Transportation, Routing and Freight Analytics

Introduction

Transportation is the physical execution of logistics strategy. It connects suppliers, plants, distribution centres, retailers and customers, while determining a substantial share of total logistics cost, delivery reliability, inventory exposure and environmental impact. Transportation analytics converts shipment, network, vehicle and customer data into decisions about mode, carrier, price, consolidation, routing, scheduling and fleet design.

The central managerial challenge is not to minimise transport cost in isolation. A low-cost option may increase transit time, stockout risk, damage, emissions or customer dissatisfaction. The appropriate objective is therefore a constrained trade-off among cost, service, capacity, risk and sustainability.

Learning Objectives

After completing this chapter, the reader should be able to:

1. Compare transport modes and carriers using cost, service, risk and emissions criteria.
2. Analyse freight rates, shipment economics and total landed cost.
3. Apply load-consolidation and vehicle-utilisation concepts.
4. Formulate and interpret vehicle-routing and scheduling models.
5. Evaluate last-mile delivery performance and operational trade-offs.
6. Measure transport emissions and incorporate them into decisions.
7. Critically assess data, technology, governance and implementation risks.

1. Transportation as an Analytical Decision System

Transportation decisions occur at three levels:

A transport decision should be evaluated using a total-cost perspective. A useful accounting identity is:

where TC is total transport-related cost; F is fixed cost; VQ is variable movement cost; I is inventory-in-transit cost; S is service-failure cost; R is risk or expected disruption cost; and E is monetised environmental cost. All terms must be expressed in the same currency and planning period. The expression is a decision framework rather than a universal accounting identity: some organisations report inventory, risk and emissions separately.

Service should be represented explicitly. Common indicators include on-time delivery, order-cycle time, delivery completeness, damage rate, tender acceptance, dwell time and customer promise adherence. A composite objective can be written as:

where C is monetary cost, OTIF is the proportion delivered on time and in full, D is a disruption or delay measure, E is emissions, and λ₁, λ₂, λ₃ are management-selected penalty weights. The weights do not arise automatically from data; they express managerial priorities and should be tested through sensitivity analysis.

2. Mode and Carrier Selection

2.1 Comparing transport modes

Road offers geographic flexibility and frequent service but is exposed to congestion, driver availability and fuel-price variation. Rail is efficient for high-volume inland movements but usually requires terminal handling and timetable coordination. Ocean has low unit cost for international bulk movement but long and variable transit times. Air provides speed and high security at a high price and with substantial emissions. Pipelines are reliable for suitable liquids and gases but lack product flexibility.

Mode selection should consider:

• shipment volume, density, value and physical characteristics;
• required delivery date and acceptable variability;
• network accessibility and terminal handling;
• damage, theft, temperature and security risk;
• contractual flexibility and capacity availability;
• fuel exposure, carbon intensity and regulatory requirements.

The relevant comparison is often total landed cost rather than quoted freight. For a shipment, an analytical approximation is:

where L is landed cost; P is product value; F is freight; H is handling and terminal cost; W is insurance; I_t is inventory carrying cost during transit; G is expected loss from damage or deterioration; and X is customs, documentation and other applicable charges. The product value may be excluded when comparing modes for the same shipment, but it should be included when analysing total supply-chain exposure.

2.2 Carrier evaluation

Carrier selection should distinguish measurable performance from marketing claims. A scorecard can include price, capacity, transit-time reliability, claims performance, safety, financial resilience, technology integration and sustainability. A weighted score is:

where S_j is the score of carrier j, w_i is the importance weight of criterion i, and r_ij is the normalised rating of carrier j on criterion i. Ratings should be placed on a common scale, such as 0 to 100. Cost criteria must be oriented consistently: a lower price should produce a higher score after normalisation. Weighted scoring is useful for screening, but it can conceal threshold failures. For example, a carrier with a high aggregate score should still be rejected if it fails a mandatory safety or temperature-control requirement.

A robust procurement process combines scorecards with scenario analysis. Rates should be evaluated under changes in fuel price, demand, accessorial charges, capacity scarcity, border delay and service penalties. Contract terms should clarify tender response, detention, demurrage, claims, data access, force majeure, audit rights and emissions reporting.

