Logistics Costing, Performance and Analytics

Introduction and Context

Logistics decisions are often evaluated too narrowly. A transport manager may select the lowest freight rate, a warehouse manager may minimize labour cost, or a procurement team may choose the lowest supplier price. Yet each decision can increase inventory, handling, customs, stockout, quality, carbon, or customer-service costs elsewhere in the supply chain. Logistics costing therefore concerns more than recording expenditure: it is the disciplined assignment, analysis, and use of cost and performance information to support integrated decisions.

The central managerial question is not simply, “What does logistics cost?” It is, “What resources are consumed to provide which level of service to which customer, through which supply-chain design, and with what financial, operational, and environmental consequences?” This question links activity-based costing, cost-to-serve, landed cost, total-cost analysis, performance measurement, and analytics.

At graduate level, logistics costing should be understood as a decision system. It must distinguish accounting cost from economic cost, fixed from variable cost, average from marginal cost, and visible from hidden cost. It must also recognize that measures can create dysfunctional behaviour. For example, a warehouse measured only on cost per order may reduce labour hours by consolidating orders, while damaging delivery reliability.

Learning Objectives

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

1. Explain the role and limitations of logistics cost information.
2. Apply activity-based costing to logistics activities and cost objects.
3. Calculate customer and product cost-to-serve.
4. Construct landed-cost and total-cost models.
5. Design a balanced logistics KPI system covering cost, service, assets, quality, and carbon.
6. Distinguish descriptive, predictive, and prescriptive logistics analytics.
7. Evaluate dashboards, data governance, artificial intelligence, and ethical risks.
8. Interpret logistics performance information for managerial decisions.

1. The Architecture of Logistics Cost

1.1 Cost objects, cost pools, and cost drivers

A cost object is anything for which cost is measured: a shipment, order, product, customer, route, facility, channel, or service proposition. A cost pool is a collection of costs associated with a common activity or resource. A cost driver is a measurable factor that explains changes in activity consumption.

Traditional costing commonly allocates overhead using broad averages such as units sold, sales value, or direct labour hours. This may be acceptable where products and customers consume resources similarly. It becomes misleading when orders differ substantially in size, frequency, urgency, delivery geography, returns, or service requirements.

Activity-based costing (ABC) addresses this problem by tracing resources to activities and activities to cost objects. Typical logistics activities include receiving, put-away, storage, replenishment, picking, packing, dispatch, transport, customs processing, returns handling, customer service, and inventory planning.

A useful distinction is:

• Direct cost: traceable economically to a cost object, such as a dedicated delivery charge.
• Indirect cost: shared across objects, such as warehouse rent.
• Fixed cost: unchanged within a relevant capacity range when activity changes.
• Variable cost: changes with activity, such as a per-kilometre toll.
• Step cost: fixed over a range but increases when capacity is expanded, such as adding a shift.
• Relevant cost: differs between alternatives and therefore affects a decision.
• Sunk cost: already incurred and not recoverable; it should not influence a forward-looking choice.
• Opportunity cost: contribution sacrificed by using a scarce resource for one alternative rather than another.

These categories are not always absolute. Warehouse rent may be fixed in the short run but avoidable over a multi-year network redesign. The correct classification depends on the decision horizon and the feasible alternatives.

1.2 Activity-based costing

For an activity cost pool, the activity rate is:

The cost assigned to a cost object is:

The cost pool and driver quantity must use compatible units. For example, an annual receiving cost measured in currency must be divided by receiving hours, pallets received, or purchase orders processed—not by an unrelated measure unless a defensible causal relationship exists. Practical capacity is generally preferable to theoretical capacity because it recognizes unavoidable downtime, maintenance, breaks, and normal disruption.

ABC is most useful when three conditions hold: activities are sufficiently heterogeneous, resource consumption differs across cost objects, and the information will improve a decision. It is not automatically superior. Excessively detailed activity models can be expensive to maintain and can create false precision.

Worked example: warehouse ABC

A distribution centre has the following annual costs:

The activity rates are:

• Receiving: $240,000 ÷ 12,000 = $20 per pallet.
• Picking: $540,000 ÷ 180,000 = $3 per order line.
• Packing: $180,000 ÷ 90,000 = $2 per parcel.
• Dispatch administration: $120,000 ÷ 30,000 = $4 per shipment.

