Transportation Network Design

Introduction

Transportation network design determines how products, information, vehicles, facilities and transport services are arranged across a supply chain. It is a strategic decision because network choices influence customer service, inventory, transportation cost, facility investment, carbon emissions, resilience and the organisation’s ability to respond to market change. Decisions include the number and location of facilities, the assignment of customers to facilities, shipment frequency, vehicle and mode selection, consolidation policy, and the use of hubs or cross-docks.

A network is not simply a map of routes. It is an integrated system of nodes and links. Nodes may include suppliers, plants, distribution centres, consolidation terminals, cross-docks, ports, rail terminals and customers. Links represent transport services with specified costs, capacities, transit times, reliability and environmental impacts.

The central design problem is to obtain an acceptable service level at minimum relevant total cost, while recognising uncertainty and operational constraints.

Learning Objectives

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

1. Distinguish direct-shipment, hub-and-spoke, consolidation, cross-docking and multimodal networks.
2. Formulate facility-location and flow-allocation models.
3. Explain the trade-off between transport, facility, inventory, service and risk costs.
4. Apply a basic network-design model to a practical situation.
5. Evaluate network alternatives using financial, operational, environmental and resilience criteria.
6. Identify the governance, technology and implementation issues associated with network redesign.

1. Network Design as a Strategic Decision

Network design is generally conducted at three levels:

The levels are interdependent. A strategic decision to close a warehouse may reduce fixed cost but increase delivery distance and inventory. Conversely, a strategically located cross-dock may raise handling requirements while reducing storage and transport costs.

A useful objective is:

where:

• TC = relevant total cost per planning period;
• F = facility opening, operating and fixed-capacity cost;
• T = line-haul, local delivery and accessorial transport cost;
• H = handling, sorting and transfer cost;
• I = inventory carrying cost, including cycle, safety and pipeline inventory;
• P = penalty cost for late, incomplete or failed deliveries;
• R = expected cost of disruption and recovery;
• E = monetised environmental cost, where the organisation uses such accounting.

The equation is a cost-accounting framework rather than a universal law. It is applicable when cost categories are measured over the same time period and double counting is avoided. For example, a carrier invoice should not be included both in transport cost and facility handling cost unless the invoice has been correctly disaggregated.

Cost minimisation alone is insufficient. A design that minimises expected cost may be unacceptable if it cannot meet contractual delivery windows or has excessive exposure to a single port, carrier or geographic region.

2. Principal Network Architectures

2.1 Direct shipment

In a direct-shipment network, each origin ships directly to its assigned destination or customer. It avoids intermediate handling and usually provides short physical lead time. It is attractive when shipment volumes are large, destinations are concentrated, products are time-sensitive, or full truckload or full container utilisation is possible.

Its disadvantages are low vehicle utilisation for small orders, many origin–destination lanes and limited pooling of demand. Direct shipment may also create complex carrier management when thousands of customer locations are served.

2.2 Hub-and-spoke

A hub-and-spoke network routes flows through one or more hubs. Shipments from several origins are consolidated at a hub and then distributed to destinations. The hub enables scale economies, standardised sorting and central control.

If there are n origins and m destinations, a simple direct structure could require up to nm origin–destination relationships. A single-hub structure requires approximately n + m hub connections, although actual links depend on the operating design. The reduction in links can simplify management, but the hub adds handling, transfer time and a potential bottleneck.

Hub-and-spoke designs are effective where flows are fragmented and network density is high. They are less attractive where the hub creates excessive detour distance or where customers require highly time-sensitive service.

2.3 Consolidation

Consolidation combines compatible shipments before line-haul movement. Compatibility may be based on destination, delivery window, product class, temperature, security requirements or legal restrictions. Consolidation reduces cost per unit when the saving from higher vehicle utilisation exceeds additional handling, waiting and coordination cost.

A consolidation decision should consider the total cost of delay. A low freight rate is not economical if waiting causes stock-outs, production interruption or customer penalties.

2.4 Cross-docking

Cross-docking transfers inbound goods directly to outbound vehicles with little or no storage. Inbound loads are received, identified, sorted and assigned to outbound departures. It is appropriate for fast-moving products, predictable demand, retail replenishment and time-sensitive flows.

