Logistics, Order Fulfilment and Supply Network Integration

Introduction and Context

Logistics is the coordinated movement and positioning of materials, products, information and related financial transactions from suppliers to customers and, where necessary, back through the network. In a digitally enabled enterprise, logistics is no longer an isolated transport activity. It is connected to procurement, production planning, inventory management, sales, finance, customer service, compliance and sustainability reporting.

Order fulfilment is the process through which a customer requirement is translated into an available, delivered and financially completed order. It includes order capture, confirmation, inventory allocation, warehouse processing, transportation planning, delivery execution, invoicing and exception management. Supply network integration extends this process across suppliers, manufacturers, logistics service providers, distributors, marketplaces, customers and reverse-logistics partners.

Enterprise resource planning (ERP) systems provide the transactional backbone for these activities. Transportation management systems (TMS) optimise freight planning and execution; warehouse management systems (WMS) control warehouse operations; electronic data interchange (EDI) and application programming interfaces (APIs) exchange information; and control towers provide cross-network visibility and coordinated intervention.

The central managerial challenge is balance. The fastest delivery is not always the lowest-cost delivery; maximum inventory availability may increase obsolescence; a low freight rate may conceal poor reliability or environmental cost. Effective digital materials management therefore combines operational efficiency with service quality, resilience, governance and sustainability.

Learning Objectives

After studying this chapter, the learner should be able to:

1. Explain the relationship between logistics, order fulfilment and supply network integration.
2. Distinguish the roles of ERP, TMS, WMS, APIs, EDI and control towers.
3. Analyse transportation and distribution decisions using relevant cost and service measures.
4. Design an integrated order-to-delivery and returns process.
5. Evaluate third-party logistics (3PL) arrangements and integration risks.
6. Interpret fulfilment, visibility, resilience and sustainability metrics.
7. Assess governance, cybersecurity, data-quality and ethical issues in digitally connected logistics networks.
8. Recommend improvements using a balanced cost–speed–visibility–sustainability perspective.

1. Logistics as a Network Capability

Traditional logistics thinking focused on point-to-point movement: a shipment left one location and arrived at another. Modern supply networks are multi-echelon systems. A product may be sourced globally, assembled regionally, stored in several facilities, sold through physical and digital channels, delivered by a carrier and later returned to a repair or recycling centre.

Three flows must be coordinated:

• Physical flow: raw materials, components, finished goods, packaging, returns and waste.
• Information flow: orders, forecasts, inventory status, shipment milestones, proof of delivery and exception messages.
• Financial flow: purchase orders, freight charges, invoices, credits, duties, taxes and settlement.

Integration exists when these flows are connected sufficiently to support consistent decisions. Integration does not mean that every partner must use the same software. It means that systems have agreed identities, processes, data definitions, message standards, controls and responsibilities.

A useful distinction is between visibility and control. Visibility means knowing what is happening, such as a vehicle’s location or an order’s status. Control means having the authority, information and alternatives to change what is happening. A dashboard that identifies a late shipment but cannot trigger reallocation, customer communication or carrier escalation provides visibility without effective control.

Push, Pull and Postponement

A push system positions inventory in anticipation of demand, usually using forecasts. A pull system responds to actual customer or downstream demand. Most commercial networks use a hybrid approach. Standard, predictable products may be pre-positioned, whereas customised products may be completed after order receipt.

Postponement delays final configuration, packaging, labelling or geographic allocation until demand is clearer. It can reduce the risk of holding the wrong finished product, but it may increase handling time or require flexible production and information systems.

The Order-Fulfilment Cycle

The order-fulfilment cycle commonly includes:

1. customer order capture and validation;
2. availability and credit checks;
3. order promising and allocation;
4. wave, batch or discrete release to the warehouse;
5. picking, packing and documentation;
6. carrier selection and transport execution;
7. delivery confirmation and exception handling;
8. invoicing, settlement and performance measurement.

Order promising may use available-to-promise (ATP), capable-to-promise (CTP) or profitable-to-promise logic. ATP considers existing and planned inventory. CTP additionally considers production or procurement capacity, lead times and constraints. In both cases, the promise is only credible if inventory records, master data, capacity data and transport assumptions are reliable.

