Logistics Network Design and Facility Location

Introduction and Context

Logistics network design determines how products, information, and sometimes returns move through a supply chain. It addresses the number, location, role, ownership, capacity, and operating policy of facilities such as plants, ports, cross-docks, distribution centres, fulfillment centres, warehouses, and service depots. These decisions shape transportation cost, inventory investment, customer responsiveness, risk exposure, carbon emissions, and the strategic flexibility of the firm.

Network design is distinct from daily routing or warehouse scheduling. Routing decides how a known shipment should move through an existing network; network design decides which facilities and links should exist in the first place. Because facilities are long-lived and difficult to relocate, design decisions are strategic, often involving substantial sunk costs and uncertain future demand.

A sound design balances at least five objectives:

1. Economic efficiency: facility, handling, inventory, transportation, labor, tax, and technology costs.
2. Service performance: delivery time, availability, order cut-off times, geographic coverage, and reliability.
3. Resilience: the ability to absorb disruption and recover acceptable performance.
4. Strategic fit: compatibility with product characteristics, market structure, and competitive priorities.
5. Sustainability and compliance: emissions, land use, labor standards, safety, data governance, and regulatory obligations.

Learning Objectives

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

• distinguish major logistics facility roles and network structures;
• formulate center-of-gravity, load-distance, capacity, and mixed-integer location models;
• evaluate service coverage, total cost, resilience, and outsourcing trade-offs;
• use scenario analysis to assess uncertainty and facility flexibility;
• interpret quantitative outputs with managerial judgment rather than treating a model as an automatic decision maker.

1. Facility Roles and Network Architectures

A plant transforms materials into finished or intermediate products. A regional distribution centre (RDC) receives consolidated inventory and replenishes markets. A fulfillment centre processes individual customer orders, often using high levels of automation. A cross-dock transfers inbound goods to outbound vehicles with little or no storage. A pool distribution centre consolidates shipments for a geographic area. A service depot holds parts and supports repair or field service. A returns centre inspects, repairs, recycles, or disposes of returned products.

Facility role should be defined by process purpose, not merely by building type. For example, one site may combine storage, postponement, returns processing, and customer service. Combining roles can reduce fixed cost but may create congestion, conflicting material flows, or incompatible labor and technology requirements.

Common architectures include:

• Direct shipping: suppliers or plants ship directly to customers. This reduces intermediate handling but may increase transport cost and shipment fragmentation.
• Plant-to-RDC-to-market: inventory is positioned regionally, usually improving delivery speed while adding facility and inventory cost.
• Centralized fulfillment: one or a few facilities serve a broad market. This can exploit risk pooling but may produce long delivery distances.
• Decentralized fulfillment: many local facilities provide rapid service but require duplicated inventory and management.
• Hub-and-spoke: flows are consolidated through hubs. Scale economies are attractive, although hub failure can create systemic disruption.
• Postponement network: generic products are positioned centrally and customized near the customer, delaying product differentiation and reducing forecast risk.

The appropriate architecture depends on product value density, demand variability, product variety, perishability, order size, lead-time requirements, and the economics of transportation relative to inventory and facility operations.

2. The Economics of Network Design

A network alternative should be assessed using relevant total cost rather than a single visible cost category. A simplified annual cost model is:

Here, TC is total annual logistics cost; F is facility fixed cost; T is transportation cost; H is handling and operating cost; I is inventory carrying cost; P is penalty or service-failure cost; R is risk and disruption cost; and E is relevant environmental or regulatory cost. Monetary terms should use the same period and currency. Some organizations report emissions separately rather than monetizing E.

A more detailed transportation representation is:

where qᵢⱼ is annual quantity shipped from origin i to destination j and cᵢⱼ is cost per unit shipped on that lane. If cost is distance-based, cᵢⱼ may be expressed as currency per tonne-kilometre, multiplied by weight and distance. The units must be consistent: tonnes × kilometres × currency per tonne-kilometre produces currency.

Inventory carrying cost can be approximated as:

where h is the annual carrying-cost rate and V is the average inventory value. The rate may include capital cost, insurance, obsolescence, shrinkage, and storage-related components. This is an approximation; inventory levels should ideally be calculated from demand, lead time, replenishment policy, and service targets.

A network with fewer facilities often has lower fixed and inventory cost through risk pooling, but higher outbound distance. A network with more facilities may lower delivery distance and improve responsiveness, but it duplicates safety stock, management, and infrastructure. The design problem is therefore a trade-off curve rather than a universal rule.

