How do I set up a demand planning platform for size and color allocation across multiple luxury stores?

To set up a demand planning platform for size and color allocation across multiple luxury stores, think of it as a forecasting + assortment + replenishment + store allocation system built around your product hierarchy and client demand patterns.

Here’s a practical blueprint.

---

1) Define the business objective

For luxury retail, the goal is usually not just “sell more,” but:

• Maximize sell-through
• Preserve brand presentation
• Reduce markdowns and end-of-season leftovers
• Improve in-store availability of key sizes/colors
• Allocate inventory by store profile, not just volume

Because luxury demand is often:

• highly skewed by store,
• sensitive to seasonality and events,
• driven by clienteling/VIP behavior,
• concentrated in specific sizes and colors,

your platform needs to plan at a SKU-store level with strong support for style-color-size hierarchies.

---

2) Build the required data foundation

You’ll need clean, reliable data from several sources.

Core data inputs

1. Product master

   • Style
   • Color
   • Size
   • Category / subcategory
   • Season
   • Material
   • Launch date
   • Retail price / cost
   • Lifecycle status

2. Store master

   • Store ID
   • Region/country
   • Store tier
   • Footfall
   • Selling area
   • Local climate
   • Client profile
   • Historical demand pattern

3. Historical sales

   • Units sold by day/week
   • Store-SKU-level transactions
   • Sell-through
   • Returns
   • Lost sales if available

4. Inventory and replenishment data

   • On-hand
   • On-order
   • In-transit
   • Transfers
   • Stockouts

5. Assortment/merchandising rules

   • Mandatory colors
   • Size curves by product type
   • Store-specific assortment depth
   • Launch pack rules
   • Minimum display quantities

6. External drivers

   • Season/weather
   • Holidays
   • Tourism
   • Local events
   • Promotions or brand activations

---

3) Create a product hierarchy for planning

Your planning model should not start at raw SKU alone. Use a hierarchy such as:

• Division
• Category
• Subcategory
• Style
• Color
• Size
• SKU

This allows you to forecast and allocate at the right level:

• Category/style level for early launch forecasting
• Color level for visual merchandising and regional preference
• Size level for fit curves and replenishment
• SKU-store level for final allocation and replenishment

For luxury, many retailers use style-color-size as the operational unit, but aggregate forecasts at higher levels when data is sparse.

---

4) Segment stores into demand clusters

Not all stores should follow the same size/color curve.

Create store clusters based on:

• sales volume,
• customer demographics,
• climate,
• tourist vs local mix,
• luxury tier,
• regional style preferences,
• historical color/size sell-through.

Example clusters:

• Flagship city stores
• Tourist-heavy boutiques
• Wealth-residential neighborhood stores
• Resort stores
• Outlet stores
• Emerging-market luxury stores

Each cluster can have its own:

• color preference profile,
• size curve,
• inventory depth target,
• replenishment frequency.

---

5) Model size and color demand separately

A strong setup usually separates demand into components:

A. Style demand

“How much of this style will sell in this store?”

B. Color mix

“What share of that style demand goes to each color?”

C. Size curve

“What share of each color/style demand goes to each size?”

This is better than forecasting every SKU independently, especially when you have limited history.

Example

If a style is forecast to sell 100 units in Store A:

• Black = 45%
• Ivory = 35%
• Red = 20%

Then for Black:

• Small = 20%
• Medium = 50%
• Large = 30%

The final allocation becomes much more reliable and explainable.

---

6) Select the forecasting approach

Use a combination of methods rather than one model for everything.

Recommended forecasting layers

1. Baseline statistical forecast

   • Moving averages
   • Exponential smoothing
   • Seasonal models
   • Hierarchical forecasting

2. Machine learning forecast

   • Gradient boosting
   • Random forest
   • Time-series models with features
   • Demand sensing models

3. Rules-based overrides

   • New launches
   • Limited editions
   • VIP-driven allocation
   • Strategic stores
   • Fashion calendar events

For luxury retail, also consider:

• sparse demand methods,
• intermittent demand forecasting,
• analog-style forecasting for new products,
• lifecycle curves by product type.

---

7) Build allocation logic

Once demand is forecast, the platform should allocate inventory by store based on constraints.

Allocation inputs

• Forecast demand by store-SKU
• Starting inventory
• Minimum presentation stock
• Safety stock
• Store rank / tier
• Pack constraints
• Warehouse availability
• Transfer opportunity

Allocation priorities

A typical logic might be:

1. Protect key stores and high-value clients
2. Cover must-have sizes/colors
3. Meet minimum launch quantities
4. Optimize sell-through
5. Avoid overstock in low-demand stores
6. Reserve inventory for replenishment if needed

Allocation methods

• Proportional allocation
• Demand-ranked allocation
• Priority scoring
• Optimization model
• Constrained integer programming for exact pack allocation

---

8) Add size curve logic

Size allocation is especially important in luxury because fit curves vary by category and region.

Build size curves by:

• category
• style family
• gender
• store cluster
• region

Example:

• Women’s shoes in Europe may have different size distributions than in Asia
• Outerwear size curves vary by climate and customer profile

Use historical sales to generate:

• base size curve,
• cluster-specific curve,
• store-specific adjustments.

