Logistics, Transportation and Distribution Analytics

Introduction

Logistics analytics concerns the systematic use of data, models and managerial judgment to move, store and fulfill products efficiently. It connects suppliers, production facilities, distribution centers, retailers and customers through decisions about transportation modes, shipment consolidation, routing, inventory positioning, warehouse operations, order fulfillment, last-mile delivery and product returns.

These decisions are interdependent. A firm may reduce transportation cost by consolidating shipments, but consolidation can increase customer waiting time and inventory. A warehouse located near customers can improve delivery speed, but may have higher rent. A faster transport mode may reduce pipeline inventory while increasing emissions and cost. Consequently, logistics analytics is not simply a cost-minimization exercise; it balances cost, service, resilience, flexibility, risk and environmental performance.

Learning Objectives

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

1. Compare transportation modes using cost, speed, reliability, capacity and emissions.
2. Apply shipment-cost, assignment and transportation-model concepts.
3. Explain routing and vehicle-routing decisions and common solution methods.
4. Evaluate warehouse productivity, space utilization, fulfillment and inventory-flow metrics.
5. Analyze last-mile and reverse-logistics systems.
6. Recommend logistics improvements while recognizing data, ethical, regulatory and implementation constraints.

1. The Logistics System and Its Trade-offs

A logistics network is a directed system of nodes and links. Nodes include plants, ports, cross-docks, warehouses, stores and customer locations. Links represent road, rail, air, maritime, pipeline or multimodal movements. Analytics typically seeks to answer five questions:

• Where should inventory be positioned?
• Which mode and carrier should be used?
• Which orders should be consolidated?
• What route and delivery sequence should be selected?
• How should facilities and labor be scheduled?

A useful total-cost perspective is:

where C_T is transportation cost, C_H handling and warehousing cost, C_F facility cost, C_I inventory-related cost, C_R risk or service-failure cost, and C_E environmental cost when it is monetized. The expression is a managerial accounting framework rather than a universal law. Its components must be defined consistently to avoid double counting.

Important service measures include order-cycle time, on-time-in-full delivery (OTIF), fill rate, damage rate, delivery-window adherence and perfect-order percentage. A cost improvement that worsens a strategically important service measure may destroy customer value.

2. Transportation Modes and Carrier Selection

2.1 Mode characteristics

Mode selection should consider shipment weight, volume, value density, perishability, security, customs requirements, lane distance, reliability and carbon intensity. The cheapest quoted freight rate may not produce the lowest total landed cost because delays, damage, handling and inventory carrying costs also matter.

2.2 Shipment economics

For a shipment with freight charge F, goods value V, annual holding-rate i, and transit time t measured in years, the approximate pipeline-inventory cost is:

This assumes the goods value remains tied up throughout transit and that the holding rate is applicable to that value. If t is expressed in days and the rate is annual, use t/365. The formula excludes damage, insurance and stockout effects.

The relevant transportation comparison is therefore approximately:

where C_H is additional handling, C_D expected damage or loss cost, and C_S expected service-failure cost.

Worked example. A shipment has goods value of $120,000. Air freight costs $8,400 and takes 2 days; ocean freight costs $2,100 and takes 24 days. The annual holding rate is 18%, and additional ocean handling is $400. Ignoring other differences:

Air pipeline cost = $120,000 × 0.18 × 2/365 = $118.36.

Air relevant cost = $8,400 + $118.36 = $8,518.36.

Ocean pipeline cost = $120,000 × 0.18 × 24/365 = $1,420.27.

Ocean relevant cost = $2,100 + $1,420.27 + $400 = $3,920.27.

Ocean is less costly under these assumptions. The conclusion could change if delay, spoilage or stockout consequences were material.

3. Transportation and Assignment Models

3.1 The transportation model

The classical transportation model allocates supply from origins to destinations at minimum cost. Let:

• x_ij = units shipped from origin i to destination j;
• c_ij = cost per unit on that lane;
• s_i = available supply at origin i;
• d_j = demand at destination j.

subject to:

If total supply equals total demand, the model is balanced and equality can be used in both constraint sets. If supply exceeds demand, unused supply may be represented by a dummy destination; if demand exceeds supply, unmet demand may be represented by a dummy origin with a penalty cost. The model assumes known costs, divisible flow, compatible products and no lane capacity beyond the stated supply unless additional constraints are added.

