Multi-Echelon, Aggregate, and Capacity Planning

Introduction

Inventory decisions become substantially more difficult when products move through several locations and channels. A firm may hold inventory at suppliers, plants, regional distribution centres, stores, third-party logistics providers, marketplaces, and customer-facing channels. Decisions made at one echelon affect service, cost, risk, and capacity throughout the network. Optimising each location independently can therefore produce a poor overall result.

This chapter examines integrated planning across the supply network. It focuses on multi-echelon inventory coordination, inventory pooling, postponement, allocation, distribution requirements planning (DRP), aggregate planning, and capacity constraints. The central managerial problem is to balance responsiveness and availability against working capital, operating cost, capacity, and risk.

The chapter is particularly relevant to entrepreneurial and growth-oriented firms. Start-ups and expanding ventures often operate with limited cash, uncertain demand, constrained suppliers, and rapidly changing channels. Their inventory systems must therefore be designed not merely to minimise inventory, but to place the right inventory at the right stage and location.

Learning Objectives

By the end of this chapter, readers should be able to:

1. Explain the structure and logic of multi-echelon inventory systems.
2. Evaluate inventory pooling and postponement decisions.
3. Distinguish allocation planning from replenishment planning.
4. Construct and interpret a basic DRP schedule.
5. Explain aggregate planning alternatives and their relationship to capacity.
6. Identify bottlenecks and assess capacity feasibility.
7. Integrate inventory, demand, capacity, and channel decisions.
8. Critically evaluate technology, governance, and implementation risks.

1. The Multi-Echelon Supply Network

An echelon is a stage in a supply network at which inventory may be stored or transformed. A simple network may contain a supplier, a manufacturing plant, a regional distribution centre, and retailers. Inventory at a downstream echelon is physically closer to the customer, whereas inventory at an upstream echelon can often serve several downstream locations.

A multi-echelon system differs from a collection of independent warehouses in three important ways:

• Demand is transmitted through linked locations rather than observed independently everywhere.
• Replenishment decisions at one echelon affect lead times and availability at other echelons.
• The relevant safety stock is determined by total network uncertainty, not simply by the uncertainty at one site.

The inventory position at a location is commonly defined as:

where IP is inventory position, OH is on-hand inventory, OO is on-order inventory, and BO is backorders. All quantities must use the same units and time basis. The expression is useful for replenishment decisions, but it does not by itself reveal where inventory is located or whether capacity is available.

Echelon Inventory

Echelon inventory represents inventory at a particular stage plus inventory held at all downstream stages supplied by that stage. If a plant supplies two distribution centres, the plant's echelon inventory includes plant stock and the stock held at both distribution centres, subject to the network definition being used.

Echelon stock is useful because upstream decisions should reflect the inventory already positioned downstream. It prevents managers from treating every echelon as if it were independent. However, echelon calculations require accurate bills of distribution, shipment visibility, lead times, and ownership information.

Centralised and Decentralised Control

Under decentralised control, local facilities make replenishment decisions using local forecasts and service targets. This can be responsive to local conditions but may cause duplicated safety stock, inconsistent priorities, and competing orders.

Under centralised or coordinated control, a planning authority considers total network demand, inventory, capacity, and constraints. Centralisation improves visibility and pooling, but may be slower, less sensitive to local knowledge, and more dependent on data quality.

A hybrid model is often preferable. Strategic parameters—such as service policies, postponement points, inventory targets, and capacity reservations—may be centrally governed, while local teams manage exceptions and customer-specific information.

2. Inventory Pooling and Risk Aggregation

Inventory pooling combines demand streams so that uncertainty is shared rather than separately protected at every location. Pooling is especially valuable when demands are not perfectly positively correlated.

For independent demand streams, the standard deviation of aggregate demand is:

where σ_T is the standard deviation of total demand and σ_i is the standard deviation of demand stream i. The equation assumes that demand streams are independent.

More generally, when correlations matter:

Here, ρᵢⱼ is the correlation between demand streams i and j. Positive correlation reduces the risk-reduction benefit of pooling; negative correlation increases it. The formula is dimensionally consistent because variances have units of squared demand.

