Logistics Network and Facility Design

Introduction

Logistics network design determines how products, information, and associated services move from suppliers to customers. It concerns the number, location, capacity, ownership, and roles of facilities such as plants, warehouses, distribution centres, cross-docks, ports, rail terminals, and last-mile depots. Unlike daily routing, network design is primarily a strategic decision: it commits capital, shapes service capability, influences emissions, and often remains difficult to reverse for many years.

A well-designed network balances apparently conflicting objectives. More facilities may reduce transport distance and improve delivery speed, but they increase inventory, labour, management, and fixed costs. Centralisation can exploit scale economies, while decentralisation can improve responsiveness and resilience. The correct design therefore depends not only on average cost, but also on uncertainty, capacity, customer-service commitments, disruption exposure, regulation, and sustainability.

This chapter develops the principal models used to design and evaluate logistics networks. It connects facility location with allocation, hub structures, gravity analysis, scenario planning, capacity evaluation, and network restructuring.

Learning objectives

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

1. Distinguish strategic, tactical, and operational network decisions.
2. Explain the cost and service trade-offs associated with facility location and allocation.
3. Formulate and interpret facility-location, transportation, hub-location, gravity, and scenario models.
4. Evaluate warehouses, distribution centres, ports, terminals, capacity, and geographic service coverage.
5. Assess network alternatives under demand uncertainty, disruption, regulatory constraints, and sustainability objectives.
6. Recommend a network restructuring approach using financial, operational, resilience, and governance criteria.

1. The architecture of a logistics network

A logistics network consists of nodes and links. Nodes include supply points, production sites, consolidation centres, ports, terminals, warehouses, fulfilment centres, and customers. Links represent transport services by road, rail, sea, air, pipeline, or multimodal combinations. Each node has a role, a capacity, a cost structure, and a service boundary.

The network may be:

• Centralised, with a small number of large facilities serving broad markets.
• Decentralised, with facilities positioned near demand.
• Multi-echelon, with flows through more than one level, such as plant → regional distribution centre → local depot → customer.
• Cross-docking oriented, where inbound shipments are rapidly sorted and transferred with limited storage.
• Hub-and-spoke, where flows are consolidated through strategically selected hubs.
• Omnichannel, combining store, warehouse, marketplace, and direct-to-consumer fulfilment.

The design problem is not simply to minimise distance. A facility that is geographically close may have poor road access, inadequate labour, high land costs, weak customs capability, or insufficient expansion potential.

Figure: A multi-echelon network linking supply, regional inventory, local depots, and customers.

2. Strategic, tactical, and operational decisions

Strategic decisions usually include facility location, network topology, ownership, mode structure, and long-term capacity. Tactical decisions include inventory positioning, allocation rules, carrier contracts, safety capacity, and seasonal use of third-party logistics providers. Operational decisions include daily order assignment, routing, dock scheduling, and shipment consolidation.

These levels interact. A strategic decision to close a warehouse may reduce fixed cost but increase tactical safety stock and operational delivery complexity. Conversely, a sophisticated routing system cannot compensate for a network whose facilities are fundamentally misplaced.

Important design criteria include:

• Total landed logistics cost: transport, handling, storage, inventory, facility, customs, taxes, and expected disruption cost.
• Service: delivery time, order-cycle reliability, coverage, availability, and flexibility.
• Capacity: throughput, storage positions, dock doors, labour, equipment, port or terminal slots, and transport availability.
• Risk: exposure to floods, earthquakes, strikes, cyberattacks, geopolitical restrictions, and supplier failure.
• Sustainability: fuel use, emissions, land use, congestion, noise, and community effects.
• Strategic fit: proximity to markets, suppliers, talent, technology ecosystems, and future growth.

3. Facility-location and allocation models

3.1 The uncapacitated facility-location model

Suppose candidate facilities are indexed by j and customer zones by i. Let yⱼ equal 1 if facility j is opened and 0 otherwise. Let xᵢⱼ be the quantity shipped from facility j to customer i. A basic mixed-integer model is:

subject to:

Here, Fⱼ is the fixed annual cost of opening facility j (currency/year), cᵢⱼ is variable transport and handling cost per unit from j to i, and Dᵢ is demand at customer zone i (units/year). The linking constraint prevents shipments from a closed facility. The model assumes known demand, divisible flow, and sufficient facility capacity. It is “uncapacitated” because no upper bound has been imposed on total throughput.

3.2 Capacitated facility location

If facility j has annual capacity Kⱼ, add:

This constraint is dimensionally consistent: both sides are units/year. Capacity may need to be expressed in pallets, cases, tonnes, orders, or handling hours; these are not interchangeable without conversion factors.