3. Freight Rates and Shipment Economics

Freight pricing may be based on distance, weight, volume, density, pallet, container, shipment, lane, dimensional weight, declared value or a combination of these. The billable quantity is often the greater of actual and dimensional weight. A generic representation is:

where Q_b is billable quantity, Q_a is actual weight and Q_d is dimensional weight, using the carrier’s stated dimensional divisor and units. The divisor is contract- and mode-specific; it must not be assumed without checking the tariff.

A simple shipment charge may be represented as:

where B is a base charge, r is the rate per billable unit, A is accessorial cost, F_s is a fuel surcharge and T is tax or other applicable charge. Real tariffs may contain minimum charges, brackets, discounts, currency adjustments and peak surcharges.

Worked example: mode comparison

A company must move 10,000 kg over 800 km. Road costs 0.085 currency units per tonne-kilometre, with an additional fixed terminal charge of 240. Rail costs 0.045 per tonne-kilometre and terminal charges of 1,100. Road transit is two days and rail transit is four days. Inventory value is 120,000, and the annual carrying rate is 18%, using 365 days.

Road freight is:

Rail freight is:

The additional two days of rail inventory cost are:

Thus, the incremental economic cost of rail relative to road is approximately 658.36. If rail has materially lower damage, emissions or disruption exposure, the decision may nevertheless differ. The calculation is valid only if the tonne-kilometre rates apply to the stated shipment and the inventory value and carrying rate are appropriate.

4. Load Consolidation and Fleet Utilisation

Consolidation combines shipments to improve vehicle utilisation and reduce cost per unit. It can be temporal, geographic, product-based or cross-docking based. Benefits arise from sharing vehicle capacity and reducing the number of dispatches. Costs may arise from waiting, additional handling, inventory delay and loss of shipment flexibility.

Vehicle utilisation has several distinct meanings:

Capacity may be measured by weight, volume, pallets, loading positions or legal axle limits. A truck can be 95% full by volume but overloaded by weight, or full by weight while using little cubic space. Therefore, utilisation should be reported by the binding constraint and, where relevant, by both weight and cube.

For a route with vehicle capacity K and loaded quantity q, load utilisation is q ÷ K, provided both quantities use the same unit. Empty running is:

Backhauls, collaborative transport, pool distribution and shipment appointment coordination can reduce empty running. However, collaboration requires clear rules for data confidentiality, liability, service ownership and benefit sharing.

The economic value of consolidation can be assessed as:

This should be evaluated by lane and customer segment rather than assumed universally.

5. Vehicle Routing and Scheduling

5.1 Routing problem structure

The vehicle-routing problem (VRP) determines routes for vehicles serving customers from one or more depots. Its variants include the capacitated VRP, vehicle-routing problem with time windows, multi-depot VRP, pickup-and-delivery VRP, heterogeneous-fleet VRP and stochastic or dynamic VRP.

A basic capacitated VRP can be represented using binary variable x_ijk, equal to 1 if vehicle k travels directly from node i to node j, and 0 otherwise. Let c_ij be travel cost, q_i customer demand, Q_k vehicle capacity, and y_ik indicate whether vehicle k serves node i. A simplified objective is:

subject to each customer being served once, route continuity being maintained, vehicle capacity not being exceeded, and each route beginning and ending at an authorised depot. Additional subtour-elimination constraints are necessary; otherwise, the mathematical solution may contain disconnected cycles that do not represent feasible depot-based routes.

For capacity:

For time windows, if customer i must be served between a_i and b_i, service start t_i must satisfy:

If travel time from i to j is τ_ij and service time at i is s_i, route sequencing requires, when the arc is used:

In mixed-integer models this implication is usually implemented with a sufficiently large constant or an indicator constraint. The constant must be chosen carefully: an excessively large value weakens computational performance, while an insufficient value can remove feasible solutions.

5.2 Exact and heuristic methods

Small routing problems can be solved with mixed-integer optimisation. Large operational problems commonly use savings algorithms, insertion heuristics, local search, tabu search, simulated annealing, genetic algorithms or adaptive large-neighbourhood search. A heuristic may produce an excellent solution quickly without proving optimality. Managers should therefore examine feasibility, stability, explainability and sensitivity rather than relying only on the objective value.

