Merchandise Hierarchy by Vertical: The Planning Levels

What a merchandise hierarchy is

It is not the navigation tree on the website, and it is not the folder structure in the product development system. The merchandise hierarchy is the coordinate system of the planning process. A figure in a merchandise plan carries no meaning until you know which level it belongs to: the same number is a strategic bet at division level, a purchase commitment at option level, and statistical noise at unit level.

Every merchandising organization has one. Not every merchandising organization has chosen one. A hierarchy inherited from an ERP implementation, extended twice by people who needed a new grouping for a report and never revisited, is how a brand ends up planning at a level nobody buys at, and reconciling by hand every month.

The level you plan at decides what you can see and what you can do

Two constraints fall out of the level you choose, and they pull in opposite directions.

The first is visibility. You cannot see a variance below the level you planned at. If the plan was written at department level, a department that finishes the season down against plan tells you the department missed. It does not tell you whether three classes missed badly and the rest held, or whether every class drifted slightly. The plan level is the finest resolution at which the word "variance" means anything, because a variance is a difference against an intent, and no intent was ever recorded below that level.

The second is actionability. You cannot act above the level you buy at. A department-level miss has no corresponding action, because nobody writes a purchase order for a department. Somebody has to translate the miss down into a set of styles, colors, models or items to add, cut or reflow — and if that translation happens in a side conversation rather than in the plan, the plan is a scoreboard rather than a control system.

The level that satisfies both is the level where the decision actually lives. That is why hierarchy design is a planning question and not a data-modeling question.

The reporting hierarchy and the planning hierarchy are different objects

These are routinely conflated, because both are drawn as trees hanging off the same product master, and because the reporting hierarchy is the one the ERP ships with. Where each tree sits inside the wider schema of products, locations and plans is the subject of the retail data model; the definitional split itself is covered in the merchandise hierarchy glossary entry.

A reporting hierarchy aggregates facts upward. It can be as deep and as branched as the source data supports, because every level is populated by actuals: units sold, receipts landed, margin realized. Adding a level costs nothing but a column. Attributes that have nothing to do with the buy — fabric content, country of origin, sustainability flag, price band — belong here, and reporting is better for having them.

A planning hierarchy pushes intent downward. Every level in it is a level at which a human will commit to a number that does not exist yet. Adding a level costs real work every cycle, and adds a place where the plan can fail to reconcile. Attributes that nobody plans at do not belong in it, however useful they are in a report.

The practical test is direction of travel. If a level only ever answers "what happened", it is a reporting level. If someone has to fill it in before the season starts, it is a planning level. Treating the deepest reporting level as the planning level produces the same symptom everywhere: a plan with tens of thousands of cells, most of them zero, maintained by nobody and trusted by no one.

Why the lowest level is not always the right level

There is a persistent intuition that finer planning is more accurate planning. It is wrong, and the reason is statistical rather than philosophical.

A forecast needs history with enough volume that the signal exceeds the noise. Aggregate demand at class level is relatively stable — the class has many items, and their individual variations partially offset. Push the same forecast down to unit level and each cell holds a small share of that volume, with none of the offsetting. The cell's week-to-week variation is now dominated by randomness: one wholesale order, one influencer post, one store's replenishment run. Forecasting each cell independently produces a set of numbers that sum to something worse than the aggregate you started with.

This is why mature planning processes plan at an aggregate level and distribute to the finer ones. Apparel plans a style-color and applies a size curve; it does not forecast each size independently. Footwear plans a model-color and applies a run; beauty plans a franchise and distributes across the shade ladder. The distribution step is a rule, not a forecast, and it is far more stable than forecasting the same cells directly.

The lowest level still matters — it is what ships, what gets counted, and what the warehouse knows about. It is simply not where judgement is added.

What breaks when the number of levels is wrong

The failure modes are asymmetric, so it is worth naming both.

Too few levels: the variance is not actionable. A hierarchy of division, department and SKU has a gap in the middle. Everything visible is too coarse to buy against; everything buyable is too fine to plan against, and the breadth versus depth trade has no level to be argued at. Merchants fill the gap informally, with private groupings held in spreadsheets, and those groupings differ between people. The business ends up with several unofficial hierarchies and one official one that nobody uses for decisions.