A practical model may also include minimum operating volume, labour availability, service-time restrictions, opening dates, facility closure decisions, and multiple products. For multiple products p, flow becomes xᵢⱼₚ and capacity may be measured in equivalent handling hours:

where aₚ is handling time per unit of product p (hours/unit) and Hⱼ is available annual handling time (hours/year).

3.3 Allocation and service constraints

The cost-minimising allocation may assign a customer to a distant facility if that facility is cheaper. Service constraints prevent unacceptable outcomes. For example, if customer i may only be served by facilities within a permitted set Nᵢ:

Alternatively, a maximum delivery-time constraint can be represented by allowing only arcs whose travel time tᵢⱼ is no greater than the promised limit Tᵢ. Such constraints convert a purely financial model into a service-feasible model.

3.4 A small illustrative calculation

Consider two candidate warehouses. Warehouse A has fixed cost $400,000 per year; Warehouse B has fixed cost $250,000. If all 10,000 annual units are served by A, transport and handling cost is $42/unit. If all are served by B, it is $58/unit. Ignoring inventory and service differences:

• A total cost = $400,000 + (10,000 × $42) = $820,000.
• B total cost = $250,000 + (10,000 × $58) = $830,000.

A is preferred by $10,000 per year despite its higher fixed cost. The result is sensitive to demand. The break-even volume Q solves:

Thus:

Above 9,375 units, A is cheaper; below it, B is cheaper. This is a special two-alternative comparison, not a substitute for a full network model.

4. Gravity models and geographic screening

Gravity models provide an initial estimate of a facility’s economically attractive location. A centre-of-gravity calculation using demand weights is:

Here, (xᵢ, yᵢ) are coordinates of demand points and wᵢ is a weight such as annual volume, shipment frequency, or volume multiplied by transport cost. The output (X,Y) is a screening point, not necessarily the optimal site. It assumes a continuous geographic plane, proportional distance cost, and no barriers, congestion, zoning, taxes, or discrete site availability.

A more realistic p-median model chooses p facilities to minimise demand-weighted distance. Gravity analysis is useful before detailed modelling because it identifies broad geographic regions, reveals the effect of demand concentration, and supports data-quality checks. It should not override feasibility investigation.

5. Hub location and consolidation

A hub consolidates flows and may reduce line-haul cost through scale economies. In a hub-and-spoke network, origin–destination flows are routed through one or more hubs. Hub design must account for:

• inbound and outbound handling;
• line-haul savings from consolidation;
• additional transfer distance and handling time;
• hub capacity and time windows;
• vulnerability to disruption;
• economies or diseconomies of scale.

A simplified hub cost for flow q may be represented as:

where d₁ and d₂ are inbound and outbound distances, α and β are mode-specific cost rates, and h is hub handling cost per unit. The expression is useful only when costs are approximately linear. Real networks may have fixed vehicle costs, stepwise capacity, congestion, and nonlinear service penalties.

Hubs are especially attractive when shipment volumes are geographically dispersed but sufficiently large to support frequent consolidated departures. They are less attractive when demand is urgent, highly asymmetric, or exposed to a single-point-of-failure risk.

6. Capacity, service coverage, and facility evaluation

6.1 Capacity is multidimensional

A warehouse’s nominal storage capacity does not equal its practical throughput capacity. Evaluation should include:

• storage capacity: pallets, cartons, cubic metres, or locations;
• receiving and shipping throughput: units or pallets per day;
• dock-door capacity and appointment constraints;
• picking productivity and replenishment capacity;
• labour availability by shift;
• yard and parking capacity;
• automation performance and maintenance downtime;
• peak, average, and seasonal capacity.

A useful utilisation measure is:

Effective capacity should reflect downtime, labour constraints, changeovers, safety restrictions, and peak operating conditions. Sustained 100% utilisation is generally infeasible because variability causes queues and service deterioration. The appropriate target depends on process variability and the cost of waiting.

6.2 Coverage and accessibility

Coverage may be measured as the proportion of demand reachable within a specified time:

Dᵢ is demand in zone i, tᵢ is travel time to the assigned facility, T is the service threshold, and I(·) is an indicator equal to 1 when the condition is true and 0 otherwise. This metric is demand-weighted; population-weighted and customer-count-weighted versions answer different questions.

Ports and terminals require additional measures: berth productivity, crane moves per hour, truck turnaround time, rail dwell time, customs clearance time, gate capacity, storage dwell, and vessel or train schedule reliability. A port with low ocean distance may be inferior if landside congestion creates excessive total transit time.

