Research question: Under which conditions does reducing an assortment's complexity improve contribution, stock quality and industrial reliability without damaging consumer choice or the distinctive value of the collection?
Working propositions
- SKU count is a weak standalone measure of complexity; variant structure, material commitments and supply-chain dependencies matter more.
- Reported gross margin can hide development costs, minimum-order exposure, allocation imbalances, return costs and markdown risk.
- Simplification creates value only when displaced demand is retained, eliminated complexity is genuinely removable, and core brand attributes are preserved.
- A useful decision combines commercial performance, product role, industrial burden and the reversibility of change.
The unit of analysis
“Reduce SKUs by 20%” is an administratively attractive objective. It is rarely a sufficient investment thesis. First, the organisation needs to decide which unit it is counting.
A style is a product design with a defined silhouette and construction. A style-colour introduces a colourway with its own material, forecasting and stock implications. A commercial SKU is commonly the saleable combination of style, colour and size, although conventions vary by enterprise system. A style offered in four colours and five sizes may generate 20 saleable SKUs while relying on one block, one fabric quality and one supplier. Another style with only five saleable SKUs may require a unique fabric, specialist finishing, a separate manufacturing line and additional quality tests.
For decision purposes, maintain distinct counts for styles, style-colours, size-level SKUs, unique materials, components and manufacturing routes. Collapsing these into one number obscures the cost drivers.
An additional complication is the role of a product. A permanent carryover, traffic generator, hero silhouette, commercial basic, image-building item and experimental capsule should not be judged by the same near-term volume threshold. The role must be declared in advance; otherwise unsuccessful items are too easily reclassified as “strategic” after sales results become known.
What product profitability must include
A minimum model starts with realised revenue and separates the expenses that genuinely vary with the product from fixed costs or allocations that may not disappear if the SKU is removed.
For a defined period and channel, a simplified contribution measure is:
SKU contribution before specifically attributable fixed costs = net realised revenue − landed cost of units sold − variable fulfilment/handling − expected returns or claims costs not already netted from revenue − other incremental selling costs.
The exact components depend on the company's accounting and channel structure. Revenue must be net of discounts and returns where those are recognised as revenue reductions; the same amount must not be deducted twice. Inventory write-downs or markdown effects must be captured consistently in revenue, cost of sales or a separate risk measure, not double-counted. VAT should not be mixed with net-of-tax business measures.
Contribution is not identical to gross margin or cash generated. It does not automatically capture the capital tied up in unsold units, capacity consumption, the costs of design and development, supplier minimums, obsolescence or the opportunity cost of a production slot.
A useful portfolio review therefore keeps four ledgers distinct:
Sell-through is often described as units sold divided by units available or received, but business definitions differ. The reporting window, treatment of transfers, returns, replenishments and inventory at the start of the period must be stated. A sell-through figure without a shared denominator is not fit for cross-category comparison.
Where complexity actually accumulates
Material commitments
A unique yarn, fabric, colour standard or trim can create minimum-order exposures and obsolete balances even when finished-garment demand is small. Material interchangeability and stock-reuse options affect this exposure. A fabric shared by five styles may be less risky than one confined to a single colourway, provided quality and product identity remain appropriate.
Pattern and construction proliferation
New blocks, grading rules, trims, bonding or assembly procedures introduce technical work and manufacturing variation. This is particularly visible in categories with specialised equipment or tolerance management. In knitwear, gauge, yarn count, stitch architecture, linking and finishing are not interchangeable decisions; two seemingly similar designs can require different production routes.
Forecast fragmentation
When demand for one commercially coherent family is distributed among many close substitutes, forecasts are split and size/colour imbalances become harder to control. Small planned quantities can increase exposure to supplier minimums or create unfavourable logistics economics. Yet consolidation may also cause demand loss where consumers genuinely value differentiated choice.
Operational coordination
Each distinct item may produce additional master-data, order, approval, fitting, testing, quality and replenishment activities. Their economic effect depends on whether tasks are automated, shared or genuinely incremental. Cost should be estimated from observed process time or supplier charges where possible, not invented from an arbitrary “cost per SKU.”
A four-lens decision model
The proposed working method reviews every style or product family through four lenses, each with explicit evidence rather than a single opaque score.
1. Economic contribution: realised performance by comparable channel and season, expected contribution under a realistic markdown scenario, forecast uncertainty and cash exposure.
2. Customer and brand role: distinctive demand, fit or size coverage, collection coherence, halo effects, repeat-purchase patterns and the customer's ability to substitute an alternative.
3. Industrial burden: dedicated development effort, material/trim overlap, supplier minimums, lead-time and setup constraints, quality risk and rework rates.
4. Transition cost and reversibility: contractual commitments, remaining stock, customer communication, pattern/tooling investment and the possibility of a controlled pilot.