Too many levels: every cell is too thin. Consider an illustrative case — the numbers below are invented purely to show the arithmetic. A department carries twelve classes. Each class averages eight options, and each option resolves to seven sizes. Planning every intersection is 12 × 8 × 7 = 672 cells for one department in one period. Fill that grid from a season of history and most cells hold a handful of units. They cannot be forecast, they cannot be reviewed in the time available, and the reconciliation between the grid and the department total becomes a weekly chore. The plan is more detailed and less accurate at the same time.

Between the two failures sits a simple design rule: plan at the coarsest level from which a decision can still be executed without re-interpretation, and treat everything finer as distribution.

The invariant: four slots, two axes, ten sets of contents

Across every category, four functions have to be occupied by some level, and two axes sit between the decision level and the fulfillment level — occupied in some categories and genuinely empty in others. The names change entirely. The functions do not.

The four slots are always occupied by something. The two axes are not, and the categories where an axis is empty are the ones a borrowed hierarchy damages most, because the borrowed structure arrives carrying a level with nothing in it.

Two of the four slots frequently collapse into one another, and that is fine when it is deliberate. In a brand where the decision level and the fulfillment level are the same object — a single-configuration hard good, for example — no translation step is needed, and the plan executes directly. In apparel they are never the same, which is why the size curve exists.

What follows is the same four slots and two axes, filled in ten times. The vocabulary in each section is the vocabulary of that category. It is worth saying explicitly that the apparel words do not travel: pushing size curves onto handbags or seasons onto furniture creates empty structure that degrades everything downstream. Each of these categories has a dedicated view on the industries we plan for.

Apparel

Apparel runs the deepest of the ten, and its vocabulary is the one most planning literature is written in — which is why it travels badly onto categories that do not share its axes. The levels in order: division, department, class, subclass, style, style-color, style-color-size.

The decision level is the style-color — the option. A merchant does not buy a style; they buy a style in a specific color, and the same style in two colors can sell through at completely different rates. This is the style-color matrix. Color occupies the variant axis. Size occupies the size axis, and it is a true multiplicative axis: every style-color resolves across a size run, so the fulfillment level is the style-color-size.

The lifecycle unit is the season — spring/summer and fall/winter, subdivided into deliveries or drops, with a carryover core in some brands. The depth axis is constrained by the size curve: a buy of a given unit quantity is not a free number, it is a quantity distributed across sizes, and getting the curve wrong strands units in the tails while the middle sells out. Depth and breadth trade against each other within a fixed option count, which is the central tension of assortment planning. The size curve template is a working version of that distribution step.

A recurring failure mode in apparel: buying at style-color and planning at style. It happens because the line plan is naturally written by style, and color decisions arrive later. The consequence is that every color-level variance is a surprise — the style hit its number while half its colors missed badly — and there is no plan line to attach the correction to. The related error is treating size as a planning level rather than a distribution level, which multiplies the grid by the size run for no forecasting benefit. See apparel brands and line planning vs assortment planning.

Footwear

Footwear looks like apparel and is not. The levels: division, category (performance, lifestyle, work), silhouette or model, model-color, size run, pair.

The decision level is the model-color, bought as a run rather than as a unit count. Colorway occupies the variant axis, and the word is native here. The size axis is occupied by the size run, and it carries a second dimension wherever a brand fits properly: length and width multiply, which is why a footwear grid gets large fast and why brands that carry widths cannot flatten them into an average without losing the fit proposition they sell on.

The lifecycle unit is a seasonal drop against a carryover core, and in wholesale the prebook is the commitment that matters. Depth is constrained by run integrity: pairs are bought against a run curve, and a broken run — the middle sizes gone, the ends intact — is functionally dead inventory long before the units say so. That makes the useful measure of a footwear position "complete runs remaining", not units.

Where this structure breaks in footwear: planning depth in units instead of pairs against a run, so the plan reports healthy coverage while the sellable position has collapsed in the middle of the curve. The second is collapsing width out of the hierarchy to keep the grid manageable, which works until returns and fit complaints start arriving at a level the plan cannot see. See footwear brands and assortment planning for footwear brands.

Accessories & bags

The levels: division, category (handbags, small leather goods, belts, hats, scarves), family or collection, style, style-colorway.

The decision level is the style-colorway. Colorway occupies the variant axis, and the structure that matters is the split between hero colors — the few that carry the season's newness — and the evergreen core that runs continuously.

