Transport Networks and Service Design

Introduction

Transport network design is the process of configuring facilities, routes, schedules, vehicles, service promises, and operating rules so that freight moves between origins and destinations at an acceptable total cost and service level. It is a strategic and operational problem: strategic decisions determine the location and role of terminals, while operational decisions determine frequency, routing, consolidation, dispatching, and exception handling.

A network is not optimal merely because it minimizes transport cost. A low-cost design may increase transit-time variability, inventory, missed deliveries, emissions, congestion, or customer dissatisfaction. Conversely, a highly responsive network may be economically unsustainable. Graduate-level analysis therefore evaluates transport as a portfolio of trade-offs involving cost, time, reliability, capacity, risk, resilience, and environmental performance.

This chapter develops the principal network forms—direct, hub-and-spoke, milk-run, cross-docking, relay, and consolidation—and presents methods for terminal location, service-frequency planning, mode choice, time–cost analysis, and last-mile design.

Learning Objectives

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

1. Distinguish major transport-network architectures and select an appropriate design for a given demand pattern.
2. Formulate basic location, routing, consolidation, and service-frequency decisions.
3. Evaluate transport modes using generalized cost, time, reliability, capacity, and risk.
4. Analyze the effect of terminals and handling points on total landed cost and lead time.
5. Design last-mile services that account for delivery density, customer windows, urban restrictions, and failed-delivery risk.
6. Use scenario analysis and optimization outputs critically rather than treating a model solution as automatically managerial best practice.

1. The Network as a Service System

A transport network consists of:

• Nodes: suppliers, plants, distribution centres, cross-docks, ports, rail terminals, parcel depots, pickup points, and customers.
• Links: road, rail, maritime, air, pipeline, or multimodal connections.
• Flows: physical shipments, information, empty equipment, and returns.
• Resources: vehicles, drivers, docks, handling equipment, containers, and terminal capacity.
• Rules: schedules, consolidation policies, priority classes, routing logic, access restrictions, and service guarantees.

Transport service has several dimensions:

A useful design principle is to separate physical architecture from service design. Two networks may use the same terminals and modes but provide different services because they differ in dispatch frequency, cut-off times, delivery windows, priority rules, or shipment eligibility.

Figure: Direct transport uses origin–destination links, whereas hub-and-spoke transport routes flows through an intermediate hub.

2. Principal Network Architectures

2.1 Direct transport

In a direct network, a vehicle travels from an origin to a destination without intermediate handling. It is attractive when shipment volumes are large, destinations are concentrated, the required service is time-sensitive, or product handling is costly or risky.

The principal advantages are short physical handling chains, low transfer risk, simple visibility, and potentially short transit time. The disadvantages are low utilization when demand is fragmented, limited departure frequency, and a large number of origin–destination links. If there are n origins and m destinations, a fully connected directional network may require up to n × m origin–destination relationships. In practice, not all links operate daily, but the coordination burden can still be substantial.

2.2 Hub-and-spoke

Hub-and-spoke networks consolidate shipments at a central or regional hub. Freight moves from origins to the hub and then to destinations. The architecture reduces the number of line-haul connections and can improve vehicle utilization, but it introduces handling, sorting, hub capacity constraints, and additional time.

Hubs are effective when demand is geographically dispersed but sufficiently dense in aggregate. They are less attractive when origins and destinations are close, shipment volumes are large enough for direct service, or service requirements prohibit transfer. Hub design should consider throughput, dock-door capacity, sorting technology, labor, congestion, security, and the consequences of a hub failure.

2.3 Milk-run networks

A milk run is a planned route that serves multiple suppliers or customers in one tour. In an inbound milk run, one vehicle collects materials from several suppliers; in an outbound milk run, it delivers to several customers. The design seeks to increase vehicle utilization and synchronize collections or deliveries with production and customer windows.

The problem resembles a vehicle-routing problem with time windows. A route must satisfy vehicle capacity, maximum working time, customer requirements, and legal driving limits. Milk runs are particularly valuable where individual shipments are small and geographically clustered. They may be unsuitable where demand is highly variable, stops are widely dispersed, or delivery windows overlap too narrowly.

2.4 Cross-docking

Cross-docking transfers freight from inbound vehicles to outbound vehicles with little or no storage. In a pre-distribution design, goods are already assigned to customers before arriving at the cross-dock. In a post-distribution design, goods are sorted and assigned after arrival.