Cross-docking is not merely a warehouse with less storage. It requires synchronised schedules, accurate advance shipment information, compatible packaging, reliable suppliers and disciplined appointment management. Poor synchronisation produces congestion, missed departures and emergency transport.

2.5 Multimodal and intermodal networks

A multimodal network uses two or more transport modes within a supply chain. In common usage, intermodal transport often means movement in the same loading unit across modes, whereas multimodal may involve a broader contractual or operational combination. The precise legal meaning varies by jurisdiction and contract.

Road offers accessibility and flexibility; rail offers efficient long-haul movement where terminals and volumes support it; maritime transport offers low unit cost for large international flows but usually longer and more variable transit; air provides speed at high cost and environmental intensity. A multimodal design must include terminal handling, drayage, transshipment time, schedule reliability, documentation and mode-specific capacity.

2.6 Comparative assessment

3. Facility Location and Flow Allocation

The classical facility-location problem decides which candidate facilities to open and how demand should be assigned. A basic capacitated model uses binary and continuous variables.

Let:

• i index candidate facilities and j index customers;
• f_i be the fixed cost of opening facility i, in currency per period;
• c_ij be variable transport and service cost per unit from i to j;
• d_j be demand at customer j, in units per period;
• u_i be capacity of facility i, in units per period;
• y_i = 1 if facility i is opened, otherwise 0;
• x_ij be quantity shipped from i to j.

The capacitated facility-location model is:

subject to:

The first constraint requires all demand to be served. The second prevents shipments from a closed facility and limits an open facility to its capacity. The model assumes known demand, known costs, divisible flow and a single planning period. It does not automatically represent delivery-time constraints, multiple commodities, inventory, uncertainty, route sequencing or economies of scale in transport. Those features require extensions.

If each customer must be served by exactly one facility, a binary assignment variable z_ij can be used with ∑ᵢz_ij=1. If split deliveries are allowed, continuous flow allocation is more appropriate. This distinction is managerial as well as mathematical: split allocation can reduce cost but may complicate customer service and billing.

Total-cost trade-offs

Opening more facilities generally reduces outbound distance and response time but increases fixed facility cost, duplicated safety stock and management complexity. A simplified inventory relationship is:

For a constant demand rate, pipeline inventory is approximately:

The relationship applies when demand rate and transit time are expressed in compatible units. For example, units per day multiplied by days gives units. Reducing transit time can therefore lower pipeline inventory, although safety stock may increase if reliability deteriorates.

A practical design compares alternatives rather than assuming that one cost category determines the answer. Relevant measures include total landed cost, order-to-delivery time, on-time-in-full performance, capacity utilisation, emissions, exposure to disruption and scalability.

4. A Worked Facility-Location Example

A firm must serve two customer zones with demands of 60 and 40 units per day. It can open Facility A or Facility B, but not both. Each facility has capacity of 100 units per day. The fixed daily cost is 420 for A and 360 for B. Unit delivery costs are:

Demand must be fully served and splitting is not necessary because each facility can serve the entire demand.

If A is opened:

If B is opened:

Facility A is preferred under the stated costs, even though its fixed cost is higher. The decision is driven by the combined allocation cost. This conclusion changes if capacities, service requirements, risk costs or demand forecasts change. For example, if Zone 1 requires delivery within a short time that only B can provide, the feasible set changes and B may become the correct choice.

5. Network-Design Methodology

A robust design process proceeds through the following stages.

Step 1: Define the decision scope

Specify products, customers, time horizon, service promises, geographic boundaries, modes, facilities and constraints. Clarify whether the decision concerns normal operations, peak season, disruption planning or all three.

Step 2: Build a clean demand and cost baseline

Demand should be expressed by location, product family and time period. Cost data should distinguish fixed, variable, accessorial, handling, inventory and penalty components. Geographic data should use consistent coordinates and distance calculations. Historical freight invoices should be reconciled with shipment records.