A practical workflow is: validate data; construct a feasible baseline; optimise; impose operational rules; test disruptions; obtain dispatcher review; release routes; monitor execution; and re-optimise when conditions change. Route plans should include realistic loading, breaks, traffic, parking, service and legal driving constraints.

6. Scheduling, Fleet Utilisation and Last-Mile Delivery

Scheduling coordinates vehicles, drivers, loading resources, docks and customer appointments. A route that is mathematically short can fail if loading docks are congested or a driver cannot legally complete the duty. Important constraints include driver hours, vehicle availability, maintenance, loading compatibility, delivery windows, school or urban restrictions and refrigeration requirements.

Fleet sizing should consider demand variability rather than average demand alone. A rough planning relationship is:

The result must be rounded up and adjusted for reserve vehicles, maintenance, absenteeism and route imbalance. Effective capacity is less than nominal capacity when legal, operational or product-mix restrictions apply.

Last-mile delivery is often the most expensive and least predictable segment. Its cost drivers include low drop density, failed delivery attempts, parking and access time, customer time windows, reverse logistics and urban congestion. Useful metrics include cost per stop, stops per route hour, first-attempt success, average dwell time, kilometres per parcel, delivery promise adherence and return rate.

Alternative designs include parcel lockers, pickup points, micro-depots, crowdsourced delivery, scheduled windows, dynamic routing and electric vehicles. These solutions must be assessed against customer accessibility, security, labour conditions, data protection and total lifecycle cost.

7. Transport Emissions Analytics

A basic activity-based emissions calculation is:

For freight, activity may be tonne-kilometres, vehicle-kilometres, litres of fuel or electricity consumed. If activity is measured in tonne-kilometres, the units of the emission factor must be kg CO₂e per tonne-kilometre. For fuel-based calculation:

The analyst must define whether the factor is well-to-wheel or tank-to-wheel, whether empty distance is included, the treatment of renewable electricity, and the allocation method for shared loads. For a shared truck, allocation by mass, volume, distance or revenue produces different results. The chosen method should be documented and consistently applied.

A transport emissions intensity measure is:

where freight work may be tonne-kilometres or delivered parcels. Intensity can improve while total emissions increase if overall activity grows, so both absolute emissions and intensity should be monitored.

Emission reduction measures include better consolidation, reduced empty running, modal shift, aerodynamic improvement, driver training, alternative fuels, route optimisation, electrification and renewable energy. Electrification analysis must include vehicle purchase cost, battery range, payload effects, charging downtime, electricity source and residual value. Carbon reporting should not be used to justify service reductions that create hidden social or economic harm.

8. Technology, Data and Artificial Intelligence

Transport management systems integrate orders, rates, carriers, vehicles, geospatial data, telematics, proof of delivery and billing. GPS and telematics support actual-time analysis of location, speed, idling, harsh events, temperature and dwell. Application programming interfaces can connect shippers, carriers, marketplaces and customers.

AI and machine learning can forecast travel time, predict failed deliveries, estimate arrival times, identify abnormal fuel use and recommend dynamic routes. Their value depends on data quality and operational integration. Common risks include biased historical decisions, missing data, unstable predictions during unusual events, opaque recommendations and automation without accountability. Human override, model monitoring, version control, cybersecurity and audit trails are essential.

Data governance should specify ownership, access rights, retention, security, quality standards and permitted secondary use. Location and driver data may constitute personal data under applicable law. Analytics programmes must comply with transport safety, employment, competition, customs, environmental and privacy regulation.

9. Illustrative Case: Northstar Consumer Products

Background

Northstar distributes 1,200 pallet-equivalents each week from one regional distribution centre to 42 retailers. It operates a dedicated fleet for core routes and contracts carriers for overflow. Average truck fill is 68% by cube, empty running is 19%, and on-time-in-full performance is 91%. Urban deliveries generate high parking and dwell times.

Problem

Management observes that freight cost has increased despite stable demand. It is considering a cheaper carrier, route optimisation software and electric vehicles, but lacks a common analytical basis.

Analysis

The team first cleans order, route, fuel, loading and proof-of-delivery data. It separates weight and cube utilisation and discovers that several trucks are cube-constrained while others are weight-constrained. It calculates cost per pallet, cost per stop, empty kilometres, failed attempts, dwell time and CO₂e per tonne-kilometre. Carrier evaluation includes a minimum safety threshold and a weighted score for price, reliability, claims and emissions.