Cross-docking can reduce storage, inventory dwell time, and order-cycle time. It requires accurate advance shipping notices, synchronized arrival and departure schedules, compatible packaging, fast scanning, sufficient staging space, and disciplined exception procedures. It does not eliminate inventory or handling; it changes where and how handling occurs.

2.5 Relay operations

In relay transport, a shipment or trailer continues over a long route while drivers or tractors are exchanged at designated points. Relay can increase asset utilization, reduce driver waiting, and support near-continuous movement. It depends on standardized equipment, secure relay yards, accurate schedules, driver availability, and regulatory compliance.

Relay is distinct from transshipment: the cargo may remain in the same trailer while the operating crew or tractor changes. Its economic value is greatest on long corridors with predictable volumes and strict transit-time requirements.

2.6 Consolidation and deconsolidation

Consolidation combines several small shipments into a larger movement. Deconsolidation separates the movement near final destinations. Consolidation may reduce line-haul cost per unit, but it introduces waiting time and handling. The decision depends on shipment size, density, service commitment, and the cost of delay.

A common practical design is: supplier pickup → consolidation terminal → full truckload or intermodal line haul → destination deconsolidation → local delivery. Consolidation should be evaluated against direct transport using total cost and not merely the line-haul rate.

3. Terminal Location and Network Configuration

Terminal location is a strategic decision because it affects transport distance, access to demand, labor, land, congestion, resilience, and future expansion. A location model normally includes fixed facility cost, variable transport cost, handling cost, capacity, and service constraints.

A simplified capacitated facility-location model can be expressed as:

Subject to:

Here, Fᵢ is the fixed cost of opening facility i per planning period; yᵢ equals 1 if facility i is opened and 0 otherwise; cᵢⱼ is the variable cost of serving one unit of demand at customer j from facility i; xᵢⱼ is the quantity assigned; dⱼ is customer demand; and Kᵢ is facility capacity. The model assumes known demand, linear variable cost, divisible flows, and no explicit route sequencing. It is a location-allocation model, not a complete vehicle-routing model.

Real applications extend the model with:

• multiple products and temperature regimes;
• inbound and outbound transport costs;
• maximum delivery-time constraints;
• labor and dock capacity;
• carbon costs and emissions limits;
• disruption scenarios;
• inventory positioning and safety stock;
• fixed-charge transportation and economies of scale.

The centre-of-gravity method can provide an initial geographic estimate, but it should not be treated as a final location decision. With customer coordinates (xⱼ, yⱼ) and weights wⱼ, the weighted centre is:

The weights may represent demand, tonne-kilometres, or another relevant measure. The method assumes a continuous plane and generally ignores roads, terrain, land prices, congestion, legal constraints, and service territories. It is therefore a screening tool rather than an optimization result.

4. Service Frequency, Time, and Inventory

Frequency affects both customer responsiveness and cost. If departures are evenly spaced every H hours and demand arrives uniformly, the expected waiting time before the next departure is approximately:

This result assumes uniformly distributed demand, fixed schedules, and no missed cut-off. If demand is lumpy or customers place orders immediately before cut-off, actual waiting time may differ substantially.

A higher frequency usually requires more vehicle departures, terminal labor, and possibly lower load factors. A lower frequency increases waiting time and may require inventory or emergency transport. A useful total-cost representation is:

where f is departures per planning period, C₀ is frequency-independent cost, C₁f represents costs that increase with departures, and C₂ ÷ f represents waiting or inventory-related cost that falls as frequency increases. This is a stylized planning function, not a universal law. If C₁ > 0 and C₂ > 0, the continuous minimum occurs at:

The result must be rounded to feasible integer frequencies and checked against capacity, schedules, labor, and customer requirements.

Transport lead time should be decomposed rather than treated as a single number:

Each term has a different management lever. For example, reducing line-haul time may have little benefit if terminal waiting or failed delivery dominates total lead time.

5. Mode Choice and Generalized Cost

Mode selection should compare the complete logistics consequence of each alternative. Relevant criteria include freight rate, fuel exposure, transit time, reliability, damage risk, shipment size, network coverage, emissions, regulatory restrictions, and flexibility.

A generalized cost model can be written as:

where GCₘ is generalized cost for mode m; Rₘ is transport rate; Hₘ is handling and terminal cost; Iₘ is inventory and time-related cost; Bₘ is expected disruption or failure cost; Eₘ is monetized environmental cost where appropriate; and Qₘ is the monetary equivalent of quality or service penalties. All terms must be expressed in the same currency and planning basis.