These are decision hypotheses, not universal rules. A protected image product may be economically legitimate, but its role should be funded consciously rather than hidden inside the average profitability of the assortment.
Illustrative scenario (not client data)
Consider a hypothetical fashion brand reviewing eight low-volume colour variants of one carryover style. The question is whether demand can migrate to remaining colourways and whether the proposed simplification reduces avoidable work.
Assumptions: Each candidate variant sold 100 units in the period; net realised selling revenue per unit was €135, landed cost per sold unit €58, incremental fulfilment €6 and a further returns/claims allowance €8 per sold unit, with no amounts duplicated elsewhere. For this simple illustration, all eight variants have identical unit economics.
The calculation does not show that eliminating all eight variants would save €50,400: that would be the contribution potentially placed at risk. A responsible counterfactual must model retained or displaced demand, savings on genuinely avoidable development and operating tasks, unrecoverable raw materials, changes in markdown exposure and any substitution toward higher-performing core variants.
Suppose an analytical scenario—not an observation—assumes 70% of the 800 units of demand would migrate to existing colourways at the same €63 unit contribution, while 30% would disappear. Contribution retained in this simplified model would be 560 × €63 = €35,280; €15,120 would be at risk before considering possible avoidable costs, capacity value or price effects. If only €5,000 of extra complexity costs could truly be eliminated, the direct short-run economic case for deletion would remain negative under those assumptions. If a different portfolio configuration saved materially more cost or released valuable capacity, the result could reverse. These scenarios are sensitive to assumptions; they are not evidence of actual customer substitution.
A second example concerns inventory, not contribution. If a removed colourway leaves 300 units on hand with €58 recorded unit cost, €17,400 of carrying-value exposure exists before any recoverable resale value or write-down. That amount is neither an immediate cash saving from deletion nor necessarily a full loss. The treatment depends on subsequent sales, reuse and accounting policy.
This is why SKU rationalisation is a scenario comparison, not a percentage-reduction target.
Implementation and experiment design
Step 1 — Define the portfolio perimeter. Separate core carryovers, seasonal variants, capsules, wholesale exclusives and development samples. Record which counts and financial periods are in scope.
Step 2 — Reconcile product data. Match sales, returns, stock, costs and supplier information to one controlled identifier. Flag missing sizes, cross-season carryovers, product substitutions and inconsistent unit definitions. Unreconciled data should not be forced into precise rankings.
Step 3 — Assign product roles in advance. Agree which styles serve economic, assortment, brand or channel objectives. Document exceptions to simple profitability thresholds.
Step 4 — Map shared and unique commitments. Material, pattern, machinery, quality and logistics burdens should be examined with technical and production teams. Commercial data alone cannot identify the true avoidable cost of simplification.
Step 5 — Model at least three choices. Maintain status quo, selective rationalisation, and redesign/replatforming. Use consistent demand and price assumptions; report ranges where evidence is weak.
Step 6 — Pilot reversible changes. Limit the first reduction to a manageable product family or channel, where seasonality and customer effects can be observed. Define success and reversal thresholds before launch.
Step 7 — Reassess after real demand and operational data. Inspect retained demand, gross-to-net price, contribution, aged stock, development workload, quality and delivery reliability. A reduction in item count alone is not an outcome.
Limitations and research agenda
This working paper does not provide a validated universal optimal SKU count. Assortment economics vary by price tier, product category, replenishment model, channel, demand volatility and supplier architecture. A 40-style luxury capsule cannot be compared naively to a multi-brand retail network or a wholesale footwear programme.
Attribution is also challenging: a reduction in styles may coincide with price changes, promotions, store openings or macroeconomic conditions. Before-versus-after results without controls cannot isolate the effect of assortment simplification. Return rates can depend on size, fit and channel mix, not just SKU proliferation.
A further investigation should measure actual time spent on development and approvals, track the residual value of shared materials, test substitution between variants and estimate the contribution of capacity released for other products. Results should be re-evaluated over more than one product cycle where feasible.
The governing question is not how many SKUs an organisation can remove. It is whether the assortment's economic and industrial structure is proportional to the value each product creates.
Related Atlas reading: Collection Planning Discipline · Margin Is Designed Upstream · Product Architecture & Industrial Governance.
For product-system development: Creative & Product Office.
Editorial scope and method
This Working Paper develops an independent, testable financial and industrial comparison of assortment decisions. It is distinct from the existing Atlas material on collection planning and upstream margin by explicitly separating realised contribution, stock exposure, industrial commitments and the economics of demand substitution. The four-lens review is an analytical proposal, not a field-validated proprietary diagnostic.
All euro values and customer-substitution assumptions are hypothetical. They demonstrate how an alternative might be evaluated; they are not client data, observed industry averages or verified financial outcomes. Economic conclusions depend on reliable retailer records, actual avoidable costs and evidence of retained demand. Relevant Virgili Studio material is linked in Related Atlas reading above.