The size axis is largely empty here, and this should be stated plainly rather than papered over. A handbag has one size. Belts have waist sizes, hats sometimes have a shallow size run or a fitted option, gloves have a size run — outside those, there is no size axis to plan across. A hierarchy that carries an empty size level because apparel had one makes every report sparser and every cell thinner for no benefit.

Because there is no size axis, depth concentrates: a single style-colorway can absorb the buy that apparel would have spread across seven sizes. The constraint on depth is upstream instead — minimum order quantities, material commitments and hide or fabric yield, which set the smallest sensible buy per colorway and effectively cap how many colorways a season can carry.

The lifecycle unit is a season for the fashion colors and continuity for the core, which means two rhythms in one hierarchy. Where this structure breaks in accessories: planning the evergreen core on a seasonal cadence because everything else is planned that way. The core has no season, needs a replenishment logic rather than a seasonal buy, and carries no markdown exposure from newness, because it never exits. See accessories brands and planning accessories lines.

Home & furniture

The levels: division, category (upholstery, case goods, textiles, decor), collection, model or frame, configuration, finish or covering.

The decision level is a model in a specific configuration and finish. Finish and fabric occupy the variant axis — a frame in oak and the same frame in walnut are different objects with different costs and different sell-through. Configuration plays a size-like role: sofa, loveseat, sectional component; queen, king. It behaves less like a size run than like a set of related products, because customers choose one configuration rather than a brand covering all of them for one customer.

The lifecycle unit is the model year, introduced at market, not a season. Depth is constrained by container quantity and landed cost: receipts arrive in containers, cubic volume is the real limit, and a plan expressed in units without a cube view will not survive contact with logistics.

A recurring failure mode in home and furniture: treating every finish and covering as a stocked variant of the model. Some are, and some are cut-to-order — made after the customer commits. Those two things sit at the same level in the product tree and behave completely differently in a plan. Stocked variants consume inventory; made-to-order variants consume capacity and lead time. A hierarchy that cannot distinguish them will over-plan inventory on the made-to-order side and under-plan the production window it actually needed. See home and furniture brands and merchandise planning for home and furniture brands.

Outdoor

The levels: division, category (packs, shelter, sleep, insulation, footwear), family, model, model-color, and a specification axis where one applies.

The decision level is the model-color for anything with a color choice and the model itself for hardgoods that ship in one colorway. That split inside a single hierarchy is characteristic of outdoor and is the thing generic structures handle badly.

The size axis is occupied by a specification rather than a garment size, and the specification differs by family: torso length on packs, capacity in liters, sleeper count on tents, temperature rating and length on bags, and conventional apparel sizing on the soft-goods side. These are not interchangeable, and a single shared "size" attribute across them produces a column that means five different things.

The lifecycle unit is the model year, with model-year transitions rather than season exits, dealer prebooks setting the commitment, and counter-seasonal buying — the order lands long before the sell window. Depth is constrained by the prebook and by MAP policy, which limits price as an exit lever, so a depth error stays on the books longer than it would in a category that can mark down freely.

Where this structure breaks in outdoor: collapsing model year and season into one level, so a carryover model that has not changed reads as a new introduction, or a genuinely new model gets planned on the outgoing model's history. Both errors point the buy at the wrong reference. See outdoor brands and merchandise planning for outdoor brands.

Health & beauty

The levels: division, category (skincare, color, fragrance, hair, tools), franchise or line, product, shade or variant, format and fill size.

The decision level is the shade or variant within a franchise. The franchise is the level a merchant defends and funds — it occupies the merchandising grouping slot — but the money is committed shade by shade, which is why the decision level sits one step below the franchise. Shade occupies the variant axis, and it is not analogous to color in apparel in one crucial respect: shades form a ladder that has to be complete, which makes them behave more like a size run than like a color choice. A foundation range with gaps at the ends of its tone range is not experiencing a stock-out, it is failing at merchandising, and the reputational cost is not captured anywhere in a sell-through number.

Format and fill size occupy the size axis and are a genuine level — travel, standard, jumbo, plus refills. Kits and gift sets sit awkwardly across the tree, because a kit is its own sellable unit that consumes the same component inventory as the singles it contains.

The lifecycle unit is the launch window, layered over a continuity base, with sampling programs and gifting peaks. Depth is constrained by shelf life and batch minimums: period after opening and expiry mean an overbought shade is not simply a markdown problem, it is a write-off with a date attached.