If safety stock is based on a normal-demand approximation, a basic expression is:

where SS is safety stock, z is the standard-normal service factor, and σ_L is the standard deviation of demand during lead time. This formulation assumes an appropriate normal approximation, a defined service measure, stable lead-time conditions, and a replenishment policy for which the calculation is relevant. If demand or lead time is highly intermittent, skewed, or dependent, simulation or a non-normal model may be more appropriate.

Pooling may be achieved through:

• A central warehouse serving multiple markets.
• Common components used in several finished products.
• Shared safety stock across sales channels.
• Virtual inventory visibility that allows orders to be fulfilled from the best available location.
• Vendor-managed inventory or postponement arrangements.

Pooling has costs. A central facility may increase delivery distance, transportation time, disruption exposure, or dependence on a single site. Pooling should therefore be evaluated using total cost and service performance rather than safety stock alone.

3. Postponement and the Strategic Location of Variety

Postponement delays product differentiation until demand information is clearer. A firm may produce a common base product and add colour, packaging, software configuration, or regional labelling closer to the customer.

Postponement separates two types of demand:

1. Demand for the common or generic form.
2. Demand for the final configuration.

The generic form benefits from pooling because it can be allocated to whichever final variant is needed. The trade-off is that late-stage operations may require specialised equipment, labour, information, or transportation.

A postponement decision should consider:

• The number and volatility of final variants.
• The value and obsolescence risk of finished goods.
• The cost and speed of late configuration.
• Regulatory and labelling requirements.
• Required customer lead time.
• Available capacity at the postponement point.

Postponement is not simply a manufacturing technique. It can involve packaging, pricing, channel assignment, customs processing, software activation, or service bundling. In entrepreneurial firms, postponement can reduce cash tied up in low-probability variants while preserving market responsiveness.

4. Allocation Planning and Replenishment Planning

Replenishment asks: how much should be ordered or produced? Allocation asks: which customer, channel, market, or location should receive scarce supply?

Allocation becomes important when available supply is less than total committed or forecast demand. A mathematically simple allocation model is:

where xᵢ is quantity allocated to demand segment i, mᵢ is contribution margin per unit, aᵢ is the scarce resource consumed per unit, C is available capacity or supply, and dᵢ is eligible demand. This model assumes linear contribution, divisible quantities, known resource consumption, and no interaction effects. Real allocation policies may also include contractual commitments, strategic customers, fairness, minimum quantities, and service obligations.

Common allocation rules include:

• Pro rata allocation: supply is distributed in proportion to demand.
• Priority allocation: higher-priority customers or channels are served first.
• Margin-based allocation: supply is directed to the highest contribution opportunities.
• Strategic allocation: supply protects launch markets, contractual relationships, or long-term positioning.
• Fair-share allocation: supply is balanced using predetermined service or historical criteria.

No rule is universally superior. A margin-maximising rule may damage a critical customer relationship. A purely pro rata rule may allocate scarce products to low-value or low-strategic channels. Governance should specify the decision rights, data, review frequency, and escalation process.

5. Distribution Requirements Planning

Distribution requirements planning translates a demand or deployment plan into time-phased replenishment requirements across a distribution network. DRP is the distribution analogue of material requirements planning (MRP), although the assumptions and data structures differ.

A basic DRP record includes:

• Beginning projected available balance.
• Forecast or gross requirements.
• Scheduled receipts.
• Safety-stock requirement.
• Projected available balance.
• Net requirements.
• Planned order receipts.
• Planned order releases, offset by transportation and replenishment lead time.

The projected available balance for period t can be represented as:

where PAB_t is projected available balance, SR_t is scheduled receipt, POR_t is planned order receipt, and GR_t is gross requirement in period t. A planned receipt is normally triggered when projected stock would fall below the policy level. The planned release is then the receipt shifted backward by the relevant lead time.

