly like a demand signal — and demand signals size the next buy. A single structural break therefore prices itself into two seasons, and the second is the one nobody attributes to it.

The mechanism behind all four is worked through in the handoff problem, and the object that removes them — one versioned set of numbers every function reads and changes through one path — is the plan of record.

Five questions that locate the seam

None of this requires a stack audit to act on. The diagnosis is five questions, and it is usually finished inside an hour — because the people who can answer them already know the answers and have simply never been asked in this order.

One: what is the buy actually decided on?

Not what the plan is expressed in — what the last argument before commitment was about. If the recurring argument is the size run, the container, the dating or the approval, that is the attribute. The tell is a decision that gets remade every season by the same two functions and is never recorded anywhere the following season can find it.

Two: which system authors it?

Authors, not stores. Copies are fine and unavoidable; what matters is where the value is first written and by whom. If two systems both write it, there is no system of record for it, only two defensible numbers — and the object-by-object version of that assignment is the system of record map.

Three: how does it reach the plan?

Trace the actual path: an export, an email, a portal download, a weekly file, a person who knows. Count the hands rather than the systems — a two-system path with three people in it is a three-handoff path. Then ask what happens to that path when the person who knows is on leave, because a path that depends on a person is a path with an availability condition on it.

Four: can the plan group by it?

This is the label-versus-dimension test, and it is the fastest one. Ask for last season’s performance grouped by the attribute — by width fitting, by container, by dating window, by licensed property, by metal. If the answer takes a person and a spreadsheet, the attribute is a label, and every decision made on it this season is being made on a rebuilt view whose assumptions nobody reviewed.

Five: what does a change to it touch?

Take one real change from last season — a consolidation, a slipped approval, a metal move, a standards change — and list everything that should have moved with it. The gap between that list and what actually moved is the cost of the seam, stated in the business’s own terms rather than in an industry figure. That number is worth more than any benchmark, because it is yours and nobody can argue about the denominator.

What to do with the answers depends on where the gap is, and the sequencing — one seam at a time, with an owner per phase — is laid out in the implementation playbook. If the question is which category of software to evaluate at all, the vendor-neutral question set is in the apparel software buyer’s guide, and the structured version of the diagnosis above is the operating model assessment.

Several of these seams have arithmetic you can work before connecting anything — open-to-buy, size curves, sell-through and markdown. The free retail-plan.com calculators let you model one seam in isolation first.

What naming the seam does not do

Locating the seam is not the same as closing it, and it is worth being plain about the difference. A handoff disappears only when both sides of it read the same versioned record — which is sequencing work, decision-rights work and data work, in roughly that order, rather than a purchase. Several of the attributes above will continue to be authored outside the commercial stack no matter what is bought, because a forwarder will still consolidate the container and a market will still move the metal.

What is achievable is narrower and more useful: carrying the attribute as a dimension the plan can group by, and making the change to it visible in the same place the decision is made. That does not stop a container being re-consolidated. It stops the floor set being planned as though it had not been. The accountability half of that — who is allowed to change what, and on which clock — is in the decision rights map and the operating cadence, and the level each decision belongs at is in the planning grain.

One last caution about reading across verticals at all. The ten stacks are genuinely similar and the temptation is to conclude that a practice which works in one transfers to another. The categories transfer; the attribute does not. A size-curve discipline built for apparel does not become a width discipline for footwear by renaming a column, and a phasing discipline built for a monthly receipt plan does not become a container discipline by changing the date. Borrow the question, not the answer.

Frequently asked questions

Does the systems stack really differ by vertical, or only the vocabulary?

The software categories are the same in all ten: a product system, a consumer-facing product record, a transaction system, a planning system, allocation, a warehouse system, an order system, a wholesale exchange, a storefront and a reporting layer. What differs is the attribute each vertical’s buy is actually decided on — width fitting in footwear, container consolidation in home and furniture, period-after-opening dating in beauty, licensor approval in toys, metal cost in jewelry — and whether the stack carries that attribute as a dimension or as a label in a text field. A stack that carries it as a label works until someone has to plan on it.

Why does footwear break where apparel does not?