Customer A generates 600 pallets received, 8,000 order lines, 4,000 parcels, and 1,000 shipments. Its assigned warehouse cost is:

• Receiving: 600 × $20 = $12,000.
• Picking: 8,000 × $3 = $24,000.
• Packing: 4,000 × $2 = $8,000.
• Dispatch administration: 1,000 × $4 = $4,000.

Total assigned cost is $48,000. The calculation reveals why shipment volume alone would be a weak allocation basis: Customer A consumes several different activities at different intensities.

2. Cost-to-Serve Analysis

2.1 Concept and purpose

Cost-to-serve (CTS) measures the resources required to serve a particular customer, segment, channel, product family, or order profile. It combines commercial revenue with operational consumption. A customer with high revenue may be unprofitable if it requires small frequent deliveries, extensive manual order changes, special packaging, expedited transport, high returns, and prolonged payment terms.

A simplified customer contribution equation is:

When the analysis focuses only on logistics:

The analysis should specify whether costs are fully allocated or only incremental. Fully allocated CTS supports long-run pricing and network decisions. Incremental CTS supports short-run choices such as accepting an additional order when capacity already exists. Confusing these purposes can lead to poor decisions.

2.2 Building a cost-to-serve model

A robust CTS project generally follows these steps:

1. Define the decision and the cost object.
2. Segment customers or orders by meaningful service characteristics.
3. Map the order-to-delivery process.
4. Identify activities, resources, and cost drivers.
5. Separate avoidable, unavoidable, fixed, variable, and opportunity costs.
6. Assign costs using causal relationships.
7. Validate results with operational managers and customers.
8. Test scenarios such as minimum order quantities, delivery frequency, price changes, or service redesign.

A customer-level model should include logistics costs that are often omitted: delivery failures, claims, returns, redelivery, order amendments, pallet exchange, special labelling, customer-specific inventory, and the working-capital cost of inventory.

2.3 Interpretation and managerial use

CTS is not merely a tool for terminating “unprofitable” customers. It can identify mutually beneficial redesign opportunities. Examples include delivery-day consolidation, electronic ordering, standard packaging, revised minimum order quantities, regional stocking, differentiated service tiers, or customer collection. A negative current contribution may become positive after a lower-cost service proposition is agreed.

A critical issue is joint cost and causality. A warehouse may serve multiple customers simultaneously, and transport costs may be shared across loads. Allocating every cost to individual customers can create arbitrary results. Managers should therefore disclose allocation rules, sensitivity ranges, and costs that remain common rather than presenting uncertain estimates as facts.

3. Landed Cost and Total-Cost Analysis

3.1 Landed cost

Landed cost estimates the cost of acquiring and making a product available at a specified destination. A generic structure is:

The exact components depend on the transaction, trade terms, tax treatment, and destination. Recoverable value-added tax may be excluded from economic cost, whereas nonrecoverable taxes should be included. Currency conversion must use a stated exchange-rate assumption and date. Incoterms® determine which party bears specified obligations, risks, and costs, but they do not by themselves determine every accounting treatment or ownership issue.

Landed cost is often calculated per unit:

The denominator must be carefully defined. If 1,000 units are ordered but only 980 are saleable after damage or quality rejection, a decision model may require cost per saleable unit rather than cost per ordered unit.

Worked landed-cost example

A firm imports 1,000 units at $40 each. Origin charges are $1,200, international freight is $3,500, insurance is $300, duty is 6% of the customs value, destination charges are $900, and inland delivery is $1,100. Assume the customs value is purchase value plus origin charges plus international freight and insurance, and that all listed amounts are nonrecoverable costs.

Purchase value = 1,000 × $40 = $40,000.

Customs value = $40,000 + $1,200 + $3,500 + $300 = $45,000.

Duty = 0.06 × $45,000 = $2,700.

Total landed cost = $40,000 + $1,200 + $3,500 + $300 + $2,700 + $900 + $1,100 = $49,700.

Unit landed cost = $49,700 ÷ 1,000 = $49.70 per unit.

The calculation is valid only under the stated customs and tax assumptions. Different jurisdictional rules or trade terms could change the result.