Time-related inventory cost can be approximated as:

where V is the value of goods in transit, r is the annual carrying-cost rate expressed as a decimal, and Tₘ is transit time in days. This estimate assumes a simple proportional carrying cost and excludes pipeline inventory variability and stockout consequences.

For example, goods worth $120,000 are in transit for 3 days, with an annual carrying-cost rate of 18%. The time-related cost is:

If air transport costs $5,800 and rail costs $3,900, while air takes 3 days and rail takes 9 days, the rate difference is $1,900. The six-day inventory difference is only approximately $355.07 at the stated inventory rate. Unless reliability, stockout avoidance, or another service factor justifies the premium, the inventory value alone does not support air transport.

However, averages can mislead. A mode with a mean transit time of 4 days but high variance may be less useful than a 5-day mode with reliable delivery. Decision makers should therefore examine percentile transit time, on-time performance, failure probability, and the cost of deviation from the customer promise.

6. Last-Mile Service Design

The last mile is often the most expensive and operationally difficult segment because deliveries are fragmented, stop density varies, access is constrained, and customers may impose narrow time windows. Last-mile design involves four linked decisions:

1. Service promise: same-day, next-day, scheduled, evening, pickup-point, or economy service.
2. Fulfilment position: central depot, urban micro-fulfilment site, store, locker, or direct ship point.
3. Route and resource design: vehicle type, driver shifts, route sequence, and delivery territory.
4. Failure management: reattempt, locker diversion, customer pickup, safe place, or return-to-depot policy.

A basic delivery-density measure is:

where N is the number of stops and L is route distance. Higher density can reduce distance per stop, although congestion and service time may offset this advantage.

A route with N stops can be approximated by:

where T_drive is driving time, sₖ is service time at stop k, and T_break is required break or waiting time. The route is feasible only if it respects vehicle capacity, driver-hour rules, road restrictions, and customer time windows.

Last-mile sustainability is not achieved simply by replacing a diesel van with an electric vehicle. The design should consider route length, payload, charging availability, battery reserve, weather, vehicle utilization, and the emissions associated with electricity production. Consolidated pickup points and lockers can reduce failed deliveries and repeated residential travel, but they may shift travel to customers and reduce convenience.

7. Optimization and Scenario Analysis

Transport network models commonly use mixed-integer programming, vehicle-routing algorithms, simulation, and multi-criteria analysis.

A vehicle-routing formulation generally uses binary variable xᵢⱼᵏ, equal to 1 when vehicle k travels from node i to node j, and 0 otherwise. A simplified objective is:

Typical constraints require each customer to be visited, vehicle flow to be conserved, vehicle capacity not to be exceeded, and subtours to be eliminated. Time-window constraints add arrival-time variables and ensure that service occurs within permitted intervals. Exact optimization becomes computationally difficult as the number of customers grows, so practical systems use heuristics, metaheuristics, decomposition, and rolling-horizon reoptimization.

Optimization outputs should be interpreted through scenarios rather than accepted as a single forecast. Useful scenarios include:

• demand growth or decline;
• peak-season volume;
• fuel-price changes;
• labor shortages;
• terminal closure;
• border or port delay;
• customer migration to time-definite service;
• new emissions regulation;
• electric-vehicle range limitations.

For each scenario, managers should compare cost, service, utilization, emissions, capacity slack, and recovery time. Sensitivity analysis asks how much a parameter must change before the preferred network changes. This identifies fragile assumptions and the value of additional information.

8. Illustrative Case: Northstar Components

Background

Northstar Components supplies repair parts from two plants to 36 industrial customers across a country. Customer orders are small, demand is uneven, and 70% of customers are concentrated in three metropolitan regions. The current design uses direct less-than-truckload shipments from each plant. It provides acceptable average transit time but has low trailer utilization and high handling charges.

Problem

Management is considering a regional hub near the central metropolitan region. The hub would receive consolidated line-haul shipments, cross-dock urgent parts, and dispatch scheduled local routes. The proposed design must preserve a next-business-day promise for priority parts and a two-day promise for standard parts.

Analysis

The project team compared three alternatives:

The team first estimated flows by customer region, then tested facility capacity, dock throughput, line-haul schedules, and last-mile routes. It separated priority and standard shipments rather than imposing one service on all products. A disruption scenario closed the central hub for two days. The two-hub design was more resilient, but its annual fixed cost was materially higher.