Worked DRP Example

A regional distribution centre begins week 1 with 120 units. Its safety-stock target is 40 units. Forecast requirements are 70, 90, 60, and 100 units in weeks 1–4. A scheduled receipt of 80 units arrives in week 2. Replenishment lead time is one week. Assume lot-for-lot ordering.

For week 3, without a planned receipt, projected stock would be 40 − 60 = −20, which is below the 40-unit target. A 60-unit receipt restores ending stock to 40. The same logic produces a 100-unit receipt in week 4. The schedule is feasible only if upstream supply and transportation capacity can support the releases.

DRP should not be treated as an automatic answer. It depends on forecast quality, lead-time reliability, order policies, capacity, minimum shipment quantities, and the accuracy of inventory records. A DRP system can calculate a precise plan from inaccurate inputs and thereby create false confidence.

6. Aggregate Planning

Aggregate planning determines medium-term production, procurement, workforce, inventory, and subcontracting decisions for product families rather than individual stock-keeping units. The planning horizon often ranges from several months to approximately two years, depending on the industry.

The aggregate plan links strategic intentions to operational feasibility. It asks how the organisation will respond to expected demand while managing workforce and capacity constraints.

Main Planning Strategies

Level strategy. Production or workforce is kept relatively stable, with inventory and backorders absorbing demand variation. This can support workforce stability but may create high holding costs or poor service.

Chase strategy. Production or workforce follows demand more closely. This reduces finished-goods inventory but may increase hiring, training, overtime, layoffs, or subcontracting costs.

Mixed strategy. The firm combines regular production, overtime, inventory, temporary labour, subcontracting, and backlog controls.

Demand-management strategy. Prices, promotions, reservations, product design, or delivery promises are used to influence demand timing or volume.

A simplified aggregate-planning objective is:

subject to a flow balance such as:

Here, Iₜ is ending inventory, Bₜ is backlog, Pₜ is regular production, Uₜ is additional supply such as overtime or subcontracting, Dₜ is demand, and the c terms are relevant unit costs. Workforce and capacity constraints must be added separately. The objective is illustrative rather than a universal model; practical plans may include setup costs, minimum batch sizes, service constraints, cash limits, and nonlinear costs.

Capacity Feasibility

Capacity is the maximum output or workload that a resource can provide under stated conditions. It is not identical to theoretical machine hours. Planned capacity should account for maintenance, breaks, changeovers, absenteeism, quality losses, and other losses.

A basic capacity relationship is:

where N is the number of parallel resources, H is scheduled hours per resource, A is availability, and E is effective performance or utilisation factor. The factors must be defined consistently; double-counting losses will understate capacity.

For a product requiring rᵢ hours per unit, total required load is:

where qᵢ is planned quantity of product i. A plan is capacity-feasible only when load does not exceed available capacity for each constrained resource and period. Capacity feasibility does not guarantee material availability, labour competence, cash availability, or customer service.

Bottlenecks and the Theory of Constraints

A bottleneck is a resource whose capacity limits system throughput. The Theory of Constraints proposes five broad actions: identify the constraint, exploit it, subordinate other activities to it, elevate its capacity, and repeat after the constraint moves.

The managerial implication is significant: improving a non-bottleneck may increase local efficiency without increasing total throughput. For example, producing more components upstream of a constrained packaging line may only increase work-in-process inventory. Capacity planning must therefore examine the whole flow, not isolated resource utilisation.

7. Integrating Inventory, Allocation, DRP, and Capacity

These planning processes should form a hierarchy rather than operate as disconnected systems:

1. The business plan establishes financial, market, and strategic boundaries.
2. The aggregate plan converts demand expectations into family-level production and capacity decisions.
3. The master production or deployment schedule specifies time-phased supply.
4. DRP distributes replenishment requirements across locations.
5. Allocation rules manage shortages and competing demand.
6. Execution systems monitor orders, inventory, capacity, and exceptions.

A change at one level should be tested against the others. For example, a marketing campaign may increase demand at retail locations. DRP may generate replenishment orders, but the aggregate plan must verify production capacity, procurement must verify component availability, and allocation rules may be required if supply is insufficient.