Step 3: Generate feasible alternatives

Alternatives may include direct shipment, a regional distribution-centre network, a national hub, cross-docking, supplier consolidation, rail–road combinations or outsourced logistics. A model should not be allowed to select an operationally impossible alternative merely because it has a low nominal cost.

Step 4: Optimise allocation and capacity

Use linear programming, mixed-integer programming, network-flow models, simulation or a combination. Optimisation identifies attractive structures; simulation tests their behaviour under variability, queues and disruptions.

Step 5: Conduct sensitivity and scenario analysis

Important scenarios include demand growth, fuel-price changes, wage inflation, facility failure, port closure, border delay, carrier capacity shortage and changes in customer service requirements. Sensitivity analysis asks how much a parameter must change before the preferred design changes.

Step 6: Evaluate implementation feasibility

Examine land, labour, permits, information systems, carrier contracts, inventory transition, employee consultation, customer communication and business continuity. A mathematically optimal design may fail if it cannot be implemented within the organisation’s change capacity.

Step 7: Pilot, measure and govern

Pilot selected lanes or customer segments. Establish ownership of master data, model assumptions, service metrics and exception procedures. Revisit the design when demand, regulation, technology or risk exposure changes.

6. Applied Case: Northstar Home Appliances

The following is an illustrative case created for educational purposes.

Background

Northstar sells appliances from two plants to 180 retailers. Demand is seasonal, products are bulky, and retailer orders are relatively small. The existing system uses direct deliveries from both plants. Vehicle fill is low outside peak season, and deliveries are frequently late in remote regions.

Problem

Management is considering three alternatives:

1. retain direct shipment;
2. establish one national cross-dock near the main market;
3. establish two regional distribution centres and use rail for the long-haul leg where feasible.

The decision must consider annual transport cost, facility cost, inventory, delivery performance, emissions and disruption exposure.

Analysis

The direct system has minimal handling but creates many low-density lanes. The cross-dock can combine plant output and sequence retailer deliveries, but its success depends on synchronised inbound schedules. The two-centre design reduces final-mile distance and improves regional responsiveness, but duplicates inventory and fixed operating cost. Rail can reduce line-haul emissions and cost for predictable volume, but road drayage is still required at both ends.

Northstar constructs a mixed-integer model with facility-opening variables, plant-to-facility flows, facility-to-retailer flows and mode choices. It includes capacity limits and a service constraint requiring at least 95% of retailer demand to arrive within the promised window. It then tests peak demand and a scenario in which one major corridor is unavailable for ten days.

Approach and solution

The analysis indicates that the national cross-dock has the lowest expected cost under average demand but becomes congested during the peak season. The two-centre design costs more in the average case but provides better service and lower disruption exposure. Northstar therefore adopts a phased regional design: one centre is opened initially, a contracted overflow cross-dock is retained for peak periods, and a rail–road lane is introduced for stable long-haul volume.

The implementation includes barcode scanning at receipt and dispatch, appointment scheduling, retailer delivery-window data, and a monthly capacity review. The company also establishes a contingency carrier panel and minimum alternative-route requirements.

Outcome and learning points

The case illustrates that network design is a portfolio decision rather than a search for the cheapest average freight rate. Capacity, seasonality, service, resilience and implementation risk materially affect the preferred architecture. It also demonstrates the value of combining optimisation with scenario analysis and staged implementation.

7. Technology, Data and Artificial Intelligence

Transportation-management systems can combine orders, rates, carrier capacity, GPS positions, appointment schedules and proof-of-delivery data. Geographic information systems support spatial analysis of demand, accessibility, congestion and facility catchments. Digital twins and discrete-event simulation can represent queues, dock doors, loading patterns and disruption scenarios more realistically than a static optimisation model.

Artificial intelligence can assist with demand forecasting, ETA prediction, anomaly detection, dynamic consolidation and carrier selection. However, AI does not remove the need for sound data, causal reasoning and governance. A model trained on historical allocations may reproduce past service inequalities or systematically favour easily served customers. Human review is needed for unusual events, contractual commitments, safety decisions and high-impact customer consequences.

Model governance should document data sources, assumptions, objective functions, constraints, version changes, validation results and override authority. Decisions should be explainable enough for procurement, finance, operations, regulators and affected customers to understand.