3. Center-of-Gravity and Load-Distance Methods

The center-of-gravity method provides an initial geographic estimate for a facility serving multiple demand points. It is most useful as a screening tool when transportation cost is approximately proportional to volume and distance, roads are reasonably represented by geographic coordinates, and no major geographic constraints dominate.

For demand points i = 1, …, n, with coordinates (xᵢ, yᵢ) and weights wᵢ, the weighted center is:

Weights may represent annual tonnes, pallets, orders, or another relevant flow measure. Both coordinate axes must use the same distance system, and weights must measure the same type of flow. The method assumes linear distance cost and does not automatically account for road networks, congestion, land prices, labor availability, taxes, political risk, or service constraints.

Worked example. Three markets have coordinates and annual shipment volumes as follows:

Total volume is 400 tonnes. Therefore:

x* = (100×10 + 200×40 + 100×20) ÷ 400 = 27.5

y* = (100×20 + 200×30 + 100×60) ÷ 400 = 35

The initial location is (27.5, 35). It is not a final site recommendation. The analyst must test nearby feasible sites using road distances, property costs, labor conditions, flood exposure, zoning, and service requirements.

The load-distance method compares candidate sites using weighted distance:

where LD_k is the load-distance score for candidate site k, wᵢ is the flow to or from demand point i, and dᵢₖ is the distance between site k and point i. The preferred site has the lowest score only when other costs and constraints are comparable. Road distance or travel time is generally more meaningful than straight-line distance.

4. Service Coverage and Facility Capacity

Location is inseparable from service design. A facility may be considered capable of serving a customer if the customer lies within a maximum travel time, delivery-time promise, or distance threshold. A binary coverage parameter can be defined as:

If yₖ equals 1 when candidate site k is opened and xᵢₖ equals 1 when demand point i is assigned to site k, a basic coverage constraint is:

This requires every demand point to be covered by at least one opened facility. It is appropriate for a set-covering interpretation, but it does not by itself guarantee capacity, redundancy, or profitable service.

Capacity must be included when facilities cannot serve unlimited demand:

Here, dᵢ is demand at point i, Capₖ is capacity of site k, and xᵢₖ is the assigned fraction or binary assignment. If split deliveries are not allowed, xᵢₖ is binary and each demand point is assigned to one site. If splitting is allowed, xᵢₖ may be continuous between zero and one, with ∑ₖ xᵢₖ = 1.

Practical capacity is usually less than nameplate capacity because of peaks, maintenance, product incompatibility, labor variability, and desired operating slack. A utilization ratio is:

High average utilization can be financially attractive but operationally fragile. The acceptable value depends on process variability and the consequences of delay; a parcel fulfillment centre and a bulk chemical plant should not use the same target.

5. Optimization Models

A classical capacitated facility-location model chooses which facilities to open and how to assign demand. A simplified mixed-integer formulation is:

subject to:

In this formulation, fₖ is the fixed cost of opening site k, cᵢₖ is the cost of serving demand point i from site k, and Z is modeled cost. The formulation assumes known demand, additive costs, one-period decisions, and assignment feasibility. It omits inventory interactions, congestion, uncertainty, nonlinear tariffs, and multi-period investment unless these are added explicitly.

A richer model may include plant production, supplier-to-plant flows, plant-to-DC flows, DC-to-market flows, facility capacities, minimum throughput, opening dates, service penalties, carbon limits, and multiple products. Binary variables make these models computationally difficult at scale, so analysts may use commercial mixed-integer solvers, decomposition, heuristics, or scenario reduction. Solver output is not automatically a valid managerial recommendation: data quality, omitted constraints, and model assumptions remain decisive.

Multi-echelon and inventory considerations

In a multi-echelon network, inventory at one level may reduce or increase inventory elsewhere. Centralization can reduce safety stock through risk pooling, while decentralization can reduce replenishment lead time. A simple safety-stock relationship under a continuous-review model is:

where SS is safety stock in units, z is the standard-normal service factor, and σ_L is the standard deviation of demand during lead time. This formula assumes an appropriate probabilistic model, a specified cycle-service interpretation, and stable lead-time and demand parameters. If daily demand is independent with standard deviation σ_d and constant lead time L days, then σ_L = σ_d√L. Correlated demand, variable lead time, promotions, and intermittent demand require more careful methods.