Useful performance measures include:

• Fill rate and case-fill rate.
• On-time-in-full delivery.
• Inventory turns and days of supply.
• Total supply-chain cost.
• Forecast error and forecast bias.
• Capacity utilisation and bottleneck throughput.
• Backorder age and cancellation rate.
• Cash-to-cash cycle time.
• Obsolescence and write-off value.

Metrics should be balanced. Measuring only inventory turns can encourage understocking; measuring only service can encourage excessive inventory and costly expediting.

8. Illustrative Case: NovaHome’s Omnichannel Launch

Background

NovaHome, an illustrative entrepreneurial company, sells modular smart-lighting kits through its website, specialist retailers, and a large online marketplace. Each kit can be configured in four finishes and three regional plug specifications. Forecast error is high because the product is new. The firm has one assembly plant and two regional distribution centres.

Problem

NovaHome plans to launch a national advertising campaign. The commercial team wants all twelve finished variants available at both distribution centres. Finance argues that finished-goods inventory will consume scarce cash. Operations reports that final configuration and packaging are performed on a single constrained line. The marketplace also requires a high service level and may impose penalties for stockouts.

Analysis

Independent stocking of all finished variants at both distribution centres would duplicate safety stock. Demand streams are not perfectly correlated, so a central pool of common kits could reduce risk. The company also has reliable information about regional plug specifications but weak information about finish preferences. This supports postponing finish selection while completing regional electrical configuration earlier in the process.

Capacity analysis shows that the final packaging line is the bottleneck. Producing large quantities of all finished variants before demand is known would consume capacity and increase obsolescence. A DRP schedule based on channel forecasts would also be unstable if marketplace orders are highly promotional.

Approach and Solution

NovaHome adopts a coordinated policy:

• Hold common semi-finished kits at the plant and at one central postponement location.
• Configure regional plug specifications before shipment where regulatory requirements demand it.
• Apply finish-specific packaging after demand signals become clearer.
• Use a defined allocation hierarchy: contractual safety obligations first, strategic retail accounts second, and margin-adjusted discretionary demand third.
• Use DRP for time-phased deployment from the central location to regional distribution centres.
• Reserve packaging capacity for high-service channels and impose a weekly allocation review during the launch.
• Use a mixed aggregate plan combining regular production, limited overtime, and a qualified subcontractor for non-critical packaging.

Outcome

The policy reduces duplicated finished-goods inventory while preserving access to a common pool. It also exposes the true constraint: packaging capacity rather than assembly capacity. Management can now evaluate whether overtime, subcontracting, an additional packaging shift, or a narrower launch assortment provides the best economic result.

Learning Points

The case demonstrates that postponement, pooling, allocation, DRP, and aggregate planning are interdependent. A technically correct replenishment schedule may still be infeasible if it ignores capacity. Conversely, a capacity expansion may be unnecessary if common inventory and postponement reduce the required number of finished variants.

9. Technology, Data, and Artificial Intelligence

Enterprise resource planning, warehouse-management systems, transportation-management systems, and advanced planning platforms can provide shared visibility across echelons. Application programming interfaces and electronic data interchange can connect suppliers, marketplaces, carriers, and internal systems.

Artificial intelligence and machine-learning tools may improve demand sensing, anomaly detection, lead-time estimation, and allocation recommendations. Their value depends on data quality and decision design. A model trained on historical stockouts may interpret lost sales as low demand. A model optimising short-term margin may systematically deprioritise vulnerable customers or strategic channels.

Good governance requires:

• Clear definitions of inventory, demand, service, and capacity.
• Traceable data lineage.
• Human review of exceptional or high-impact allocations.
• Monitoring for forecast bias and segment-level discrimination.
• Audit logs for recommendations and overrides.
• Cybersecurity and access controls.
• Scenario testing before deployment.

Technology should support managerial judgement, not conceal unresolved policy choices.