2. Digital Architecture for Integrated Logistics

ERP as the Transactional Backbone

An ERP system normally maintains core records and transactions such as materials, suppliers, customers, locations, purchase orders, sales orders, inventory movements, invoices and accounting entries. It establishes the commercial and financial consequences of logistics events.

However, ERP systems are not always designed to perform detailed route optimisation, real-time vehicle dispatch, dock scheduling or warehouse task sequencing. Specialist systems therefore complement ERP. Integration should preserve a clear system of record and avoid duplicate, conflicting transactions.

WMS and TMS

A WMS manages warehouse-level execution. Its functions may include location control, receiving, put-away, replenishment, picking, packing, cycle counting, labour assignment, wave planning and shipping confirmation.

A TMS manages transportation planning and execution. Common functions include shipment consolidation, mode and carrier selection, route planning, tendering, appointment scheduling, freight audit, tracking, claims and carrier performance analysis.

The relationship can be expressed operationally as follows:

• ERP creates or receives the order and establishes commercial rules.
• WMS confirms what can physically be picked and shipped.
• TMS determines how the shipment should move.
• Carriers and 3PLs provide execution events.
• ERP receives confirmation for inventory, customer service and financial settlement.

EDI and APIs

EDI exchanges structured business messages using agreed standards and transaction sets. Examples include purchase orders, order acknowledgements, advance shipping notices, invoices and functional acknowledgements. EDI is valuable where partners have stable, high-volume, standardised transactions and established compliance requirements.

APIs expose system functions or data through software interfaces. They support near-real-time queries and event exchange, such as checking inventory, requesting a freight quotation or receiving a delivery milestone. APIs may be more flexible than traditional batch EDI, but they require authentication, version control, monitoring, rate management and robust error handling.

Integration design should address message ownership, unique identifiers, timestamps, units of measure, time zones, status codes, retries, duplicate prevention and reconciliation. A message that arrives twice must not create two shipments or two invoices. This requires idempotent processing, transaction identifiers and exception queues.

Control Towers

A control tower is a coordinated visibility and decision-support capability that combines data from multiple systems and partners. It generally includes:

• data ingestion from ERP, WMS, TMS, carriers, suppliers, sensors and external services;
• common data models and event timelines;
• dashboards and alerts based on thresholds or predictive analysis;
• workflows for ownership, escalation and resolution;
• performance analytics and simulation.

A mature control tower moves through four levels:

1. Descriptive: What has happened?
2. Diagnostic: Why has it happened?
3. Predictive: What is likely to happen?
4. Prescriptive: What action should be taken?

A control tower should not become merely a “wall of screens.” Its value depends on data quality, decision rights, response playbooks and measurable business outcomes.

3. Transportation and Distribution Decisions

Transportation decisions involve mode, carrier, route, consolidation, frequency, service level and risk. Relevant costs include freight charges, fuel surcharges, handling, inventory in transit, damage, customs, administration, expediting and environmental externalities.

A simple total logistics cost model is:

where:

• TC = total logistics cost;
• F = freight and transport charges;
• H = handling and terminal costs;
• I = inventory carrying cost associated with goods in transit and cycle stock;
• D = damage, loss and claims cost;
• A = administrative and information-system cost;
• E = expediting, disruption and other exceptional cost.

The model is a managerial classification rather than a universal accounting identity. Its components must be defined consistently and should not double-count costs.

Freight Cost per Unit

For a shipment, freight cost per unit is:

where C_u is cost per unit, C_s is shipment freight cost in currency and Q is the number of chargeable units. The calculation is meaningful only when the units are comparable and the shipment cost includes the same surcharge and accessorial assumptions.

For example, a consolidated shipment costs $2,400 and carries 800 cartons. The freight cost is:

If an expedited shipment costs $3,600 for the same quantity, its cost is $4.50 per carton. The extra $1.50 per carton should be compared with the value of improved service, avoided stockout, contractual penalty or customer retention.

On-Time-In-Full and Perfect Order

On-time-in-full (OTIF) measures whether an order arrives by the agreed date and with the complete quantity. A common order-level definition is:

The definition must specify the promised date, delivery window, treatment of customer-caused delays, partial shipments and order lines. A line-level calculation may produce a different result from an order-level calculation.

A perfect order measure may additionally require accurate documentation, no damage and correct invoicing. Metrics are useful only when the organisation prevents gaming, publishes definitions and links results to corrective action.