8. Sustainability, Ethics and Regulation

Network design affects emissions through distance, mode, vehicle utilisation, empty running, energy source and facility operations. A basic activity-based estimate is:

where l indexes transport legs, tonnes are the mass moved, distance is measured consistently, and the emission factor is expressed as carbon-dioxide equivalent per tonne-distance. The estimate is only as reliable as its boundaries and factors. Organisations should state whether they include well-to-wheel or tank-to-wheel emissions, subcontractor activity, empty returns and terminal operations.

Ethical design requires attention to worker safety, reasonable delivery schedules, fair treatment of carriers, community impacts, noise, congestion and exposure to air pollution. Regulatory requirements may include dangerous-goods rules, vehicle weight and driving-time limits, customs documentation, data protection, competition law, accessibility rules and environmental reporting. Compliance should be treated as a design constraint, not a post-design administrative task.

9. Challenges and Limitations

First, demand and travel time are uncertain. A deterministic model can produce a fragile design if uncertainty is ignored. Stochastic programming, robust optimisation, safety capacity and scenario analysis can improve decisions, although they require more data and computational effort.

Second, costs may be nonlinear. Full-truckload rates, quantity discounts, congestion, labour shifts and facility step-capacities violate simple linear assumptions. Piecewise-linear or integer formulations may be needed.

Third, model boundaries can distort results. Excluding inventory, returns, claims, taxes, carbon cost or transition expense may make a network appear cheaper than it really is.

Fourth, optimisation may conflict with organisational incentives. A transport department may prefer low freight cost while customer service values responsiveness and finance values low fixed assets. A cross-functional governance process is therefore essential.

Finally, a network is dynamic. E-commerce growth, geopolitical change, climate events, automation, urban restrictions and new transport services can alter the value of each architecture. Periodic redesign should be linked to trigger indicators rather than conducted only after performance has deteriorated.

10. Best-Practice Checklist

• Define service requirements before optimising cost.
• Use total landed cost rather than freight rate alone.
• Separate fixed, variable, handling, inventory, penalty and risk costs.
• Validate demand, distance, capacity and rate data.
• Model peak periods and realistic disruptions.
• Test split-delivery assumptions explicitly.
• Include implementation and transition costs.
• Compare carbon, safety, labour and community impacts.
• Use optimisation for structure and simulation for variability.
• Establish model governance and review triggers.
• Pilot high-risk changes before full deployment.

Chapter Summary

Transportation network design selects the structure, facilities, modes, flows and operating policies that connect supply with demand. Direct shipment minimises handovers; hubs and consolidation exploit pooling; cross-docking reduces storage when schedules are synchronised; and multimodal networks combine the capabilities of different modes. Facility-location and flow-allocation models formalise opening, capacity and assignment decisions, but their usefulness depends on valid data and realistic assumptions. Total cost must include facilities, transport, handling, inventory, service penalties, disruption and, where appropriate, environmental impacts. Effective practice combines optimisation, scenario analysis, simulation, professional judgement and staged implementation. Resilience, sustainability, ethics, regulation and AI governance are now integral to network design rather than optional additions.

Key Terms

Facility location: Selection of sites for plants, warehouses, hubs or terminals.

Flow allocation: Assignment of quantities between origins, facilities and destinations.

Direct shipment: Movement from origin to destination without an intermediate facility.

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

Consolidation: Combining shipments to improve utilisation or reduce unit cost.

Cross-docking: Receiving, sorting and dispatching goods with minimal storage.

Multimodal transport: Use of two or more transport modes in one supply chain movement.

Intermodal transport: Movement using multiple modes, often in the same loading unit.

Facility fixed cost: Cost incurred for opening or operating capacity, largely independent of short-term volume.

Variable transport cost: Cost that changes with quantity, distance, mode or service.

Capacity constraint: Limit on the quantity a facility, link or mode can process.

Service level: Degree to which promised delivery or fulfilment requirements are achieved.

Pipeline inventory: Inventory physically moving through the supply chain.

Safety stock: Inventory held to protect against demand or supply uncertainty.

Mixed-integer programming: Optimisation involving continuous variables and discrete decisions.