6.3 Total cost of ownership

Facility comparison should include more than rent or land cost. A practical annualised evaluation includes:

TLC is total logistics cost; F is annualised fixed facility cost; the remaining terms represent transport, handling, inventory, expected service failure, expected risk, and monetised externality costs. Not every organisation can estimate all terms precisely. However, explicitly identifying omitted terms prevents false precision.

7. Scenario, uncertainty, and resilience models

A single forecast is inadequate for strategic design. Scenario planning represents alternative futures such as high growth, recession, fuel-price escalation, e-commerce expansion, port disruption, carbon taxation, or a major supplier relocation.

For alternatives a and scenarios s, expected cost is:

where pₛ is the probability of scenario s, with ∑ₛ pₛ = 1, and Cₐₛ is the cost of alternative a under that scenario. Expected cost alone can favour a fragile design. Decision-makers should also examine worst-case cost, service loss, regret, and recovery time.

A robust optimisation approach may minimise the worst scenario cost:

This is conservative and may overpay for protection. A common compromise is to minimise expected cost plus a risk penalty, but the penalty must be justified rather than selected arbitrarily.

Resilience can be designed through multiple suppliers, alternative ports, postponement, strategic inventory, spare capacity, dual transport modes, and digitally visible contingency routes. Redundancy is not automatically valuable: it should be compared with the probability and consequence of disruption.

8. Network restructuring methodology

A disciplined redesign normally proceeds through the following stages.

Stage 1: Define the decision and scope

Specify the planning horizon, products, geographies, service promises, facility types, transport modes, ownership options, and constraints. Clarify whether the question is “where should facilities be opened?” or “what complete network should serve each demand point?”

Stage 2: Build and validate the baseline

Map current flows, costs, lead times, capacities, inventory, customs processes, and failure points. Reconcile shipments, sales, inventory, and transport invoices. Poor data can make an optimised model confidently wrong.

Stage 3: Generate feasible alternatives

Create candidate sites using gravity screening, market intelligence, property analysis, labour studies, infrastructure assessment, and regulatory review. Eliminate sites that fail basic access, zoning, environmental, security, or capacity requirements.

Stage 4: Model and test

Use mixed-integer optimisation for facility and allocation decisions; simulation for congestion, variability, and queueing; GIS for travel-time and coverage analysis; and scenario analysis for uncertainty. Compare alternatives using common assumptions.

Stage 5: Stress-test and value flexibility

Test demand growth, peak season, fuel cost, labour shortage, border delay, natural disaster, cyberattack, and facility outage. Consider staged investment, expandable buildings, leases rather than ownership, and temporary third-party capacity.

Stage 6: Decide and implement

The recommendation should specify facility roles, customer allocation, transport lanes, transition timing, capital, workforce, systems, permits, and exit criteria. Implementation risk can exceed model risk; a theoretically superior network may fail if transition inventory, labour recruitment, or system integration is ignored.

9. Illustrative case: Northstar Consumer Products

This is a fictional illustrative case.

Background

Northstar distributes packaged household products to 14 regional markets. It currently operates one national distribution centre near the manufacturing plant. Average annual demand is 1.2 million cases, but 30% of volume is concentrated in two distant metropolitan regions. Customers increasingly require next-day delivery, while the national facility experiences severe peak-season congestion.

Problem

Management must choose among: (1) retaining the national centre; (2) adding one eastern regional DC; or (3) adding two regional DCs. The company also faces periodic disruption at the main motorway and wants to reduce empty return kilometres.

Analysis

The baseline study found that the national-centre option has the lowest fixed cost but the highest outbound transport cost and only 78% next-day demand coverage. One eastern DC increases annual fixed cost by $1.6 million, reduces transport cost by $2.4 million, and raises coverage to 94%. Two regional DCs increase fixed cost by $3.0 million, reduce transport cost by $3.0 million, and raise coverage to 97%, but require additional inventory and management complexity.

A complete comparison must add working-capital cost, handling, systems, labour, tax, and disruption effects. Suppose incremental inventory and management cost is $0.7 million for one DC and $1.5 million for two DCs. The approximate annual benefits before resilience value are:

• One DC: $2.4m transport saving − $1.6m fixed cost − $0.7m additional cost = $0.1m net benefit.
• Two DCs: $3.0m transport saving − $3.0m fixed cost − $1.5m additional cost = −$1.5m net benefit.

The one-DC alternative is financially marginal, but it delivers a major service improvement. If the expected annual cost of motorway disruption and lost sales avoided by the regional DC exceeds $0.1 million, the investment becomes economically attractive. The analysis should also test whether a third-party overflow facility can provide peak capacity at lower commitment.