Lead Time and Reliability

Lead time is the elapsed time between defined start and end events. For transportation, it may be the interval from carrier pickup to delivery. For fulfilment, it may begin at order receipt and end at customer acceptance. Averages alone can conceal variability, so organisations should also monitor percentiles, standard deviation or the proportion within the promised window.

For a set of observed lead times t₁, t₂, ..., t_n, the arithmetic mean is:

where n is the number of observations. This measure is appropriate for summarising central tendency, but it does not describe tail risk. A 95th-percentile lead time is often more useful when planning service commitments and safety capacity.

4. Third-Party Logistics and Network Governance

A 3PL performs logistics activities for a shipper, potentially including transport, warehousing, customs, fulfilment, freight payment and returns. A lead logistics provider or fourth-party logistics arrangement may coordinate several 3PLs and provide network design, procurement and control-tower services.

Make-or-Buy Evaluation

The choice between internal logistics and outsourcing should consider more than quoted rates. Relevant criteria include:

• total landed cost and cost transparency;
• service capability and geographic coverage;
• scalability during peaks;
• technology integration and data ownership;
• regulatory and security competence;
• asset flexibility and investment requirements;
• business continuity and financial stability;
• labour practices and sustainability performance;
• strategic dependence and switching cost.

Contracts should define service-level agreements (SLAs), key performance indicators (KPIs), liability, claims, insurance, subcontracting, data protection, audit rights, business continuity, exit support and continuous-improvement obligations. Incentives should reward overall customer and network outcomes rather than isolated local measures. For example, a carrier rewarded solely for low cost may select slow services that increase stockouts or customer complaints.

Network Design and Segmentation

There is no universally optimal distribution structure. Centralised inventory may reduce stock duplication but increase delivery distance and disruption exposure. Regional distribution may improve responsiveness but raise facility and inventory costs. The appropriate design depends on demand density, product value, product variety, service promises, customs barriers and risk.

Segmentation is especially important. Critical spare parts, hazardous materials, temperature-sensitive goods, high-value electronics and routine commodities should not necessarily receive identical transport policies. Service, security, packaging, monitoring and carrier requirements should reflect consequence of failure.

5. Returns and Reverse Logistics

Returns arise from damage, wrong shipment, customer dissatisfaction, warranty claims, repair, recall, reusable packaging and end-of-life recovery. Reverse logistics is not simply forward logistics in reverse. Return flows are uncertain in timing, condition, quantity and destination.

An effective returns process includes:

1. return authorisation and reason capture;
2. customer instructions and label generation;
3. collection or drop-off;
4. receipt and inspection;
5. disposition decision;
6. refund, replacement, repair or credit;
7. recovery, resale, recycling or responsible disposal;
8. root-cause analysis.

Disposition categories may include return to stock, refurbishment, vendor return, liquidation, material recovery and destruction. The economic value of recovery must be compared with inspection, transportation, processing, storage, data erasure and compliance costs.

Returns data is diagnostically valuable. A high return rate may indicate poor product quality, inaccurate product descriptions, packaging failure, picking errors or inappropriate customer expectations. Treating returns only as a cost misses their role as a source of product and process intelligence.

6. Sustainability and Responsible Logistics

Sustainable logistics considers emissions, energy, congestion, noise, packaging, labour conditions, safety and resource recovery. Decisions should use a life-cycle perspective rather than shifting impacts from one activity to another.

A basic transport-emissions estimate is:

where E is estimated emissions, W is transported mass, D is distance and EF is an emissions factor. If W is measured in tonnes, D in kilometres and EF in kilograms of CO₂-equivalent per tonne-kilometre, E is measured in kilograms of CO₂-equivalent.

This is an approximation. A robust calculation must specify vehicle type, fuel or electricity source, load factor, empty running, allocation method, geography, operational conditions and whether upstream fuel emissions are included. It should not be presented as a precise measurement when input data is uncertain.

Sustainability options include consolidation, improved load utilisation, modal shift, route optimisation, alternative fuels, electric vehicles for suitable routes, packaging redesign, renewable energy and reduced empty mileage. However, faster delivery and fragmented shipments may increase emissions. Sustainability must therefore be integrated into service and cost decisions rather than treated as a separate reporting exercise.

7. Data, Cybersecurity and Governance

Digitally integrated logistics creates operational dependence on data. Common data-quality problems include duplicate locations, incorrect dimensions, obsolete carrier codes, inconsistent units, missing time zones, inaccurate promised dates and unstructured exception reasons.