A capacitated time-window model is then constructed. It includes delivery appointments, driver duty limits, vehicle capacities, loading compatibility and realistic service times. A baseline heuristic is compared with an optimisation solution and dispatcher review. The proposed policy consolidates geographically compatible orders, shifts two low-density long-haul lanes to rail where service permits, and introduces delivery windows with clustered urban stops.

Solution and outcome

In the illustrative scenario, the pilot produces 8% fewer vehicle-kilometres, raises average cube utilisation to 79%, reduces empty running to 12%, and increases OTIF to 95%. The improvement is not attributed solely to software: better appointment discipline, rate governance and dispatcher involvement are also required. Electric vehicles are assigned only to routes whose distance, payload and charging schedule are feasible; the company postpones broad deployment until winter-range and charging data are validated.

Learning points

The case demonstrates that transport analytics is a portfolio of linked decisions. A routing algorithm cannot compensate for inaccurate master data, unsuitable carrier contracts or unrealistic service assumptions. Performance must be evaluated through cost, service, utilisation, emissions and operational feasibility together.

10. Limitations, Governance and Best Practice

Transport models simplify reality. Distance matrices may ignore traffic and access restrictions; demand forecasts may be biased; reported capacity may be unavailable in practice; and emissions factors may not match local technology. Optimised routes may also create unacceptable driver workloads or customer inequity.

Best practice includes:

1. Establish a single definition for shipment, stop, delivery, tender and on-time performance.
2. Reconcile transport-management, telematics, billing and proof-of-delivery data.
3. Use hard constraints for safety, legal and product requirements.
4. Separate mandatory thresholds from weighted preferences.
5. Compare a baseline, an optimised solution and disruption scenarios.
6. Report both average performance and variability or service tails.
7. Validate assumptions with drivers, dispatchers, carriers and customers.
8. Monitor realised savings rather than projected savings only.
9. Document emissions boundaries, factors and allocation rules.
10. Review models periodically as network, regulation and technology change.

Chapter Summary

Transportation analytics supports strategic, tactical and operational decisions about modes, carriers, rates, consolidation, routing, scheduling, fleet and last-mile delivery. Sound analysis uses total landed cost and recognises service, risk, capacity and environmental consequences. Carrier scorecards are useful when combined with mandatory thresholds and scenario analysis. Freight charges depend on billable quantity, rate structures and accessorial costs. Consolidation can reduce freight and empty running but may create delay and handling costs.

Vehicle-routing models formalise customer, capacity, depot and time-window constraints. Exact optimisation is appropriate for smaller problems, while heuristics and metaheuristics are often practical for large or dynamic operations. Fleet utilisation must distinguish weight, cube, legal and temporal constraints. Last-mile performance depends strongly on density, dwell time, failed attempts and customer windows. Emissions analytics requires clearly defined activity boundaries and emission factors. Technology and AI improve visibility and prediction only when supported by reliable data, governance, human oversight and responsible implementation.

Key Terms

Mode selection: Choosing among road, rail, ocean, air, pipeline or multimodal alternatives.

Total landed cost: Product, freight, handling, inventory, insurance, risk and related delivered cost.

Billable weight: Quantity used to calculate freight charges, often the greater of actual and dimensional weight.

Accessorial charge: Additional charge for services such as detention, lift-gate delivery or special handling.

Carrier scorecard: Structured evaluation of carrier cost, service, risk and capability.

Consolidation: Combining shipments to improve utilisation and reduce movement cost.

Empty running: Vehicle distance travelled without revenue-generating freight.

Capacity utilisation: Used capacity divided by available capacity.

Vehicle-routing problem: Optimisation problem assigning vehicle routes to customer visits.

Time window: Permitted interval for serving a customer.

Subtour: Disconnected cycle in a routing solution that does not connect properly to a depot route.

Heuristic: Method that seeks a good feasible solution without guaranteeing optimality.

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

Dwell time: Time a vehicle spends waiting or being serviced at a location.

OTIF: On-time-in-full delivery performance measure.

Freight work: Transport activity commonly measured in tonne-kilometres.