3.2 Total cost of ownership and total logistics cost

Landed cost is an acquisition view. Total-cost analysis extends the boundary to costs caused by a decision over its relevant life or operating horizon. For a sourcing decision, this may include quality failures, lead-time variability, pipeline inventory, safety stock, obsolescence, supplier management, financing, carbon charges, and disposal.

A general comparative model is:

Not every term belongs in every decision. The model should avoid double counting. For example, if a supplier price already includes delivery, freight should not be added again. Similarly, expected disruption cost should be represented either through a probability-weighted cost or through scenario analysis, not both without justification.

Inventory carrying cost can be estimated as:

The carrying-rate percentage may include the cost of capital, insurance, shrinkage, obsolescence, and storage. It is an approximation, not a universal physical constant. If average inventory is $500,000 and the annual carrying rate is 20%, estimated annual carrying cost is $100,000.

Total-cost analysis is especially important when comparing transport modes. Air freight may have a higher freight charge but lower pipeline inventory, less safety stock, or fewer stockouts. Conversely, a low-cost ocean option may create long lead times and high uncertainty.

4. Logistics Performance Measurement

4.1 From isolated measures to a performance system

A KPI is useful only when its purpose, owner, definition, data source, frequency, target, and response are explicit. A performance system should connect strategic objectives to operational measures without encouraging local optimization.

Five dimensions are particularly important:

1. Cost: logistics cost as a percentage of sales, cost per order, cost per unit shipped, transport cost per tonne-kilometre, and cost-to-serve.
2. Service: on-time in-full (OTIF), order-cycle time, fill rate, perfect-order rate, and customer complaint response time.
3. Assets: inventory turns, days of inventory, warehouse utilization, vehicle utilization, and cash-to-cash cycle time.
4. Quality and reliability: picking accuracy, damage rate, claims rate, return rate, forecast error, and process capability where appropriate.
5. Carbon and sustainability: emissions per shipment, emissions per tonne-kilometre, energy per order, empty-running percentage, waste, and renewable-energy share.

A balanced system includes leading and lagging indicators. Delivery failures are lagging outcomes; dock congestion, backlog age, and forecast bias may be leading signals.

4.2 Definitions and formulas

A common OTIF measure is:

The organization must define “on time,” “in full,” treatment of customer-requested delays, partial shipments, and order lines versus orders. Different definitions produce noncomparable results.

Inventory turnover is commonly calculated as:

The numerator and denominator should cover the same period and valuation basis. Days of inventory is approximately:

For a 365-day year, inventory of $2 million and annual cost of goods sold of $12 million give 60.83 days.

A basic perfect-order measure may require correct product, correct quantity, no damage, complete documentation, and on-time delivery. It is stricter than OTIF and should not be interpreted as interchangeable with it.

Carbon accounting requires a defined boundary and emission factor. A simplified activity-based estimate is:

For transport, activity data might be tonne-kilometres and the emission factor kilograms of CO₂e per tonne-kilometre. The result depends on vehicle type, fuel, load factor, route, allocation method, and whether well-to-tank emissions are included. Reported emissions should state methodology and data quality.

4.3 Dashboard design

A dashboard is not a collection of attractive charts. It is an interface for prioritizing decisions. Effective dashboards:

• show a limited number of strategically relevant KPIs;
• distinguish actual, target, forecast, and prior-period values;
• expose trends and variance, not only current status;
• permit drill-down from network to facility, route, customer, and order;
• display definitions and data timestamps;
• identify owners and escalation thresholds;
• show trade-offs, such as cost versus OTIF or utilization versus lead time.

A useful dashboard architecture has three levels. The executive level shows strategic outcomes. The management level shows drivers and exceptions. The operational level supports daily intervention. A dashboard should not make a decision automatically merely because a threshold is crossed; context and data quality must be reviewed.

5. Analytics for Logistics Decisions

5.1 Descriptive analytics

Descriptive analytics answers “What happened?” It includes data cleaning, aggregation, variance analysis, segmentation, Pareto analysis, control charts, route visualization, and performance trends. For example, a logistics team may discover that 20% of customers generate 65% of returns or that late deliveries cluster on two routes and one carrier.