10. Challenges, Limitations, and Ethical Issues

Multi-echelon planning is limited by uncertainty. Forecasts are imperfect, lead times vary, capacity can fail, and demand may change in response to availability. Models based on normal demand, fixed lead times, or linear costs may be useful approximations but should not be treated as facts.

Important limitations include:

• Bullwhip effects caused by forecast updating, batching, promotions, and shortage gaming.
• Inaccurate inventory records and unreported shrinkage.
• Correlated disruptions across supposedly independent suppliers or markets.
• Hidden capacity constraints at suppliers and logistics providers.
• Conflicting incentives between sales, operations, finance, and channel partners.
• Excessive centralisation that weakens local responsiveness.

Ethical and governance concerns include fair allocation during shortages, truthful delivery promises, responsible use of customer data, labour impacts of capacity changes, and avoiding deliberate scarcity manipulation. Regulatory requirements may affect safety stock, traceability, product labelling, recall capability, and cross-border inventory ownership.

11. Best-Practice Implementation Framework

A practical implementation sequence is:

1. Map the physical and information network, including ownership and lead times.
2. Define service measures and customer priorities.
3. Segment products by value, volatility, criticality, and postponement potential.
4. Establish common master data and inventory definitions.
5. Identify bottlenecks and validate capacity assumptions.
6. Design pooling, postponement, and allocation policies.
7. Build aggregate and DRP planning processes with exception management.
8. Test plans under demand, supply, and capacity scenarios.
9. Pilot the policy in one product family or region.
10. Review outcomes and revise parameters regularly.

The strongest systems combine analytical discipline with clear decision rights. They make trade-offs visible rather than pretending that every objective can be maximised simultaneously.

Chapter Summary

Multi-echelon inventory planning coordinates stock, demand, replenishment, and capacity across interconnected locations. Echelon thinking prevents managers from optimising sites independently. Inventory pooling can reduce aggregate uncertainty, but its benefit depends on demand correlation, transportation, disruption, and service requirements. Postponement delays final differentiation and is valuable when product variety and demand uncertainty are high.

Allocation planning determines how scarce supply is distributed, whereas replenishment planning determines when and how much supply is ordered. DRP creates time-phased deployment plans, but its outputs are only as reliable as the demand, inventory, lead-time, and capacity data behind them. Aggregate planning balances production, workforce, inventory, subcontracting, and backlog over a medium-term horizon. Capacity analysis identifies whether the plan is feasible and directs attention to bottlenecks.

Effective practice integrates strategic policy, aggregate planning, DRP, allocation, execution, and performance measurement. Technology and AI can improve visibility and decision support, but governance, data quality, human accountability, and ethical allocation remain essential.

Key Terms

Echelon inventory: Inventory at one stage plus inventory held at downstream stages supplied by that stage.

Multi-echelon system: A supply network containing several connected inventory and transformation stages.

Inventory pooling: Combining demand or inventory resources to share uncertainty.

Demand correlation: The statistical relationship between demand movements in two or more streams.

Safety stock: Inventory held to protect against uncertainty in demand, supply, or lead time.

Postponement: Delaying final product differentiation until demand information is clearer.

Allocation: The process of assigning scarce supply to competing demands.

Replenishment: The process of restoring inventory through purchasing, production, or transfers.

Distribution requirements planning: Time-phased planning of replenishment across a distribution network.

Projected available balance: Expected inventory after receipts and requirements in a period.

Aggregate planning: Medium-term planning for product families, capacity, workforce, and inventory.

Level strategy: A planning approach that keeps production or workforce relatively stable.

Chase strategy: A planning approach that adjusts production or workforce to demand.

Capacity: The output or workload a resource can provide under defined conditions.

Bottleneck: A resource that constrains system throughput.

Throughput: The rate at which a system generates output or contribution.

Fill rate: The proportion of demand supplied immediately from available inventory.

Bullwhip effect: Amplification of demand variability as orders move upstream.

Omnichannel inventory: Inventory coordinated across multiple customer-facing channels.

Inventory position: On-hand inventory plus on-order inventory minus backorders.