The base-case model favored the central cross-dock because consolidation reduced line-haul cost and improved vehicle utilization. However, the scenario analysis showed that the solution became unattractive if priority demand exceeded the forecast or if the hub’s outbound dock capacity fell below the peak requirement. The final recommendation was therefore conditional: adopt the central cross-dock, retain a qualified direct-shipping option for priority emergencies, and contract backup cross-dock capacity in the southern region.

Outcome and learning points

The case illustrates five principles. First, network design must distinguish service classes. Second, cross-docking succeeds only when schedules and information are synchronized. Third, a lower expected cost can conceal unacceptable disruption exposure. Fourth, capacity slack has economic value. Fifth, resilience may be purchased through optionality rather than through duplicating every facility.

9. Governance, Regulation, and Technology

Transport networks operate within legal and ethical boundaries. Important issues include driver hours, vehicle weight and dimensions, dangerous-goods rules, customs and security, data protection, accessibility, labor standards, and environmental requirements. Outsourcing does not remove the shipper’s responsibility to select competent providers and monitor compliance.

Network decisions can also create distributional effects. Delivery restrictions may reduce urban congestion but disadvantage customers without private vehicles. Algorithmic routing may systematically allocate undesirable work to particular driver groups or neighborhoods. Data-driven service design should therefore be audited for privacy, fairness, and explainability.

Digital technologies support, but do not replace, managerial judgment. Transport-management systems integrate orders, rates, tenders, tracking, and settlement. Telematics provide location, driving, temperature, and vehicle-health data. Digital twins and simulation can test network changes before implementation. Artificial intelligence can forecast demand, predict arrival times, detect anomalies, and propose routes. Its outputs remain dependent on data quality, model drift, constraint accuracy, and human oversight. A route recommendation that ignores legal driving limits or accessibility requirements is not operationally intelligent.

10. Best-Practice Design Process

A robust network-design process follows these steps:

1. Define the service promise, products, customers, and planning horizon.
2. Map demand, shipment size, density, time windows, and variability.
3. Identify feasible modes, terminals, carriers, and regulatory constraints.
4. Develop a small set of architectures rather than optimizing one preconceived design.
5. Estimate total cost, including handling, inventory, failure, carbon, and fixed facility cost.
6. Test capacity, peak demand, disruptions, and implementation constraints.
7. Validate results with operations staff, carriers, customers, and finance.
8. Pilot the design in a limited geography or service class.
9. Monitor cost, on-time performance, utilization, damage, emissions, and customer outcomes.
10. Reoptimize as demand, regulations, technology, and service expectations change.

The most important professional discipline is to make assumptions explicit. A model should state whether demand is deterministic or stochastic, whether capacity is daily or hourly, whether costs include empty repositioning, and whether service penalties are monetized. Without this transparency, numerical precision can create false confidence.

Chapter Summary

Transport network design configures nodes, links, resources, flows, schedules, and service rules. Direct networks minimize handling but may suffer from low utilization; hub-and-spoke networks consolidate dispersed flows but create transfer and dependency risks. Milk runs serve clustered suppliers or customers through planned tours. Cross-docking reduces dwell time when inbound and outbound operations are synchronized. Relay networks extend operating hours through tractor or driver exchange, while consolidation combines small shipments to achieve scale.

Terminal location should be evaluated through fixed cost, variable transport cost, capacity, access, service, resilience, and environmental impact. Frequency creates a trade-off between departure cost and waiting or inventory cost. Mode choice should use generalized cost rather than freight rate alone. Last-mile design must account for density, time windows, vehicle constraints, failed delivery, and urban conditions. Optimization models support location, allocation, routing, and scheduling decisions, but scenario analysis is essential for understanding uncertainty and resilience. Effective network design combines quantitative analysis with operational validation, governance, ethical judgment, and disciplined implementation.

Key Terms

Direct transport: Movement from origin to destination without an intermediate transfer.

Hub-and-spoke: Network in which flows are consolidated through a central or regional hub.

Milk run: Planned vehicle tour serving several suppliers or customers.

Cross-docking: Transfer of freight from inbound to outbound vehicles with minimal storage.

Relay: Long-distance operating method involving exchange of drivers or tractors.

Consolidation: Combining several shipments into a larger movement.

Deconsolidation: Separating a consolidated movement for local delivery.

Terminal: Facility where freight, vehicles, equipment, or information is transferred or processed.

Service frequency: Number of scheduled departures or delivery opportunities in a period.

Transit time: Elapsed time between shipment acceptance and delivery.

Reliability: Consistency of service performance, often measured by on-time delivery.

Generalized cost: Monetary representation of transport, handling, time, risk, quality, and environmental costs.