Descriptive analysis must distinguish correlation from causation. A correlation between congestion and late delivery does not establish whether congestion caused delay, whether both resulted from peak demand, or whether a third variable such as staffing explains both.

5.2 Predictive analytics

Predictive analytics answers “What is likely to happen?” Applications include demand forecasting, estimated-time-of-arrival prediction, stockout probability, failure prediction, and return-risk scoring.

A simple moving-average forecast is:

where F_t+1 is the forecast for the next period, D_t is demand in the current period, and n is the number of periods averaged. This method assumes recent observations are informative and gives equal weight to them. It performs poorly when strong trend, seasonality, structural change, or intermittent demand is present.

Forecast accuracy can be assessed using measures such as mean absolute error:

where D_t is actual demand, F_t is forecast demand, and n is the number of observations. MAE is expressed in demand units and is less sensitive to large errors than squared-error measures. No accuracy measure is universally best; the choice should reflect decision purpose and data characteristics.

5.3 Prescriptive analytics

Prescriptive analytics answers “What should we do?” It combines forecasts, constraints, objectives, and decision variables. Typical applications include inventory positioning, vehicle routing, load consolidation, workforce scheduling, facility location, mode selection, and order allocation.

A simplified transport-mode model can be written as:

subject to capacity, delivery-time, legal, service, and operational constraints. The objective function is not inherently correct; management must decide whether carbon is monetized, constrained, or reported separately. An optimization result is only as credible as its data, assumptions, objective weights, and feasible solution space.

Analytics maturity should progress from reliable descriptive data to validated prediction and then to controlled prescription. Automating an inaccurate process merely accelerates bad decisions.

6. Integrated Case: Northstar Home Appliances

Background

Northstar Home Appliances, an illustrative regional distributor, serves retail chains, independent stores, and direct-to-consumer customers. Management reports strong revenue but declining operating margin. The company measures freight cost per shipment and warehouse cost per unit, while customer service is measured separately through complaint counts.

Problem

The retail-chain segment places large planned orders but demands extensive compliance labelling. Independent stores place frequent small orders and often request changes after order cut-off. Direct-to-consumer orders have high packing and returns costs. The existing average allocation makes all segments appear similarly profitable.

Analysis

Northstar maps activities and identifies receiving pallets, order lines, parcels, shipments, delivery kilometres, returns, and customer-service contacts as relevant drivers. ABC shows that independent stores consume disproportionate order-processing and delivery activity. CTS then adds returns, redelivery, claims, and account-management time. Landed-cost analysis identifies a supplier whose low purchase price is offset by long lead times, higher safety stock, and damage.

The KPI review finds that freight cost per shipment has improved because shipments are being consolidated, but OTIF has fallen. Inventory turnover has also declined. The dashboard had encouraged cost reduction without showing service and asset effects.

Approach and solution

Northstar introduces three service propositions:

• scheduled consolidated delivery for independent stores;
• compliant, forecast-based delivery windows for retail chains;
• a standardized parcel and returns policy for direct customers.

Pricing is revised using customer-level CTS bands rather than a single average logistics surcharge. A cross-functional dashboard reports CTS, OTIF, inventory days, damage, returns, and emissions per order. Predictive analytics forecasts late deliveries using route, carrier, weather, backlog, and depot variables, while planners retain authority to override predictions with documented reasons.

Outcome

In the illustrative scenario, small-order delivery frequency falls, customer-level contribution improves, and the company accepts slightly higher average freight cost in exchange for improved OTIF and lower emergency shipments. The principal outcome is not a single cost reduction but better alignment between service promises, resource consumption, and customer economics.

Learning points

1. Average cost concealed heterogeneous resource consumption.
2. Customer profitability required operational rather than purely financial data.
3. A low local KPI can worsen system performance.
4. Service redesign can be more constructive than customer exit.
5. Analytics must support managerial judgement rather than replace accountability.

7. Governance, Ethics, and Limitations

Cost and performance systems influence behaviour and therefore have ethical consequences. If a carrier is rewarded only for low cost, it may use unsafe schedules or underreport service failures. If warehouse workers are assessed only on speed, picking accuracy and safety may deteriorate. If carbon metrics exclude subcontracted transport or empty return legs, environmental performance may be overstated.