Where this structure breaks in beauty: distributing depth across the shade ladder from a house-level average curve. A shade ladder behaves like a size curve — the same uneven depth across a run that only makes sense as a set — but it is bell-shaped nowhere, so it has to be built per franchise and per door tier, as merchandise planning for health and beauty brands sets out. Shade demand is driven by the customer mix of each door or channel; a national average applied everywhere over-serves the middle of the ladder and starves the ends in exactly the locations where the ends are the volume. See health and beauty brands.

Sporting goods

The levels: division, discipline, category, model, model-color or specification, and a discipline-specific fit axis.

The decision level is the model within a discipline, and the discipline is doing most of the merchandising work — it is the level a buyer defends, because coverage of a sport is judged as a whole. Colorway occupies the variant axis where color choice exists, which it does on the soft-goods side and generally not in hardlines.

The size axis is the interesting one: the size axis is different in every discipline. Bat length and weight, racquet grip size, glove size, frame size, board length, shaft flex. These are all legitimately "the thing that varies within a model", and none of them is convertible into the others. A hierarchy that imposes one shared size attribute across disciplines produces a field whose meaning depends on which branch you are standing in, and every cross-discipline report built on it is meaningless.

The lifecycle unit is the model year, tied to the sport's own calendar, sold heavily through a dealer channel that prebooks. Depth is constrained by the model-year transition, because the outgoing model has to clear before the incoming one lands and the two compete directly.

A recurring failure mode in sporting goods: forcing a single size dimension across disciplines in the name of consistency, which trades a real merchandising distinction for a tidy-looking schema. The correct structure gives each discipline its own fit axis and does not pretend they roll up. See sporting goods brands and merchandise planning for sporting goods brands.

Toys & games

The levels: division, category (construction, plush, dolls, games and puzzles, collectibles), age grade, property or brand, item, packaging SKU.

The decision level is the item in its packaging configuration, and the orderable unit is the case or inner pack — not the each. This is the defining structural fact of the category, and it is why toy plans that are built in eaches and rounded to cases at purchase-order time silently rewrite their own assortment.

Nothing occupies the variant axis in the way color does elsewhere. Variety within a line comes from character, property or model rather than color, and where color appears it is set inside the case assortment mix rather than chosen independently. Nothing occupies the size axis either. Age grade is the nearest analogue and it is a merchandising grouping, not a variant axis — an item has one age grade; it does not span a run of them.

The lifecycle unit is the holiday-anchored season, crossed with the licensed window when a property is involved, and gated by the retailer's planogram reset. Two external parties own the calendar. Depth is constrained by case and inner pack multiples, so planned depth has to resolve to an orderable quantity at every level of the plan — the pre-pack versus singles question, answered before the plan is written rather than at purchase-order time.

Where this structure breaks in toys and games: putting the license or property above the category, so licensed and evergreen items cannot be compared. The licensed window is a lifecycle attribute, not a hierarchy level; promoting it breaks the comparison between a licensed item and the evergreen item it competes with for the same shelf. See toys and games brands and merchandise planning for toy and game brands.

Baby & juvenile

The levels: division, category (gear, furniture, feeding, safety), system or collection, model, fashion or pattern, SKU.

The decision level is the model in a specific pattern. Pattern occupies the variant axis — the textile treatment on a stroller, car seat or high chair frame — and it functions like a colorway with one important difference: it is attached to a certified product.

Nothing occupies the size axis. What sits in the adjacent position is stage — infant, convertible, booster — defined by weight and height ranges, and stage behaves as a merchandising grouping rather than a variant axis, because a product occupies one stage or spans a declared range rather than being bought across a run.

The lifecycle unit is the model year, paced by the certification cycle rather than a fashion calendar, with registry demand producing a purchase pattern that is anchored to a date the customer chooses. Depth is constrained by certification and lot traceability: inventory is not fungible across production runs, because a recall reaches into specific lots and the business has to be able to find them.

A recurring failure mode in baby and juvenile: treating pattern as a costless variant hidden beneath the model. A pattern change on a certified frame can carry its own certification, its own supplier lead time and its own traceability requirement. A hierarchy that hides it loses the object that recall management and compliance actually operate on, and the recovery is manual every time. See baby and juvenile brands and merchandise planning for baby and juvenile brands.