Sensitivity analysis: Examination of how results change when assumptions vary.

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

Total landed cost: Combined cost of acquiring, moving, handling and delivering a product.

Model governance: Controls for assumptions, data, validation, use and revision of analytical models.

Emission factor: Quantity of greenhouse-gas emissions per unit of transport activity.

Review Questions

Short Answer Questions

1. What is the principal difference between direct shipment and hub-and-spoke design?
2. Why does opening additional facilities not always reduce total cost?
3. Define cross-docking and identify one condition necessary for its success.
4. What do the variables xᵢⱼ and yᵢ represent in the facility-location model?
5. Why should transport network design include scenario analysis?

Long Answer Questions

1. Compare direct shipment, hub-and-spoke, consolidation and cross-docking networks in terms of cost, service and operational requirements.
2. Explain the capacitated facility-location model, including its objective, constraints, assumptions and limitations.
3. Discuss how inventory, facility, transport and disruption costs interact in network design.
4. Explain how multimodal transport can improve a network and identify the operational risks it introduces.
5. Develop a professional methodology for redesigning a national distribution network.

Analytical / Application Questions

1. A company serves three customer regions from two candidate facilities. Design the variables, objective and constraints required to evaluate which facilities should open, and state what additional data are needed.
2. A low-cost single-hub solution fails during peak periods. Analyse at least three alternative responses and explain the trade-offs among cost, service and resilience.
3. A firm is considering replacing part of a road movement with rail. Specify the cost, service, emissions and risk measures that should be included before making the decision.

Multiple Choice Questions with Answers

Question 1

Which network architecture routes flows through an intermediate central facility?

A. Direct shipment

B. Hub-and-spoke

C. Point-to-point parcel delivery

D. Single-mode procurement

Correct Answer: B. Hub-and-spoke

Question 2

In the capacitated facility-location model, what does yᵢ = 1 usually indicate?

A. Customer j is fully served

B. Facility i is opened

C. Vehicle i is empty

D. Demand i is zero

Correct Answer: B. Facility i is opened

Question 3

A facility has a fixed daily cost of 300 and serves 50 units at a variable cost of 4 per unit. What is its total daily cost?

A. 350

B. 500

C. 1,200

D. 1,500

Correct Answer: B. 500

Question 4

Which condition is most important for successful cross-docking?

A. Large quantities of long-term storage

B. Highly synchronised inbound and outbound schedules

C. Elimination of shipment information

D. Maximum inventory at the terminal

Correct Answer: B. Highly synchronised inbound and outbound schedules

Question 5

Which cost is most directly associated with goods moving through the supply chain during transit?

A. Pipeline inventory cost

B. Facility depreciation only

C. Advertising cost

D. Product development cost

Correct Answer: A. Pipeline inventory cost

Question 6

What is the main purpose of consolidation?

A. To reduce vehicle utilisation

B. To combine compatible shipments and improve load utilisation

C. To eliminate all handling

D. To guarantee zero transit time

Correct Answer: B. To combine compatible shipments and improve load utilisation

Question 7

In a facility-location model, the constraint ∑ⱼ xᵢⱼ ≤ uᵢyᵢ primarily ensures that:

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

B. Every customer receives twice its demand

C. All facilities must open

D. Transport cost is zero

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

Question 8

Which transport mode generally offers high accessibility and flexibility for final delivery?

A. Road

B. Ocean only

C. Pipeline only

D. Rail terminal only

Correct Answer: A. Road

Question 9

Why is scenario analysis useful in transportation network design?

A. It removes the need for data validation

B. It tests performance under changing demand and disruption conditions

C. It guarantees the mathematically cheapest outcome

D. It replaces all managerial judgement

Correct Answer: B. It tests performance under changing demand and disruption conditions

Question 10

In the activity-based emissions expression CO₂e = ∑ₗ(tonnesₗ × distanceₗ × emission factorₗ), the emission factor must be expressed per:

A. Facility opened

B. Unit of transport activity such as tonne-distance

C. Customer account only

D. Calendar year regardless of movement

Correct Answer: B. Unit of transport activity such as tonne-distance