Approach and solution

Northstar selects one eastern DC, subject to a staged lease, a performance review after 18 months, and a contingency contract with a nearby third-party operator. Customers are allocated by service zone, but the national centre remains an emergency supply point. The company introduces a transport-management system, measures actual coverage by postcode, and establishes a monthly capacity and resilience review.

Outcome and learning points

The case demonstrates that the lowest fixed-cost network is not necessarily the lowest total-cost or highest-value network. Service coverage, disruption exposure, peak capacity, contractual flexibility, and implementation risk materially affect the decision. It also illustrates why a regional facility should be evaluated as a portfolio of cost and capability rather than as a transport-saving project alone.

10. Technology, governance, and contemporary developments

GIS platforms convert coordinates, road networks, travel times, demographics, and hazard layers into location evidence. Optimisation solvers evaluate discrete facility and allocation decisions. Digital twins and discrete-event simulation can test congestion, labour variability, dock queues, and disruption recovery. Machine-learning forecasts can improve demand scenarios, but forecasts remain uncertain and should not be treated as facts. AI-generated recommendations require traceability, validation, and human accountability.

Governance is essential. Location decisions can affect land use, traffic, air quality, worker safety, tax bases, and community equity. Environmental impact assessment, planning permission, customs law, labour regulation, hazardous-material rules, accessibility requirements, and data-protection obligations may constrain feasible sites. Organisations should document assumptions, protect commercially sensitive data, audit optimisation inputs, and provide decision rights for exceptions.

Contemporary design increasingly incorporates decarbonisation. Modal shift to rail or inland waterway, urban consolidation, electric vehicles, renewable-powered facilities, packaging redesign, and facility energy management can reduce emissions. However, emissions reductions must be measured across the full system; a local delivery improvement may be offset by additional replenishment kilometres or inventory transfers.

Challenges and limitations

Network models simplify reality. Common limitations include:

• uncertain demand and travel times;
• inaccurate or inconsistent master data;
• linear cost assumptions despite economies of scale;
• omitted congestion and queueing;
• difficulty valuing service reliability and brand effects;
• political, social, and environmental constraints that are hard to monetise;
• model size and computational complexity;
• behavioural resistance during implementation;
• overconfidence in a single “optimal” solution.

Sensitivity analysis is therefore indispensable. Decision-makers should vary demand, rates, capacity, service thresholds, opening costs, and scenario probabilities. A good design is often not the mathematically cheapest alternative, but one that remains acceptable across a credible range of futures.

Best-practice checklist

1. Define service requirements before optimising cost.
2. Use validated shipment, demand, inventory, and capacity data.
3. Separate strategic, tactical, and operational decisions while modelling their interactions.
4. Distinguish nominal, effective, and peak capacity.
5. Combine optimisation with GIS, simulation, and managerial judgement.
6. Include inventory, handling, risk, externality, and transition costs.
7. Test multiple scenarios and identify assumptions driving the recommendation.
8. Preserve flexibility through staged investment and qualified alternatives.
9. Evaluate resilience, safety, legality, community impact, and emissions.
10. Monitor the implemented network and redesign when demand or risk changes.

Chapter Summary

Logistics network and facility design determines the structural configuration through which supply and demand are connected. Its central trade-off is between scale efficiency and proximity, with service, capacity, inventory, resilience, sustainability, and flexibility shaping the final decision. Facility-location models select sites and allocate flows; capacitated and service-constrained variants make the analysis more realistic. Gravity models support geographic screening, while hub models examine consolidation and transfer economics. Capacity must be evaluated in storage, throughput, labour, dock, yard, and peak terms. Scenario and robust approaches prevent a network from being designed around one fragile forecast. Effective restructuring combines validated data, optimisation, GIS, simulation, governance, implementation planning, and sensitivity analysis. The strategic objective is not merely minimum transport cost, but a feasible and resilient network that delivers required service at acceptable total lifecycle cost and societal impact.

Key Terms

Facility location: Selection of sites for logistics or production facilities.

Allocation: Assignment of demand or flows to facilities.

Distribution centre: Facility that receives, stores, processes, and dispatches products.

Multi-echelon network: Network containing two or more facility levels.

Hub-and-spoke: Network in which flows are consolidated through central hubs.

Fixed facility cost: Annualised cost that does not vary directly with shipment volume.

Capacitated facility location: Location model with facility throughput limits.

Service coverage: Share of demand meeting a specified geographic or time threshold.

Centre of gravity: Demand-weighted geographic screening point.

p-median model: Model selecting a specified number of facilities to minimise weighted distance.

Cross-docking: Transfer of inbound goods to outbound vehicles with limited storage.