Review Questions

Short Answer Questions

1. What is an echelon, and how does echelon inventory differ from inventory at a single location?
2. Why does demand correlation affect the value of inventory pooling?
3. Distinguish allocation planning from replenishment planning.
4. What information is included in a basic DRP record?
5. Why can improving a non-bottleneck fail to increase total throughput?

Long Answer Questions

1. Explain the advantages and disadvantages of centralised versus decentralised multi-echelon inventory control.
2. Discuss how postponement can reduce working capital while preserving customer responsiveness.
3. Compare level, chase, mixed, and demand-management strategies in aggregate planning.
4. Explain how DRP, aggregate planning, and capacity planning should be integrated.
5. Evaluate the governance and ethical risks of using AI for inventory allocation.

Analytical / Application Questions

1. A firm serves three markets with demand standard deviations of 40, 30, and 50 units. Assuming independent demand, calculate the standard deviation of pooled demand and explain the managerial implication.
2. A distribution centre has 500 units available and receives eligible demands of 300, 250, and 200 units from three channels. Design and justify an allocation rule using service, margin, and strategic considerations.
3. A packaging resource has two machines, 40 scheduled hours per week each, availability of 90%, and effective performance of 85%. Calculate weekly effective capacity and explain how the result should be used in an aggregate plan.

Multiple Choice Questions with Answers

Question 1

Which statement best describes echelon inventory?

A. Inventory held only at the retail store

B. Inventory at a stage plus inventory held downstream from that stage

C. Inventory that has been physically damaged

D. Inventory that is reserved exclusively for one customer

Correct Answer: B. Inventory at a stage plus inventory held downstream from that stage

Question 2

If two demand streams are independent, pooled demand variability is calculated by:

A. Adding their standard deviations directly

B. Multiplying their standard deviations

C. Taking the square root of the sum of their variances

D. Subtracting the smaller standard deviation from the larger

Correct Answer: C. Taking the square root of the sum of their variances

Question 3

What is the principal purpose of postponement?

A. To increase the number of finished variants held in stock

B. To delay final differentiation until demand information improves

C. To eliminate all transportation lead time

D. To replace capacity planning

Correct Answer: B. To delay final differentiation until demand information improves

Question 4

Allocation planning primarily determines:

A. Which demand receives scarce available supply

B. The accounting value of fixed assets

C. The design of warehouse buildings

D. The historical accuracy of a forecast

Correct Answer: A. Which demand receives scarce available supply

Question 5

In a basic DRP schedule, a planned order release is generally:

A. The planned receipt shifted backward by replenishment lead time

B. The gross requirement multiplied by safety stock

C. The beginning inventory minus demand

D. The scheduled receipt plus backlog

Correct Answer: A. The planned receipt shifted backward by replenishment lead time

Question 6

A level aggregate-planning strategy generally attempts to keep which factor relatively stable?

A. Every customer’s demand

B. Production or workforce

C. Transportation distance

D. Forecast error

Correct Answer: B. Production or workforce

Question 7

A bottleneck is best defined as:

A. The resource with the highest purchase price

B. A resource that limits system throughput

C. Any warehouse with safety stock

D. A supplier located overseas

Correct Answer: B. A resource that limits system throughput

Question 8

A resource has two machines, 40 scheduled hours each, 90% availability, and 85% effective performance. Its effective weekly capacity is:

A. 61.2 hours

B. 68.0 hours

C. 72.0 hours

D. 80.0 hours

Correct Answer: A. 61.2 hours

Question 9

Which factor generally reduces the risk-reduction benefit of inventory pooling?

A. Independent demand

B. Negative demand correlation

C. Positive demand correlation

D. Shared common components

Correct Answer: C. Positive demand correlation

Question 10

Why should AI-generated allocation recommendations be governed by human review?

A. Algorithms cannot process any numerical data

B. Historical data may contain bias, stockout distortion, or inappropriate incentives

C. Human review always eliminates inventory cost

D. AI systems are designed only for accounting

Correct Answer: B. Historical data may contain bias, stockout distortion, or inappropriate incentives