6. Scenario Analysis, Uncertainty, and Resilience

Demand forecasts, fuel prices, labor rates, exchange rates, transport capacity, and disruption probabilities are uncertain. A single “most likely” forecast can conceal the risk of a design that performs poorly under plausible alternatives. Scenario analysis evaluates candidate networks across internally consistent futures, such as:

• baseline growth with normal transport conditions;
• high-demand growth and labor scarcity;
• fuel-price escalation and carbon regulation;
• port closure or regional disaster;
• rapid e-commerce growth with smaller, more frequent orders.

If scenario s has probability p_s and network alternative n has cost TC_ns, expected cost is:

Expected cost is not sufficient when severe downside matters. A decision maker may compare worst-case cost, percentile cost, service failure, or conditional value at risk. The selected risk measure should reflect managerial risk tolerance and the consequences of disruption.

Resilience can be designed through multiple suppliers, geographically separated facilities, spare capacity, alternate transportation modes, inventory buffers, postponement, digital visibility, and documented recovery procedures. Redundancy is not free, and duplicated facilities may increase emissions and fixed cost. Resilience analysis should therefore identify critical nodes and links, estimate recovery time, and compare the value of mitigation with its cost.

A useful stress test removes one facility or lane and recalculates feasible service. If a single hub carries nearly all flow, the network may be efficient in normal conditions but structurally vulnerable. However, resilience claims must be operationally credible: an “alternate” facility without trained labor, compatible systems, inventory, or transport capacity is not a genuine substitute.

7. Outsourcing and Ownership Choices

The location decision is linked to the make-or-buy decision. A firm may own a distribution centre, lease it, use a third-party logistics provider (3PL), or combine internal and outsourced facilities. Relevant comparison should include:

• fixed and variable operating cost;
• contract minimums, escalation clauses, and termination rights;
• implementation and switching costs;
• control of customer experience and data;
• provider financial health and operational capability;
• labor, safety, cybersecurity, and regulatory compliance;
• scalability and access to shared networks;
• business continuity and subcontracting transparency.

Outsourcing may convert fixed cost into variable cost and provide expertise or geographic reach. It may also create dependency, weaker process control, data exposure, and coordination costs. A low quoted price is not equivalent to low total cost. Service-level agreements should define measurable indicators, such as order accuracy, on-time dispatch, inventory accuracy, claims handling, recovery time, and reporting obligations. Governance should include audit rights, performance remedies, data ownership, ethical labor requirements, and exit planning.

8. Technology, Data, and Contemporary Developments

Modern network design uses geographic information systems, digital twins, transport-management data, warehouse-management data, demand sensing, and optimization platforms. AI and machine learning can improve demand forecasts, travel-time estimates, anomaly detection, and scenario generation. They do not eliminate the need for causal reasoning, data governance, or human accountability. Historical data may encode past service inequalities or temporary disruptions, and opaque models can make it difficult to justify a location decision.

Important data controls include versioned assumptions, traceable source data, consistent geographic definitions, documented cost allocation, validation against actual shipments, and sensitivity testing. Analysts should distinguish observed facts from assumptions and forecasts. A model that gives a precise answer from unreliable inputs creates false confidence.

Contemporary network design increasingly incorporates urban fulfillment, micro-fulfillment, electric vehicles, parcel lockers, circular flows, reverse logistics, and carbon accounting. These developments can improve service or emissions performance, but they also introduce charging constraints, building limitations, product-handling complexity, and new labor requirements. Sustainability should be measured across the relevant life cycle rather than inferred from facility count alone.

9. Illustrative Case: Northstar Consumer Health Products

Background. Northstar is an illustrative manufacturer of non-prescription health products serving four metropolitan markets. It currently operates one central distribution centre. Online orders have increased, customers expect two-day delivery, and the existing site is approaching effective capacity.

Problem. Management must choose between expanding the central facility, opening two regional fulfillment centres, or using a 3PL for one region. Demand is uncertain, and a flood-risk assessment shows that the current site has higher disruption exposure than previously recognized.

Analysis. The project team first cleansed order and transport data, separated parcel and pallet flows, and defined service as delivery within two business days for 95% of eligible orders. A center-of-gravity calculation provided an initial geographic search area, but candidate sites were then evaluated using road travel time, labor availability, lease cost, flood exposure, parcel-carrier access, and capacity. A capacitated mixed-integer model compared ownership alternatives. Scenarios represented baseline demand, 25% online growth, a fuel-price increase, and temporary closure of the central site.