3.2 Assignment decisions

An assignment problem allocates each worker, vehicle, order or job to exactly one task, often with each task receiving exactly one resource. With binary variable x_ij equal to 1 when resource i is assigned to task j, the objective is:

subject to:

In logistics, assignments may match drivers to routes, vehicles to delivery waves or orders to fulfillment centers. Real applications often require extensions for skills, time windows, vehicle capacity and labor regulations.

4. Routing and Vehicle-Routing Analytics

Routing determines the sequence in which stops are served. The traveling salesperson problem (TSP) seeks a minimum-distance or minimum-cost tour that visits each customer once and returns to the depot. The vehicle-routing problem (VRP) extends this to multiple vehicles and commonly includes capacity, time windows, driver hours, service times, heterogeneous fleets and pickup-and-delivery requirements.

For a route with arcs (i,j), distance d_ij, vehicle cost per distance k, and fixed vehicle cost f, an illustrative cost function is:

where K is the number of vehicles and x_ij indicates whether arc (i,j) is used. This is only a cost component; a complete VRP requires constraints ensuring route continuity, customer coverage, capacity and elimination of subtours.

Vehicle capacity constraints can be expressed as:

where q_i is demand at customer i, y_ik indicates service by vehicle k, and Q_k is vehicle capacity. If demand is measured in kilograms, capacity must also be measured in kilograms; volume and pallet constraints may need separate inequalities.

Exact optimization methods such as mixed-integer programming can provide optimal solutions for smaller instances. Larger operational problems commonly use heuristics and metaheuristics, including savings methods, local search, tabu search, genetic algorithms and adaptive large-neighborhood search. A practical workflow is:

1. Validate orders, locations, service times and vehicle data.
2. Define hard constraints and soft preferences.
3. Generate a feasible baseline.
4. Improve distance, cost or service using an optimization method.
5. Test sensitivity to traffic, cancellations and demand changes.
6. Dispatch, monitor and re-optimize when conditions change.

A mathematically shorter route is not always operationally superior. A route that violates realistic loading sequence, driver breaks, delivery windows or curb-access restrictions is not feasible.

5. Warehousing and Fulfillment Analytics

Warehouses transform inventory flows. Core processes include receiving, put-away, storage, replenishment, picking, packing, staging, shipping and returns. Analytics should examine both resource productivity and customer outcomes.

Common measures include:

Space utilization should not be pushed toward 100%. Congestion, blocked aisles, poor accessibility and replenishment delays can increase well before physical capacity is exhausted. Slotting places high-volume or frequently co-ordered items in accessible positions. ABC analysis prioritizes attention, but should be supplemented by cube, velocity, fragility, compatibility and seasonality.

The basic inventory turnover ratio is:

It is meaningful only when numerator and denominator use the same valuation basis. A high turnover may indicate efficiency, but can also signal understocking and lost sales.

Fulfillment design includes ship-from-store, centralized fulfillment, cross-docking, drop-shipping, micro-fulfillment centers and dark stores. Each option changes fixed cost, inventory pooling, order splitting, delivery distance and control complexity.

6. Last-Mile Delivery

Last-mile delivery is often the most expensive and variable segment because orders are fragmented across many destinations. Relevant decisions include delivery-window design, depot location, vehicle type, parcel locker deployment, delivery density, customer presence and failed-delivery management.

Useful indicators include cost per stop, stops per route-hour, first-attempt success rate, delivery-window adherence, miles per package, packages per route and customer complaints. Density is crucial: more deliveries per kilometer generally reduce average cost, but only if service times and congestion remain manageable.

Delivery promises should be based on realistic capacity and uncertainty. Overly narrow windows may increase route cost and emissions. Consolidated delivery, pickup points, lockers, electric vehicles and cargo bikes can improve both urban efficiency and environmental performance, although they require suitable infrastructure and demand patterns.