Important governance practices include:

• documented metric definitions and version control;
• role-based access to customer, employee, and supplier data;
• audit trails for data changes and manual overrides;
• separation of duties for rate approval and performance reporting;
• privacy protection and data minimization;
• transparent model documentation;
• bias testing across customers, regions, and carriers;
• human review for consequential automated decisions;
• scenario testing for disruption and model failure.

AI systems can classify delivery exceptions, predict demand, optimize routes, and generate narrative explanations. However, logistics data may contain missing scans, inconsistent master data, changing carrier practices, and historical biases. Predictive performance can degrade under disruption or structural change. Managers should monitor drift, establish fallback procedures, and avoid presenting probabilistic outputs as certainty.

Carbon accounting also presents boundary and allocation disputes. A shipment can be allocated emissions by mass, volume, distance, revenue, or marginal impact; each method answers a different question. Claims should therefore state boundaries, factors, assumptions, and uncertainty.

8. Best-Practice Implementation Road Map

A practical implementation can proceed in stages:

Stage 1: Clarify decisions. Identify which pricing, sourcing, network, capacity, or service decisions require better information.

Stage 2: Establish data foundations. Standardize customer, product, location, shipment, activity, rate, and time definitions. Reconcile operational data with financial ledgers.

Stage 3: Build a minimum viable cost model. Start with material activities and defensible drivers. Test results against known totals.

Stage 4: Integrate performance. Link cost with service, assets, quality, safety, and carbon. Assign owners and action thresholds.

Stage 5: Analyze scenarios. Test service tiers, mode changes, inventory policies, network alternatives, and disruption assumptions.

Stage 6: Validate and institutionalize. Review with finance, operations, sales, procurement, customers, and sustainability specialists. Document limitations and update the model when processes change.

The objective is not maximum model complexity. It is decision-relevant accuracy, transparency, timeliness, and organizational adoption.

Chapter Summary

Logistics costing is an integrated managerial discipline that connects resource consumption to activities, cost objects, service propositions, and supply-chain decisions. Activity-based costing improves visibility where customers and products consume logistics activities differently. Cost-to-serve extends this logic to customer profitability and can reveal redesign opportunities. Landed cost supports acquisition decisions, while total-cost analysis includes inventory, quality, risk, compliance, carbon, and end-of-life consequences within a defined decision boundary.

A balanced KPI system should cover cost, service, assets, quality, and carbon, with precise definitions and clear ownership. Dashboards should expose trends, variance, drivers, and trade-offs rather than encourage isolated optimization. Descriptive analytics explains past performance, predictive analytics estimates likely outcomes, and prescriptive analytics evaluates actions subject to objectives and constraints. All three depend on reliable data, explicit assumptions, governance, and human accountability.

Key Terms

Activity-based costing: A costing method that assigns resource costs to activities and then to cost objects using causal drivers.

Cost object: An item, customer, order, product, route, or service for which cost is measured.

Cost pool: A group of costs associated with a common activity.

Cost driver: A measurable factor that explains activity consumption or cost variation.

Cost-to-serve: The cost of providing a defined logistics and service proposition to a customer or segment.

Landed cost: The acquisition cost of making goods available at a specified destination.

Total-cost analysis: Evaluation of all relevant costs and benefits within a defined decision boundary and time horizon.

Relevant cost: A future cost that differs among alternatives.

Opportunity cost: The value of the best alternative sacrificed by using a resource.

OTIF: The percentage of due orders delivered on time and in full under a stated definition.

Perfect order: An order meeting specified requirements for accuracy, completeness, condition, documentation, and timeliness.

Inventory turnover: Cost of goods sold divided by average inventory value.

Cost-to-serve segmentation: Grouping customers or orders by their patterns of resource consumption and service requirements.

Descriptive analytics: Analysis of historical data to explain what happened.

Predictive analytics: Methods used to estimate future outcomes or probabilities.

Prescriptive analytics: Analysis that recommends actions under objectives and constraints.

Practical capacity: Usable capacity after allowing for normal downtime and operating limitations.

Carbon intensity: Emissions associated with a defined unit of logistics activity or output.