Governance should establish:

• ownership of master and transactional data;
• validation rules and approval workflows;
• data-retention and archival policies;
• audit trails and segregation of duties;
• access controls based on least privilege;
• encryption in transit and at rest;
• supplier and 3PL security requirements;
• incident response and recovery testing;
• reconciliation between operational and financial records.

Cybersecurity risk is particularly serious because logistics systems may control physical assets. A compromised interface could redirect shipments, alter quantities, expose customer information or interrupt warehouse operations. Resilience therefore requires backups, manual fallback procedures, tested recovery objectives and clear communication protocols.

Ethical issues include worker surveillance, algorithmic assignment of delivery work, unfair performance targets, privacy-invasive tracking and the unequal distribution of service quality across locations or customer groups. Digital efficiency must not be pursued without considering safety, dignity, accessibility and legal obligations.

8. Artificial Intelligence and Advanced Analytics

AI can support demand sensing, estimated-time-of-arrival prediction, carrier selection, anomaly detection, warehouse slotting, route optimisation and return classification. Its effectiveness depends on the quality and stability of historical data.

A predictive ETA model may identify that weather, congestion, route, carrier, shipment type and prior milestone delays influence arrival time. Yet prediction does not eliminate uncertainty. Human oversight is needed for unusual events, safety-critical decisions, regulatory requirements and customer communication.

Good practice includes model monitoring, explainability appropriate to the decision, bias testing, performance comparison with simple baselines, data-drift detection and documented override rules. AI should recommend actions within approved constraints; it should not independently bypass contractual, safety, customs or ethical controls.

9. Worked Application: An Integrated Omnichannel Network

Background

The following is an illustrative case. Meridian Home Systems sells appliances through dealers, its website and online marketplaces. It operates two regional distribution centres, uses three parcel carriers and a 3PL for bulky deliveries. Its ERP contains orders and financial records, the WMS manages picking, and the TMS manages freight execution. Returns are handled through email and spreadsheets.

Problem

Customer complaints have increased. Reported OTIF is 93%, but customers experience missed delivery appointments and incomplete installation kits. Carrier invoices are difficult to reconcile, return reasons are inconsistent, and managers cannot see orders delayed between warehouse release and carrier pickup. The company is considering faster delivery but is concerned about freight cost and emissions.

Analysis

A process review identifies four root causes:

1. The ERP promised date is calculated without considering carrier capacity.
2. The WMS sends shipment confirmation, but carrier pickup events are not returned consistently.
3. Product dimensions and installation-kit requirements are incomplete in master data.
4. Returns are measured by financial credit date rather than physical receipt and disposition.

The company also discovers that the 93% OTIF figure is order-line based, while several customers judge performance at appointment level. The metric therefore overstates perceived service quality.

Approach and Solution

Meridian implements the following design:

• The ERP remains the system of record for sales orders, customer commitments and billing.
• The WMS validates kit completeness and records packed dimensions.
• The TMS receives shipment attributes, obtains carrier capacity and returns pickup, in-transit and delivery events.
• APIs are used for real-time carrier milestones; EDI remains for high-volume purchase and invoice transactions.
• A control tower displays orders by risk: unallocated, packed-not-picked-up, delayed in transit, appointment risk and delivery exception.
• Exception ownership is assigned to warehouse, transport, customer service or the 3PL.
• Returns receive standard reason codes and disposition outcomes.
• The company introduces an emissions estimate by mode and shipment weight, while preserving service constraints for vulnerable customers and critical deliveries.

The new promise logic confirms a delivery window only after inventory, installation capability and carrier capacity are checked. For urgent orders, the system compares the incremental freight cost with customer priority and stockout consequences rather than automatically selecting premium transport.

Outcome

After three months, Meridian reports a reduction in packed-but-not-collected shipments, improved appointment visibility and faster return disposition. The organisation does not rely on the dashboard alone: weekly reviews examine root causes, carrier performance, data exceptions and customer complaints. Management concludes that the principal benefit is not simply faster delivery, but more reliable commitments and earlier intervention.

Learning Points

The case demonstrates that integration is a process and governance problem as much as a technology problem. Real-time data has limited value if events are not defined consistently, exceptions lack owners or master data is incomplete. It also shows why service metrics must reflect the customer’s experience rather than the easiest available system timestamp.