7. Reverse Logistics and Circular Flows

Reverse logistics manages returns, repair, refurbishment, resale, recycling and disposal. It differs from forward logistics because return timing, condition, value and reason are uncertain. A return process should classify products quickly, authorize appropriate disposition and prevent avoidable transport.

A simple expected-value decision compares disposition alternatives. If option r has recovery value v_r, processing cost p_r, transport cost t_r and probability Pr_r, its expected net value is:

The probabilities must sum to 1 across mutually exclusive outcomes. This model excludes strategic brand, regulatory and environmental effects unless they are represented as additional terms.

Return-rate analysis should distinguish customer remorse, damage, wrong item, size mismatch, warranty failure and fraudulent activity. Treating all returns as one category conceals root causes. Product design, packaging, clearer descriptions and quality improvement may reduce returns more effectively than simply optimizing reverse transportation.

8. Emissions and Sustainable Logistics

A basic freight-emissions estimate is:

where E is emissions, W shipment mass, D distance and e an emissions factor per mass-distance unit. If the emissions factor is in kilograms CO₂e per tonne-kilometer, weight must be converted to tonnes and distance to kilometers. Real calculations may require vehicle type, load factor, empty running, fuel, electricity source, refrigeration and well-to-wheel versus tank-to-wheel boundaries.

Carbon reduction options include better load utilization, mode shift, route consolidation, reduced empty miles, intermodal planning, alternative fuels, electric delivery vehicles, efficient packaging and near-customer inventory. Firms should report boundaries and assumptions transparently; otherwise apparently favorable comparisons may be misleading.

9. Technology, Data and Artificial Intelligence

Transportation-management systems (TMS), warehouse-management systems (WMS), enterprise-resource-planning systems, telematics, barcode or RFID capture, GPS and application-programming interfaces provide the data foundation for analytics. A control tower can combine shipment status, inventory, orders, weather, traffic and carrier information.

AI and machine learning can support demand forecasting, estimated-time-of-arrival prediction, anomaly detection, dynamic pricing, image-based damage inspection and route re-optimization. However, optimization quality depends on data quality and operational acceptance. A model trained on historical routes may reproduce inequitable workload allocations or fail under unusual disruptions. Human oversight is necessary when decisions affect driver safety, employment, customer access or regulatory compliance.

Good governance includes data ownership, access controls, audit trails, cybersecurity, model monitoring, explainability for consequential decisions and documented override procedures. Personal location data should be minimized, protected and retained only as long as justified.

10. Illustrative Case: FreshBox Urban Distribution

Background

FreshBox is an illustrative online grocery business serving one metropolitan area. It operates one regional fulfillment center and delivers 1,200 orders per day. Demand has grown, but on-time delivery has fallen from 96% to 88%. Delivery cost is $9.40 per order, and 7% of orders require a second delivery attempt.

Problem

Management is considering three actions: adding delivery vans, creating two neighborhood pickup points, or opening a small micro-fulfillment facility. It must improve service without increasing emissions disproportionately.

Analysis

The analytics team first cleans address, promised-window, actual-arrival, route, vehicle, order-size and failed-delivery data. It discovers that late orders are concentrated in two high-density zones during evening peaks. Routes are also built using average travel times, which underestimates Friday congestion. Failed deliveries are more common where customers receive narrow windows and apartment access is difficult.

The team evaluates three scenarios:

A time-dependent VRP is used for delivery routes. The model includes vehicle capacity, driver-hour limits, service times, congestion-adjusted travel times and delivery windows. A separate simulation tests demand variation and failed deliveries. The team also estimates emissions using shipment mass, route distance and vehicle-specific factors.

Approach and solution

The recommended pilot combines one pickup point in the densest zone, wider optional delivery windows, evening locker delivery and congestion-aware routing. The existing regional facility remains the inventory source during the pilot, avoiding the immediate complexity of a micro-fulfillment center. Customers receive a small discount for pickup-point selection, while urgent home delivery remains available.