CO₂e: Carbon-dioxide equivalent representing the warming impact of greenhouse gases.

Dynamic routing: Route adjustment based on changing demand, traffic or operational conditions.

Transport management system: Software supporting planning, execution, visibility and settlement of transport.

Review Questions

5 Short Answer Questions

1. Why should mode selection consider total landed cost rather than freight price alone?
2. Distinguish actual weight, dimensional weight and billable weight.
3. What is the difference between weight utilisation and cube utilisation?
4. State two operational constraints in a vehicle-routing problem with time windows.
5. Why should transport emissions be reported using both absolute and intensity measures?

5 Long Answer Questions

1. Explain a rigorous framework for carrier selection, including thresholds, weighted criteria and scenario analysis.
2. Discuss the economic benefits and operational risks of shipment consolidation.
3. Describe the structure and assumptions of a capacitated vehicle-routing model.
4. Evaluate the major cost and service challenges of last-mile delivery.
5. Discuss how transport-management systems, telematics and AI can improve transport decisions, including governance risks.

3 Analytical / Application Questions

1. A carrier offers a low freight rate but has poor tender acceptance and high damage rates. Design a quantitative and qualitative evaluation to determine whether it should be selected.
2. A distributor has vehicles that are full by cube but lightly loaded by weight. Analyse possible causes and recommend operational interventions.
3. Develop a pilot plan for reducing transport emissions while preserving OTIF performance. Specify metrics, data boundaries, constraints and decision rules.

Multiple Choice Questions with Answers

Question 1

Which measure best represents a transport decision that includes freight, handling, inventory and risk costs?

A. Line-haul rate

B. Total landed cost

C. Vehicle fill rate

D. Tender acceptance

Correct Answer: B. Total landed cost

Question 2

A shipment has actual weight of 800 kg and dimensional weight of 950 kg. If the tariff uses the greater quantity, what is the billable weight?

A. 150 kg

B. 800 kg

C. 950 kg

D. 1,750 kg

Correct Answer: C. 950 kg

Question 3

Which constraint ensures that a vehicle does not carry more demand than its capacity?

A. ∑ᵢ qᵢyᵢₖ ≤ Qₖ

B. ∑ᵢwᵢ = 1

C. CO₂e = Activity × factor

D. OTIF = 1 − delay

Correct Answer: A. ∑ᵢ qᵢyᵢₖ ≤ Qₖ

Question 4

What is a key limitation of a weighted carrier scorecard?

A. It cannot compare more than one carrier

B. It automatically proves mathematical optimality

C. It may conceal failure on a mandatory safety threshold

D. It excludes all service criteria

Correct Answer: C. It may conceal failure on a mandatory safety threshold

Question 5

A route travels 1,000 km, of which 180 km are empty. What is the empty-run rate?

A. 5.6%

B. 18%

C. 55.6%

D. 82%

Correct Answer: B. 18%

Question 6

Which routing problem specifically includes permitted customer service intervals?

A. Uncapacitated flow problem

B. Vehicle-routing problem with time windows

C. Economic order quantity problem

D. Assignment problem without routes

Correct Answer: B. Vehicle-routing problem with time windows

Question 7

If a truck carries 18 tonnes over 400 km, what freight work does it perform?

A. 22.2 tonne-km

B. 418 tonne-km

C. 7,200 tonne-km

D. 7,200 km per tonne

Correct Answer: C. 7,200 tonne-km

Question 8

Which action most directly reduces empty running?

A. Increasing failed delivery attempts

B. Removing backhaul planning

C. Developing compatible return loads

D. Reducing data quality controls

Correct Answer: C. Developing compatible return loads

Question 9

Why should a large constant in a mixed-integer time-window model be chosen carefully?

A. It determines fuel consumption directly

B. An excessively large value can weaken computational performance

C. It eliminates the need for service times

D. It converts weight into volume

Correct Answer: B. An excessively large value can weaken computational performance

Question 10

What is the correct basic activity-based emissions relationship?

A. CO₂e = Activity × Emission factor

B. CO₂e = Freight rate ÷ Distance

C. CO₂e = Capacity − Demand

D. CO₂e = OTIF × Vehicle count

Correct Answer: A. CO₂e = Activity × Emission factor