10. Implementation Roadmap and Best Practices

A practical implementation can proceed through the following stages:

Stage 1: Define the operating model

Clarify customer promises, network roles, decision rights, process owners, service segments and financial objectives. Document current-state processes and pain points.

Stage 2: Establish data foundations

Create common identifiers for products, locations, orders, shipments, packages, carriers and events. Standardise units, time zones, status codes and reason codes. Measure data completeness and accuracy.

Stage 3: Prioritise integration use cases

Begin with high-value, feasible use cases such as order status, inventory availability, carrier tendering, proof of delivery, freight audit or returns authorisation. Define measurable benefits and failure-handling procedures.

Stage 4: Pilot and validate

Use a limited geography, product family or carrier group. Test normal flows, duplicates, missing messages, late events, cancellations, partial shipments and system outages.

Stage 5: Scale with governance

Expand only after operational ownership, cybersecurity, training, support and performance measurement are established. Review partner capability and contract alignment.

Stage 6: Improve continuously

Use process mining, root-cause analysis, simulation and controlled experiments. Compare cost, speed, reliability, visibility, sustainability and employee or customer outcomes together.

A balanced logistics scorecard may include:

No single measure should dominate. For example, reducing freight cost while increasing late deliveries may destroy value. Likewise, achieving excellent visibility without intervention capability may create reporting activity rather than performance improvement.

Challenges and Limitations

Integrated logistics faces several persistent challenges. First, organisations may inherit fragmented legacy systems and incompatible partner capabilities. Second, real-time visibility can create false confidence when location data is inferred, delayed or incomplete. Third, optimisation models may produce locally efficient decisions that harm the broader network. Fourth, global disruptions, weather, geopolitical events and labour shortages can invalidate historical assumptions.

There are also organisational limits. Functions may protect their own KPIs, while logistics outcomes depend on shared decisions. Sales may promise dates without transport capacity; procurement may select suppliers without considering delivery variability; finance may reject resilience investment because benefits are difficult to assign to one budget.

Finally, digital systems do not remove managerial judgement. They structure information and automate repeatable decisions, but unusual events require experienced interpretation, ethical reasoning and accountability.

Chapter Summary

Logistics, order fulfilment and supply network integration connect physical movement, information exchange and financial settlement across an extended network. ERP systems provide transactional control, while WMS and TMS support warehouse and transport execution. EDI and APIs enable structured or near-real-time collaboration, and control towers combine visibility with exception management and coordinated action.

Effective transportation decisions balance freight, handling, inventory, damage, administrative, disruption and environmental costs. Service measures such as OTIF and perfect order must be precisely defined and aligned with customer experience. 3PL relationships require careful evaluation of cost, capability, technology, risk, governance and exit arrangements. Reverse logistics should be designed as a value-recovery and learning process rather than treated as an administrative afterthought.

Digital integration creates opportunities for predictive analytics and AI, but also increases dependence on data quality, cybersecurity, privacy and sound governance. Sustainable logistics requires attention to emissions, load utilisation, mode, packaging, labour and life-cycle effects. The most effective network is not necessarily the fastest or cheapest in isolation; it is the one that delivers an appropriate balance of cost, speed, visibility, resilience, responsibility and customer value.

Key Terms

API: A software interface that allows systems to request data or services and exchange events.

Available-to-promise (ATP): A commitment based on available and planned inventory.

Control tower: An integrated visibility and decision-support capability for monitoring and coordinating a supply network.

Electronic data interchange (EDI): Structured computer-to-computer exchange of standard business documents.

Exception management: Identification, ownership, escalation and resolution of events outside expected conditions.

Fulfilment: The activities required to receive, process, deliver and financially complete a customer order.

Inventory in transit: Stock that has left one location but has not yet been received at its destination.

Lead logistics provider: A coordinating organisation that manages multiple logistics providers and broader network activities.

On-time-in-full (OTIF): A measure of delivery by the agreed time and with the complete required quantity.

Order promising: The process of determining whether and when a customer order can be fulfilled.

Postponement: Delaying final product configuration or allocation until demand is clearer.

Reverse logistics: The movement and management of returns, repairs, reusable assets, recovery and end-of-life materials.

Service-level agreement (SLA): A contractual specification of expected service performance and responsibilities.