Outcome

In the pilot design, expected home stops fall in the densest zone, route density improves, and failed first attempts decline. The business should not treat modeled results as realized benefits: it must compare a pre-intervention baseline with pilot performance and monitor customer adoption, total cost, service, emissions and equity of access.

Learning points

The case illustrates that route optimization alone may not solve a structural last-mile problem. Delivery-window design, customer behavior, facility choices and failed deliveries must be analyzed together. It also demonstrates why scenario analysis is preferable to selecting a technology before understanding the cause of poor performance.

11. Implementation, Limitations and Professional Judgment

Logistics models are vulnerable to inaccurate addresses, missing service times, unrealistic travel speeds, stale capacities, poor demand forecasts and inconsistent cost definitions. Optimization can also create local improvements that worsen the network—for example, minimizing warehouse labor while increasing transportation and customer delay.

Implementation should therefore proceed through baseline measurement, stakeholder validation, pilot testing, exception monitoring and post-implementation review. Managers should distinguish hard constraints from preferences, record assumptions, conduct sensitivity analysis and communicate uncertainty.

Regulatory and ethical issues include driver working-time rules, vehicle weight limits, dangerous-goods regulations, customs documentation, accessibility, worker surveillance, algorithmic discrimination, privacy and environmental claims. Cost reduction must not depend on unsafe driving, unlawful hours, misleading delivery promises or exclusion of customers in less profitable locations.

Best-Practice Checklist

• Define service, cost and emissions measures before optimizing.
• Use total landed cost rather than freight price alone.
• Validate data with drivers, warehouse staff and customer-service teams.
• Model capacity, time windows, breaks, loading sequence and uncertainty.
• Separate planned performance from exception performance.
• Test solutions under peak demand and disruption scenarios.
• Track emissions boundaries and units explicitly.
• Pilot major network changes before full deployment.
• Keep human accountability for consequential automated decisions.
• Reassess models as demand, regulations, fleet technology and customer behavior change.

Chapter Summary

Logistics and distribution analytics integrates transportation, warehousing, fulfillment, last-mile delivery and reverse logistics. Transportation-mode selection balances cost, speed, reliability, capacity, inventory exposure and emissions. Transportation models allocate flow between origins and destinations, while assignment models match resources to tasks. Routing and vehicle-routing problems add sequence, capacity, time-window and labor constraints and are commonly solved with optimization and heuristics.

Warehouse analytics uses measures such as inventory accuracy, space utilization, picking productivity, order-cycle time and perfect-order rate. Last-mile performance depends strongly on delivery density, congestion, delivery windows and failed attempts. Reverse logistics requires classification, recovery-value analysis and attention to return causes. Technology and AI can improve visibility and prediction, but require reliable data, governance and human oversight. Effective decisions balance financial cost with service, resilience, safety, equity and environmental consequences.

Key Terms

Transportation mode: A method of moving goods, such as road, rail, air or maritime transport.

Intermodal transport: Movement using multiple modes under coordinated planning.

Total landed cost: Purchase and logistics cost including freight, handling, inventory and related consequences.

Pipeline inventory: Goods physically in transit but not yet available for use or sale.

Transportation model: An optimization model allocating supply from origins to destinations.

Assignment problem: A model matching resources to tasks under one-to-one allocation conditions.

Vehicle-routing problem: A routing problem involving multiple vehicles and operational constraints.

Time window: The permitted interval for serving a customer or activity.

Cross-docking: Moving inbound goods to outbound shipment with little or no storage.

Slotting: Assigning warehouse storage locations to products.

Order-cycle time: Elapsed time between order release and a defined fulfillment milestone.

Perfect-order rate: Percentage of orders that are complete, timely, accurate and undamaged.

Last mile: The final delivery segment to the customer or pickup location.

Delivery density: Number of deliveries served per unit of route distance or time.

Reverse logistics: Management of product flows from customers back to recovery or disposal points.

Emissions factor: Emissions estimated per unit of activity, such as tonne-kilometer.

Control tower: A coordinated visibility and exception-management capability across logistics operations.