The expansion option had the lowest normal-condition cost because it preserved scale economies. The two-region option had higher fixed cost but achieved better service and retained partial operation when one region was disrupted. The 3PL option performed well during high growth because capacity could be acquired quickly, but its contract included minimum volume commitments and a limited audit clause.

Approach and solution. Northstar selected a staged design: expand the central facility for pallet and slower-moving products, establish a smaller regional fulfillment centre for high-velocity online products, and contract overflow capacity with a 3PL for an initial two-year period. The contract included transparent labor standards, data access, disaster-recovery tests, volume flexibility, and a defined transition process. The company also reserved space and systems capacity for a second regional site if demand crossed a trigger level.

Outcome. The design did not minimize modeled cost in every scenario. It reduced expected service failure, limited exposure to the central-site hazard, and preserved an option to expand. Key performance indicators included two-day delivery attainment, cost per order, inventory accuracy, fulfillment utilization, recovery time after disruption, emissions per order, and 3PL compliance.

Learning points. First, the center of gravity is a starting point, not a final location. Second, capacity and disruption scenarios can reverse the ranking of alternatives. Third, staged investment and contractual flexibility can be valuable even when they carry a modest premium. Finally, outsourcing must be evaluated as a governance and resilience decision, not only as a labor-cost comparison.

10. Implementation and Best Practices

A disciplined network study normally follows these steps:

1. Define the decision horizon, service promises, products, demand points, and facility roles.
2. Establish a validated baseline of current flows, costs, capacity, and performance.
3. Generate feasible candidate sites using geographic, regulatory, labor, infrastructure, and risk screens.
4. Build the simplest adequate model before adding detail.
5. Include capacity, service, inventory, transportation, emissions, and disruption assumptions.
6. Test multiple demand and operating scenarios, including downside conditions.
7. Compare total cost, service, risk, flexibility, and implementation feasibility.
8. Validate results with operations, finance, procurement, legal, sustainability, and local stakeholders.
9. Pilot or phase the design where uncertainty is high.
10. Monitor realized performance and revisit the design when triggers are reached.

Common errors include optimizing transportation while ignoring inventory, using straight-line distances for urban delivery, assuming average capacity is fully usable, treating forecast demand as certain, omitting returns, and selecting a mathematically optimal but politically or operationally infeasible site. Effective practice combines quantitative modeling with site visits, stakeholder review, sensitivity analysis, and explicit documentation of trade-offs.

Chapter Summary

Logistics network design determines the structure and capabilities of a supply chain over a strategic time horizon. Facility roles, echelon structure, centralization, postponement, and outsourcing influence cost, service, inventory, resilience, and sustainability. Center-of-gravity and load-distance methods are useful screening tools, but real decisions require road-based analysis, capacity constraints, facility fixed costs, inventory effects, and regulatory and geographic feasibility. Mixed-integer location models formalize facility-opening and assignment decisions, while scenario analysis tests uncertain demand, costs, and disruptions. Resilience requires credible alternatives, spare capacity, recovery capability, and governance. Technology and AI can improve data and modeling, but they require validation, transparency, and human accountability. The best network is not necessarily the one with the lowest normal-condition cost; it is the one whose economic, service, risk, ethical, and strategic performance fits the organization’s objectives.

Key Terms

Center of gravity: A weighted geographic estimate of a location based on demand points and flows.

Load-distance method: A site-comparison method that multiplies flow by distance and sums the results.

Facility location: The strategic choice of where facilities should be established.

Network design: The configuration of facilities, transportation links, flows, capacities, and policies.

Distribution centre: A facility that receives, stores, consolidates, and dispatches products.

Fulfillment centre: A facility designed primarily to process and ship customer orders.

Cross-dock: A facility that transfers goods between inbound and outbound vehicles with minimal storage.

Postponement: Delaying final product differentiation until demand is more certain or closer to the customer.

Risk pooling: Combining demand across locations to reduce relative variability in aggregated inventory.

Capacity constraint: A limit on the flow, volume, labor, or activity a facility can process.

Service coverage: The geographic or time-based ability to meet a specified customer service requirement.

Mixed-integer model: An optimization model containing continuous and discrete decision variables.

Scenario analysis: Evaluation of decisions under multiple plausible future conditions.

Resilience: The ability of a network to withstand, adapt to, and recover from disruption.

3PL: A third-party logistics provider performing outsourced logistics activities.

Safety stock: Inventory held to protect against demand or lead-time uncertainty.

Effective capacity: Usable capacity after allowing for variability, downtime, constraints, and operating policy.

Service-level agreement: A contract defining measurable service obligations and remedies.