Jewelry & watches

The levels: division, category (fine, fashion, watches), collection, piece, metal and stone or dial, ring size or strap and bracelet length.

The decision level is the piece in a specific metal, because metal is what moves the cost. The same ring in yellow gold, white gold and silver is three plans, three cost bases and three sell-through curves, and merging them destroys the only view that matters for margin.

The size axis is occupied by ring size, bracelet length and strap size. It is real but shallow compared with apparel, and the part of it that is made or adjusted to order rather than stocked moves the constraint from inventory to bench capacity.

The lifecycle unit is the collection, and collections behave as continuity rather than season — a fine jewelry collection can run for years, punctuated by gifting peaks that change the phasing without changing the assortment. Depth is constrained by a moving metal cost base and by serialisation: where pieces are serialised, the inventory record is per item, and unit economics have to be computed against the cost the piece was actually made at rather than a standard.

Where this structure breaks in jewelry and watches: planning margin at the collection level on a standard cost while the metal cost base moves underneath it. The plan looks healthy through the season and the realized margin arrives different, with no line in the plan that explains why. See jewelry and watch brands, merchandise planning for jewelry and watch brands and planning margin on a moving cost base.

The ten hierarchies side by side

The same four slots and two axes, filled in ten times. This table is the short answer to "what do we call the levels in our category".

The first four content columns are the four slots that every category occupies. The last three are the two axes that sit between the decision level and the fulfillment level, plus the clock the category's lifecycle runs on — and unlike the slots, an axis can legitimately be empty.

Read the table by column rather than by row and the pattern is clearer than any individual entry. The financial rollup is nearly identical everywhere, because finance asks the same questions of every category. The merchandising grouping is where each category's own logic lives. The decision level is where money is committed and is the level a hierarchy has to get right. The variant and size columns are the ones that vary most — including the entries that read "none", which are the most informative cells in the table.

The multi-category problem

This arrives as soon as a brand adds a second category, and the failure is not obvious until the second or third season. A brand with an apparel line, an accessories line and a hard-goods line cannot run one hierarchy shape across all three.

The forces pull in both directions. Finance needs one spine, because open-to-buy, inventory targets and margin are set for the business, not for a category, and three unreconcilable trees mean three unreconcilable plans. Merchandising needs three shapes, because the axes genuinely differ: the apparel line has a size run, the accessories line does not, and the hard-goods line runs on a model year rather than a season.

The workable answer is to separate the two concerns rather than compromise between them.

Share the financial rollup

Division and department — or whatever the top two levels are called — should be common, and every category should reconcile into them. This is the level where a merchant's plan meets the P&L, and it must be one number.

Let the planning levels differ beneath it

Below the shared rollup, each category gets the levels it actually decides at. The apparel branch carries class, style, style-color and size. The accessories branch carries category, style and colorway, and carries no size level at all. The hard-goods branch carries collection, model, configuration and finish, and carries model year instead of season. These branches are different depths, and that is correct — a hierarchy is not obliged to be symmetrical, and forcing symmetry is what creates the empty axes.

Make the mapping explicit

Somewhere there has to be a documented statement of what rolls into what, maintained deliberately rather than implied by a set of ERP codes. When it is implicit, the first person to add a category invents a mapping, and the second person invents a different one.

Test each branch against the four slots

For each category, name the financial rollup, the merchandising grouping, the decision level and the fulfillment level. If any of the four is empty, the branch has a gap. If two of them are the same level, confirm that is deliberate rather than accidental. If a level exists that fits none of the four and is not the variant or size axis, it is a reporting attribute and should be moved out of the planning hierarchy.

Two decisions cause disproportionate pain later, and both are about which category set the template. A brand that starts in apparel and expands will tend to push size onto categories that have none. A brand that starts in hard goods and adds apparel will tend to plan at model level and discover halfway through the first season that color was never in the plan. Neither is difficult to avoid; both are difficult to unwind once a year of history has been recorded against the wrong shape.

A last point that applies to all ten. A merchandise hierarchy is a planning decision that gets encoded as a data structure, not a data structure that constrains planning. When it is set the other way around — the tool's default shape decides how the category is planned — the shape wins, and it will keep winning for as long as the history sits underneath it. Choosing it deliberately, once, costs a fraction of what unwinding it costs after the history has accumulated underneath it. For the process this structure exists to serve, start with what merchandise planning is.