Last mile: Final movement from a distribution point to the customer.

Time window: Permitted interval for pickup or delivery.

Vehicle-routing problem: Problem of assigning and sequencing vehicle movements to serve customers subject to constraints.

Facility-location model: Model selecting facility sites and assigning demand to them.

Cross-dock dependency: Exposure created when a network relies heavily on one transfer facility.

Capacity slack: Unused capacity available to absorb variability or disruption.

Review Questions

Short Answer Questions

1. What are the principal advantages and disadvantages of a hub-and-spoke network?
2. Why can consolidation increase total service time even when it reduces transport cost?
3. Define cross-docking and identify two conditions required for its success.
4. What does service frequency mean, and how does it affect waiting time?
5. Why should last-mile design include failed-delivery management?

Long Answer Questions

1. Compare direct, hub-and-spoke, milk-run, and consolidation networks for a firm with dispersed customers and small shipments.
2. Explain how terminal location affects cost, time, capacity, resilience, and sustainability.
3. Discuss the limitations of using freight rate as the sole basis for mode choice.
4. Explain how optimization and scenario analysis should be combined in strategic network design.
5. Evaluate the operational, ethical, and regulatory implications of using AI for route planning.

Analytical / Application Questions

1. A firm has evenly spaced departures every 24 hours. Estimate expected pre-departure waiting time under the uniform-demand assumption. What operational factors could make actual waiting time higher?
2. A company is choosing between a direct route and a cross-dock design. Develop a comparison framework including transport, handling, inventory, reliability, emissions, and disruption costs.
3. Design a last-mile pilot for a dense urban district with narrow delivery windows. Specify the service promise, vehicle and route policies, failure process, performance metrics, and conditions under which the pilot should be expanded.

Multiple Choice Questions with Answers

Question 1

Which network architecture routes freight through an intermediate consolidation point before delivery?

A. Direct transport

B. Hub-and-spoke

C. Point-to-point express

D. Single-stop shuttle

Correct Answer: B. Hub-and-spoke

Question 2

A milk run is best described as:

A. A vehicle serving multiple suppliers or customers on a planned tour

B. A shipment transported only by rail

C. A direct shipment with no scheduled stops

D. A terminal storing goods for long periods

Correct Answer: A. A vehicle serving multiple suppliers or customers on a planned tour

Question 3

If departures occur every 24 hours and demand arrives uniformly, the expected waiting time before departure is:

A. 6 hours

B. 12 hours

C. 24 hours

D. 48 hours

Correct Answer: B. 12 hours

Question 4

Which condition is most critical to successful cross-docking?

A. Long storage dwell times

B. Unsynchronized inbound and outbound schedules

C. Accurate shipment information and synchronized operations

D. Elimination of all handling equipment

Correct Answer: C. Accurate shipment information and synchronized operations

Question 5

In a facility-location model, the variable yᵢ commonly represents:

A. The distance between two customers

B. Whether candidate facility i is opened

C. The transit time of a shipment

D. The number of delivery stops

Correct Answer: B. Whether candidate facility i is opened

Question 6

Which item is part of generalized transport cost?

A. Only the carrier’s quoted freight rate

B. Only fuel consumption

C. Transport, handling, time, risk, and service effects

D. Only the fixed cost of a warehouse

Correct Answer: C. Transport, handling, time, risk, and service effects

Question 7

A major advantage of relay operations is that they can:

A. Eliminate all regulatory requirements

B. Increase long-haul utilization by exchanging drivers or tractors

C. Remove the need for schedules

D. Guarantee zero shipment handling

Correct Answer: B. Increase long-haul utilization by exchanging drivers or tractors

Question 8

If goods worth $100,000 are in transit for 2 days at an annual carrying-cost rate of 18%, the approximate time-related inventory cost is:

A. $9.00

B. $49.32

C. $98.63

D. $3,600.00

Correct Answer: C. $98.63

Question 9

Which measure is most directly associated with last-mile delivery density?

A. Stops divided by route distance

B. Facility fixed cost divided by demand

C. Transit time divided by annual interest rate

D. Vehicle weight divided by driver count

Correct Answer: A. Stops divided by route distance

Question 10

Why is scenario analysis important in transport network design?

A. It guarantees that forecasts are correct

B. It replaces the need for operational validation

C. It tests how alternatives perform under uncertainty and disruption

D. It makes capacity constraints irrelevant

Correct Answer: C. It tests how alternatives perform under uncertainty and disruption