Throughput: Quantity processed by a facility during a specified period.

Utilisation: Actual output divided by effective available capacity.

Scenario analysis: Evaluation of alternatives under distinct possible futures.

Robust optimisation: Design approach seeking acceptable performance under uncertainty.

Resilience: Ability to withstand, adapt to, and recover from disruption.

Total logistics cost: Combined facility, transport, handling, inventory, service, risk, and related costs.

Network restructuring: Deliberate change to facility roles, locations, capacities, or flows.

Review Questions

Short Answer Questions

1. Why can a decentralised network improve service but increase total cost?
2. What is the purpose of the linking constraint in a facility-location model?
3. Why is a centre-of-gravity result only a screening recommendation?
4. Distinguish nominal capacity from effective capacity.
5. What is the principal risk of relying on a single demand forecast?

Long Answer Questions

1. Explain how fixed facility cost, transport cost, inventory, and service requirements interact in network design.
2. Compare centralised, decentralised, multi-echelon, and hub-and-spoke networks.
3. Describe a rigorous methodology for restructuring an existing logistics network.
4. Explain the assumptions and limitations of capacitated facility-location models.
5. Discuss how resilience and sustainability should be incorporated into strategic network decisions.

Analytical / Application Questions

1. A company has two candidate facilities with different fixed and variable costs. Formulate the break-even-volume calculation and explain which qualitative factors could invalidate a purely cost-based choice.
2. Design a scenario analysis for a port-dependent distribution network exposed to demand growth, port closure, and carbon-price changes. Specify the performance measures you would compare.
3. A warehouse reports 92% average utilisation but frequent peak-season delays. Diagnose possible causes and recommend capacity and network measures to improve service without assuming that a new facility is immediately required.

Multiple Choice Questions with Answers

Question 1

Which decision is primarily strategic in logistics network design?

A. Selecting a delivery route for one truck today

B. Scheduling a dock appointment

C. Choosing the location of a regional distribution centre

D. Releasing a daily picking wave

Correct Answer: C. Choosing the location of a regional distribution centre

Question 2

In a capacitated facility-location model, what does the constraint ∑ᵢ xᵢⱼ ≤ Kⱼyⱼ ensure?

A. Demand is always greater than supply

B. A closed facility cannot ship and an open facility cannot exceed capacity

C. Every customer must use every facility

D. Transport cost is fixed

Correct Answer: B. A closed facility cannot ship and an open facility cannot exceed capacity

Question 3

Using fixed cost and variable cost only, Warehouse A costs $400,000 plus $42 per unit and Warehouse B costs $250,000 plus $58 per unit. What is their break-even volume?

A. 5,625 units

B. 8,000 units

C. 9,375 units

D. 12,500 units

Correct Answer: C. 9,375 units

Question 4

What is the principal purpose of a centre-of-gravity calculation?

A. To prove a site is legally permissible

B. To provide an initial demand-weighted geographic screening point

C. To calculate warehouse labour productivity

D. To determine customs duties

Correct Answer: B. To provide an initial demand-weighted geographic screening point

Question 5

Which measure is most directly concerned with the proportion of demand reachable within a specified time?

A. Facility utilisation

B. Service coverage

C. Inventory turnover

D. Dock dwell

Correct Answer: B. Service coverage

Question 6

Why can sustained 100% facility utilisation be problematic?

A. It eliminates all fixed costs

B. Variability can create queues and service deterioration

C. It guarantees excess inventory

D. It prevents all automation

Correct Answer: B. Variability can create queues and service deterioration

Question 7

In expected-cost scenario analysis, what does pₛ represent?

A. The capacity of facility s

B. The probability assigned to scenario s

C. The price of transport mode s

D. The number of customers in zone s

Correct Answer: B. The probability assigned to scenario s

Question 8

Which is a potential disadvantage of a hub-and-spoke network?

A. It can never consolidate shipments

B. It eliminates all handling

C. It may create a single point of failure and add transfer time

D. It requires no capacity planning

Correct Answer: C. It may create a single point of failure and add transfer time

Question 9

Which cost is commonly omitted in a narrow transport-only facility comparison?

A. Transport distance

B. Line-haul rate

C. Inventory and transition cost

D. Shipment volume

Correct Answer: C. Inventory and transition cost

Question 10

Why should optimisation results be stress-tested?

A. Because all optimisation models are necessarily infeasible

B. Because demand, costs, capacity, and disruption conditions are uncertain

C. Because fixed costs cannot be represented mathematically

D. Because service requirements are irrelevant

Correct Answer: B. Because demand, costs, capacity, and disruption conditions are uncertain