Review Questions

5 Short Answer Questions

1. How does a fulfillment centre differ from a cross-dock?
2. What assumptions underlie the center-of-gravity method?
3. Why can adding facilities reduce transportation cost but increase inventory cost?
4. What is the purpose of a capacity constraint in a location model?
5. Identify two governance risks associated with logistics outsourcing.

5 Long Answer Questions

1. Explain how facility roles and network architecture should be aligned with product and customer characteristics.
2. Derive and interpret the center-of-gravity calculation, and explain why its result should not be treated as a final site decision.
3. Discuss the principal components of total logistics network cost and the interactions among them.
4. Explain how scenario analysis and resilience testing improve strategic network decisions.
5. Critically evaluate the advantages and limitations of using a mixed-integer facility-location model in practice.

3 Analytical / Application Questions

1. A firm serves three markets with volumes of 150, 250, and 100 tonnes at coordinates (10, 10), (50, 20), and (20, 50). Calculate the weighted center of gravity and identify two additional feasibility tests that should be performed.
2. Two candidate distribution centres have identical fixed cost. Site A has a weighted load-distance score of 82,000 tonne-kilometres and Site B has 75,000 tonne-kilometres, but Site B has a 10% probability of annual closure while Site A has a 2% probability. Design a scenario-based comparison and specify what further data are required before selecting a site.
3. A company is considering a 3PL for rapid e-commerce growth. Develop a decision framework that compares ownership, cost, service, resilience, data governance, labor standards, and exit flexibility.

Multiple Choice Questions with Answers

Question 1

Which method provides an initial geographic estimate by weighting demand-point coordinates by shipment volume?

A. Center-of-gravity method

B. Critical-path method

C. Economic order quantity method

D. First-in, first-out method

Correct Answer: A. Center-of-gravity method

Question 2

In the basic load-distance method, a candidate site's score is calculated as:

A. Fixed cost divided by capacity

B. The sum of flow multiplied by distance

C. Average inventory multiplied by service level

D. Demand divided by facility count

Correct Answer: B. The sum of flow multiplied by distance

Question 3

What does the constraint ∑ᵢ dᵢxᵢₖ ≤ Capₖyₖ ensure in a facility-location model?

A. Every customer is served by two facilities

B. A facility can operate without being opened

C. Assigned demand does not exceed the opened facility's capacity

D. Transportation cost is zero for local deliveries

Correct Answer: C. Assigned demand does not exceed the opened facility's capacity

Question 4

Which condition most directly supports use of a center-of-gravity calculation?

A. Transportation cost is approximately related to volume and distance

B. All facilities have unlimited capacity

C. Demand is completely random and unobservable

D. Road distances are irrelevant to delivery performance

Correct Answer: A. Transportation cost is approximately related to volume and distance

Question 5

A network has annual facility cost of $400,000, transportation cost of $650,000, handling cost of $250,000, and inventory cost of $100,000. Ignoring other terms, what is total annual cost?

A. $900,000

B. $1,000,000

C. $1,300,000

D. $1,400,000

Correct Answer: C. $1,300,000

Question 6

Which statement best describes risk pooling?

A. Separating all demand into independent local inventories

B. Combining demand so aggregate variability is relatively lower

C. Increasing the number of suppliers without changing inventory

D. Replacing safety stock with fixed production schedules

Correct Answer: B. Combining demand so aggregate variability is relatively lower

Question 7

Why can a network with the lowest normal-condition cost be strategically inferior?

A. It may have excessive geographic redundancy

B. It may perform poorly under disruption or demand change

C. It always requires more facilities

D. It cannot use transportation contracts

Correct Answer: B. It may perform poorly under disruption or demand change

Question 8

Which variable normally represents whether candidate facility k is opened in a basic mixed-integer location model?

A. A continuous shipment quantity

B. A binary variable yₖ

C. A distance coefficient dᵢₖ

D. A demand forecast dᵢ

Correct Answer: B. A binary variable yₖ

Question 9

If annual carrying-cost rate is 20% and average inventory value is $500,000, what is the approximate annual inventory carrying cost?

A. $25,000

B. $100,000

C. $250,000

D. $2,500,000

Correct Answer: B. $100,000

Question 10

Which outsourcing provision most directly improves an organization's ability to change providers?

A. A higher minimum-volume commitment

B. A restriction on performance reporting

C. A defined transition and exit process

D. Exclusive subcontracting without audit rights

Correct Answer: C. A defined transition and exit process