If history is weak, use:

• regional averages,
• peer store analogs,
• category norms,
• buyer overrides.

---

9) Add color allocation logic

Color preference in luxury can be highly store-specific and fashion-driven.

Color allocation should consider:

• brand core colors,
• seasonal fashion colors,
• local preferences,
• visual merchandising strategy,
• newness/novelty value,
• price point sensitivity,
• country-specific cultural preferences.

A good platform should support:

• core color coverage rules
• fashion color trial allocations
• regional color bias adjustments
• color phase-in / phase-out logic

---

10) Set inventory parameters

Define planning rules for every product/store combination:

• Minimum stock
• Maximum stock
• Reorder point
• Safety stock
• Weeks of cover target
• Presentation stock
• Replenishment lot size
• Transfer threshold

Luxury often uses tighter presentation rules:

• enough to display the collection,
• but not so much that inventory ages.

---

11) Design the platform architecture

A standard architecture looks like this:

Data layer

• ERP / POS / WMS / OMS / PLM / e-commerce feeds
• Data warehouse or lakehouse

Planning engine

• Demand forecasting
• Size/color curve engine
• Allocation optimizer
• Replenishment engine
• Exception management

User layer

• Planner dashboard
• Buyer dashboard
• Store allocation view
• Store performance view
• Override workflow

Integration layer

• APIs or batch interfaces with ERP/WMS/OMS
• Automated allocation push to distribution systems
• Store replenishment signals

---

12) Define workflow and governance

This is critical. A platform fails if the planning process is unclear.

Typical workflow

1. Load historical sales and inventory
2. Generate baseline forecast
3. Adjust for new launches/events
4. Apply store cluster size/color curves
5. Generate initial allocation
6. Review exceptions and overrides
7. Release allocation to DC/stores
8. Monitor sell-through and replenish
9. Post-season learning

Governance

• Who can override forecasts?
• Who approves launch allocations?
• What are the approval thresholds?
• How are exceptions documented?
• How are forecast errors tracked?

---

13) Track the right KPIs

Measure both planning accuracy and commercial outcome.

Forecast KPIs

• Forecast accuracy
• MAPE / WAPE
• Bias
• Forecast stability

Allocation KPIs

• Sell-through rate
• Stockout rate
• Weeks of cover
• Markdown rate
• Inventory aging
• Transfer rate
• Lost sales
• Full-price sell-through

Luxury-specific KPIs

• Presentation compliance
• Assortment completeness
• VIP fulfillment rate
• Launch performance by store tier
• Color/size fulfillment rate

---

14) Start with a phased implementation

Don’t try to solve everything at once.

Phase 1: Foundation

• Clean master data
• Build store and product hierarchies
• Standardize KPIs

Phase 2: Forecasting

• Baseline demand forecast by store/style
• Add size and color curves

Phase 3: Allocation

• Build initial allocation rules
• Add constraints and priorities

Phase 4: Replenishment

• Introduce weekly replenishment and transfers

Phase 5: Optimization

• Use scenario planning and optimization models
• Add exception automation
• Improve with machine learning

---

15) Recommended platform capabilities

Your demand planning platform should ideally support:

• Hierarchical forecasting
• Store clustering
• Size/color curve management
• New product forecasting
• Assortment planning
• Allocation optimization
• Replenishment planning
• Transfer planning
• Scenario simulation
• Exception management
• KPI dashboards
• Workflow/approval control

---

16) Practical implementation stack

Depending on your budget and scale:

Enterprise options

• Blue Yonder
• o9 Solutions
• SAP IBP
• Oracle Retail
• Kinaxis

Custom / hybrid option

• Data warehouse: Snowflake, BigQuery, Databricks
• Forecasting: Python, Prophet, XGBoost, statsmodels
• Optimization: Pyomo, OR-Tools, Gurobi
• Visualization: Power BI, Tableau, Looker
• Workflow: custom app or low-code layer

For luxury-specific needs, many companies use a hybrid: enterprise planning tool + custom logic for size/color allocation.

---

17) Common pitfalls to avoid

• Forecasting SKUs with too little history without hierarchy support
• Ignoring store clustering
• Using one size curve for all stores
• Not accounting for new launches or fashion effects
• Over-automating without planner overrides
• Weak master data on product/color/size
• No feedback loop after sell-through
• Not integrating inventory and replenishment data

---

18) A simple target operating model

A good operating model for luxury allocation looks like this:

• Merchandising defines assortment and business rules
• Demand planning forecasts style/store demand
• Allocation planning assigns color/size quantities
• Supply chain fulfills and replenishes
• Store operations provide execution feedback
• Analytics monitors performance and model learning

---

19) If you want a minimum viable setup

Start with this:

1. Clean product/store master data
2. Build weekly sales history by store-SKU
3. Cluster stores into 5–10 groups
4. Create color and size curves by cluster
5. Forecast style demand by store
6. Allocate by forecast + rules
7. Track sell-through and stockouts weekly
8. Refine curves every season

That’s enough to launch a useful planning platform without overengineering.

---

If you want, I can also help you with one of these next:

1. a system architecture diagram for the platform,
2. a data model/table design,
3. a step-by-step implementation roadmap, or
4. a sample allocation algorithm for size/color by store.