Data governance: Policies and controls ensuring data quality, security, accountability, and appropriate use.

Review Questions

Short Answer Questions

1. What is the difference between a cost pool and a cost driver?
2. Why can freight cost per shipment be a misleading logistics KPI?
3. Distinguish landed cost from total-cost analysis.
4. What does OTIF measure, and why must its definition be documented?
5. State one limitation of predictive analytics in logistics.

Long Answer Questions

1. Explain how activity-based costing can improve customer profitability analysis in a distribution network.
2. Discuss the assumptions and governance requirements of a customer cost-to-serve model.
3. Compare landed-cost analysis with total-cost-of-ownership analysis for international sourcing.
4. Design a balanced logistics KPI framework covering cost, service, assets, quality, and carbon.
5. Explain how descriptive, predictive, and prescriptive analytics can be integrated into transport management.

Analytical / Application Questions

1. A warehouse has annual receiving costs of $300,000 and practical capacity of 15,000 pallets. Customer X receives 800 pallets. Calculate the receiving activity rate and the cost assigned to Customer X. State the assumptions required.
2. A firm is comparing two transport modes. Mode A has lower freight cost but longer and more variable lead time; Mode B has higher freight cost but lower inventory and stockout risk. Develop a total-cost model and identify the data required before making a recommendation.
3. A dashboard shows falling cost per shipment, falling OTIF, and rising inventory days. Diagnose at least three possible causes and propose a cross-functional corrective-action plan.

Multiple Choice Questions with Answers

Question 1

Which feature most directly distinguishes activity-based costing from broad-average overhead allocation?

A. It assigns all overhead equally to every product.

B. It traces resources through activities using causal cost drivers.

C. It excludes warehouse overhead from product cost.

D. It uses revenue as the only allocation basis.

Correct Answer: B. It traces resources through activities using causal cost drivers.

Question 2

A cost pool of $240,000 has practical capacity of 12,000 pallets. What is the activity rate?

A. $5 per pallet

B. $20 per pallet

C. $200 per pallet

D. $2,880 per pallet

Correct Answer: B. $20 per pallet

Question 3

Which cost is most likely to be included in customer cost-to-serve but omitted from a simple freight-rate analysis?

A. Product selling price

B. Corporate share capital

C. Returns and redelivery costs

D. Supplier purchase quantity

Correct Answer: C. Returns and redelivery costs

Question 4

Under the assumptions in the chapter’s worked example, what is the unit landed cost for total landed cost of $49,700 and 1,000 units?

A. $40.00

B. $45.00

C. $49.70

D. $56.70

Correct Answer: C. $49.70

Question 5

Which measure is generally classified as an asset-related logistics KPI?

A. Inventory turnover

B. Picking accuracy

C. Customer complaint count

D. On-time delivery percentage

Correct Answer: A. Inventory turnover

Question 6

What does descriptive analytics primarily answer?

A. What happened?

B. What will happen with certainty?

C. What should be done under constraints?

D. Who is legally liable?

Correct Answer: A. What happened?

Question 7

Which statement about a KPI is most appropriate?

A. A KPI needs no documented definition if it appears on a dashboard.

B. A KPI should have a purpose, owner, data source, target, and response.

C. A KPI should measure only financial outcomes.

D. A KPI is valid only when it is calculated daily.

Correct Answer: B. A KPI should have a purpose, owner, data source, target, and response.

Question 8

If average inventory is $2 million and annual cost of goods sold is $12 million, approximate days of inventory using a 365-day year.

A. 16.67 days

B. 30.42 days

C. 60.83 days

D. 182.50 days

Correct Answer: C. 60.83 days

Question 9

Which is a principal risk of automating a logistics decision using an unvalidated predictive model?

A. It eliminates all data requirements.

B. It may accelerate inaccurate or biased decisions.

C. It guarantees better service performance.

D. It removes the need for human accountability.

Correct Answer: B. It may accelerate inaccurate or biased decisions.

Question 10

In the simplified emissions relationship, emissions are calculated as:

A. Activity data divided by revenue

B. Activity data multiplied by an emission factor

C. Revenue multiplied by inventory turnover

D. Freight cost minus insurance cost

Correct Answer: B. Activity data multiplied by an emission factor