Because footwear buys a matrix rather than a curve. An apparel buy commits units against one size scale per style-color; a footwear buy commits pairs against a size run crossed with a width fitting. A product system built around a garment’s single size scale has somewhere to put the run and nowhere to put the width, so width is carried as a suffix on the code or as a separate style. The plan can then count pairs, but it cannot roll up by width, and run integrity gets measured per code instead of per run — which is why a run short in one fitting reads as two partly healthy codes instead of one broken run.

What is the seam that breaks first in home and furniture?

The one between the purchase order and the receipt. A purchase order line carries an item, a quantity and a date; the real unit of receipt is a container, and a container is consolidated by the freight forwarder after the order is placed. When a consolidation changes which options travel together, the order’s per-line dates stay exactly as they were written while the goods behind them move. The floor-set calendar was built against those per-line dates, so half a floor set arrives on time and the other half does not, and nothing in the stack said so in advance.

What do beauty, toys and baby and juvenile have in common?

In each, a condition outside the commercial systems can stop a buy or write it off. In beauty it is dating: period-after-opening and shelf life put an expiry on depth, so a buy too deep for its dating can end as a write-off rather than a markdown. In toys it is approval and certification: a licensor approval and a safety test — CPSIA and ASTM F963 in the United States, EN 71 in the European Union — both gate whether an item may ship at all. In baby and juvenile it is regulation and recall: standards such as FMVSS 213 in the United States and ECE R129 in Europe apply to the hardgoods models the plan is built on, and a recall is a stop condition that has to reach open orders, in-transit units and allocation at the same moment. None of those conditions live in the planning system, and all three decide whether the plan is real.

Which vertical has the shortest window to correct a buy?

The prebook verticals — footwear, outdoor and sporting goods — alongside toys. Footwear, outdoor and sporting goods commit through prebooks that close before the season opens, so the order book is fixed while the season is still unreadable, and the model-year changeover then decides whether prior stock is carryover or closeout. Toys concentrate their selling into a gifting peak that cannot move, so every week spent waiting on a licensor approval or a safety certificate comes out of production rather than out of the selling window. In all four, the useful question is not how to react faster in season but what the plan has to be right about before the window closes.

Where should a brand in a non-apparel vertical start?

Name the attribute the buy turns on, then find the system that holds it and the system that decides the buy. If those are two different systems with a person between them, that gap is the first seam worth instrumenting, whatever else is on the roadmap. It is a cheaper diagnosis than a stack audit and it is usually the one that explains the recurring argument — the size run that never fills, the floor set that lands in two halves, the shade that expires, the piece nobody can sell because it is on memo somewhere else.

• ✓Every merchandise vertical runs the same software categories — product, consumer-facing product record, transaction, planning, allocation, warehouse, order, wholesale exchange, storefront and reporting. What differs is the attribute the buy is actually decided on.
• ✓Apparel turns on the size curve; footwear on the size run crossed with the width fitting; accessories on committed material lots; home and furniture on container consolidation; outdoor and sporting goods on the prebook book and the model-year changeover.
• ✓Health and beauty turns on dating — period after opening puts an expiry on depth. Toys turns on licensor approval and safety certification. Baby and juvenile turns on regulatory and recall state. Jewelry and watches turns on a metal cost base that moves on a daily market.
• ✓In nine of the ten, the attribute is authored outside the planning stack, and is carried as a label — a code suffix, a note, a field nobody plans from — rather than as a dimension the plan can group by.
• ✓The failure is silent while it happens: every system stays internally consistent, so there is no discrepancy to find until goods arrive, a shade expires or a piece cannot be sold from where it is sitting.
• ✓The expensive consequence is the following season, because a broken run, a half-arrived floor set or a stalled licensed item produces a sell-through read that looks exactly like a demand signal — and demand signals size the next buy.
• ✓Five questions locate the seam without a stack audit: what the buy is decided on, which system authors it, how it reaches the plan, whether the plan can group by it, and what one real change to it should have touched.
• ✓Naming the seam does not close it. Carrying the attribute as a dimension, and making changes to it visible where the decision is made, is the achievable version — and it is sequencing work rather than a purchase.

• Read the ten software categories one level down, inside apparel →
• See which system owns each apparel data object — the system of record map →
• Read what the plan of record holds and who holds the pen →
• Read why workflows break at handoffs rather than inside functions →
• See which decision belongs at which level — the planning grain →
• Read why the disconnected stack is structural, not an apparel quirk →
• Sequence a connected implementation one seam at a time →
• Score your own operating model with the assessment →