Third-party logistics (3PL): Outsourced logistics services provided to a shipper or supply-network participant.

Transportation management system (TMS): Software for planning, executing, tracking and analysing transportation.

Warehouse management system (WMS): Software for controlling warehouse inventory and operational tasks.

Visibility: The ability to observe relevant inventory, order, shipment and event information.

Perfect order: An order delivered on time, in full, without damage and with accurate documentation, according to the adopted definition.

Review Questions

Short Answer Questions

1. Distinguish physical, information and financial flows in logistics.
2. What are the primary differences between EDI and APIs?
3. Define a control tower and explain why visibility alone is insufficient.
4. Identify four activities normally managed by a WMS.
5. Why should returns data be analysed for root causes?

Long Answer Questions

1. Explain how ERP, WMS, TMS, EDI, APIs and control towers can be integrated in an order-to-delivery process.
2. Discuss the criteria a company should use when selecting a 3PL provider.
3. Analyse the trade-offs between centralised and regional distribution networks.
4. Explain how sustainability objectives can be incorporated into transportation and fulfilment decisions.
5. Evaluate the governance, cybersecurity and ethical risks of digitally integrated logistics.

Analytical / Application Questions

1. A company has high freight cost but poor OTIF performance. Design a diagnostic framework that separates transport, warehouse, master-data and order-promise causes.
2. Design a control-tower dashboard for a firm using two warehouses, four carriers and an external returns provider. Specify events, alerts, owners and escalation rules.
3. Using the transport-emissions model presented in this chapter, identify the data needed to compare road and rail alternatives, and explain the limitations of the resulting estimate.

Multiple Choice Questions with Answers

Question 1

Which system is primarily responsible for detailed transport planning, carrier tendering and shipment execution?

A. Transportation management system

B. General ledger

C. Product lifecycle management system

D. Human resource management system

Correct Answer: A. Transportation management system

Question 2

What is the principal distinction between visibility and control in a supply network?

A. Visibility concerns cost, while control concerns revenue

B. Visibility observes conditions, while control enables authorised intervention

C. Visibility applies only to warehouses, while control applies only to carriers

D. Visibility is retrospective, while control is always predictive

Correct Answer: B. Visibility observes conditions, while control enables authorised intervention

Question 3

A shipment costs $2,400 and contains 800 chargeable cartons. What is the freight cost per carton?

A. $2.00

B. $2.50

C. $3.00

D. $3.50

Correct Answer: C. $3.00

Question 4

Which activity is most directly associated with a warehouse management system?

A. Corporate income-tax calculation

B. Picking and location control

C. Foreign-exchange hedging

D. Product advertising design

Correct Answer: B. Picking and location control

Question 5

Which statement best describes postponement?

A. Increasing safety stock at every network location

B. Delaying final configuration until demand is clearer

C. Replacing all transport with local delivery

D. Eliminating customer order validation

Correct Answer: B. Delaying final configuration until demand is clearer

Question 6

Under the order-level OTIF definition, an order is counted as successful when it is:

A. Invoiced before the promised date

B. Picked, regardless of delivery quantity

C. Delivered on time and in full

D. Dispatched using the cheapest carrier

Correct Answer: C. Delivered on time and in full

Question 7

Which feature is most characteristic of an API integration?

A. Near-real-time service or event interaction

B. Paper-based document approval

C. Manual re-entry of every shipment

D. A requirement that all partners use one ERP instance

Correct Answer: A. Near-real-time service or event interaction

Question 8

What is a major risk of rewarding a carrier solely for the lowest freight rate?

A. It guarantees maximum network resilience

B. It may encourage slow service and increase wider supply-chain cost

C. It eliminates the need for performance measurement

D. It ensures complete shipment visibility

Correct Answer: B. It may encourage slow service and increase wider supply-chain cost

Question 9

In the simplified emissions model E = W × D × EF, what does EF represent?

A. Freight invoice value

B. Facility energy consumption

C. Emissions factor

D. Expected fulfilment time

Correct Answer: C. Emissions factor

Question 10

Which practice best supports safe and accountable use of AI in logistics?

A. Allowing the model to bypass customs controls

B. Removing all human overrides

C. Monitoring performance, data drift and approved override rules

D. Using historical data without checking for bias

Correct Answer: C. Monitoring performance, data drift and approved override rules