Micro-fulfillment center: A small, often urban facility designed for rapid order fulfillment.

Review Questions

Short Answer Questions

1. Why can the lowest freight rate fail to produce the lowest total landed cost?
2. What is the difference between a transportation model and an assignment model?
3. Identify four important warehouse-performance measures.
4. Why is last-mile delivery frequently expensive and variable?
5. What types of activities are included in reverse logistics?

Long Answer Questions

1. Compare road, rail, maritime and air transport for a high-value, time-sensitive product.
2. Explain the assumptions and constraints of the classical transportation model.
3. Discuss how warehouse slotting, fulfillment design and inventory positioning affect customer service.
4. Explain how a vehicle-routing model can incorporate capacity, time windows and driver-hour limits.
5. Evaluate the benefits and risks of AI-enabled logistics decision-making.

Analytical / Application Questions

1. A $75,000 shipment can travel by air in 3 days for $5,600 or by ocean in 21 days for $1,800 plus $250 handling. If the annual holding rate is 20%, calculate the relevant cost of each option using pipeline inventory cost and recommend an option before considering service penalties.
2. A delivery fleet serves 80 customers. Explain what additional data and constraints would be required to determine whether two larger vehicles or four smaller vehicles produce the better VRP solution.
3. Design a pilot for reducing failed last-mile deliveries in a dense urban zone. Specify baseline metrics, intervention, comparison method, service safeguards and emissions measures.

Multiple Choice Questions with Answers

Question 1

Which factor is included in total landed cost but not necessarily in the quoted freight rate?

A. Pipeline inventory cost

B. Customer name

C. Product color

D. Supplier logo

Correct Answer: A. Pipeline inventory cost

Question 2

In a balanced transportation model, total supply is:

A. Less than total demand

B. Equal to total demand

C. Always zero

D. Unrelated to demand

Correct Answer: B. Equal to total demand

Question 3

What distinguishes an assignment problem from a basic transportation problem?

A. It generally uses one-to-one resource-task allocation

B. It prohibits cost information

C. It applies only to maritime shipments

D. It cannot use binary variables

Correct Answer: A. It generally uses one-to-one resource-task allocation

Question 4

Which constraint ensures that a vehicle does not exceed its carrying capacity?

A. ∑ demand assigned to vehicle ≤ vehicle capacity

B. Total distance = 0

C. Number of depots = demand

D. Freight rate ≥ inventory value

Correct Answer: A. ∑ demand assigned to vehicle ≤ vehicle capacity

Question 5

Which measure is most directly associated with warehouse picking productivity?

A. Lines picked per labor hour

B. Annual sales revenue per customer

C. Number of suppliers per country

D. Fuel price per liter

Correct Answer: A. Lines picked per labor hour

Question 6

Why can warehouse utilization near 100% be undesirable?

A. It may cause congestion and reduce accessibility

B. It guarantees perfect-order performance

C. It eliminates replenishment

D. It always reduces labor cost

Correct Answer: A. It may cause congestion and reduce accessibility

Question 7

Which intervention can reduce failed home deliveries?

A. Offering lockers or pickup points

B. Removing all delivery notifications

C. Narrowing every delivery window

D. Ignoring apartment access constraints

Correct Answer: A. Offering lockers or pickup points

Question 8

In reverse logistics, disposition decisions may include:

A. Repair, resale, recycling or disposal

B. Only international shipping

C. Only forward production

D. Only demand forecasting

Correct Answer: A. Repair, resale, recycling or disposal

Question 9

If emissions are estimated as E = W × D × e, what does e represent?

A. An emissions factor

B. The number of customers

C. Vehicle capacity only

D. Warehouse rent

Correct Answer: A. An emissions factor

Question 10

Why should logistics AI systems retain human oversight?

A. Data and models may be biased, uncertain or unsafe in unusual conditions

B. Algorithms never use operational data

C. Human oversight always increases emissions

D. Optimization cannot process constraints

Correct Answer: A. Data and models may be biased, uncertain or unsafe in unusual conditions