Integrated Planning under Uncertainty

Introduction

Facility decisions are often presented as separate choices: where to locate a plant, how to arrange departments, how much capacity to install, how much inventory to hold, and how to distribute products through a network. In practice, these decisions are strongly interdependent. A low-cost location may have high disruption exposure; a compact layout may reduce handling cost but leave little room for expansion; high capacity may protect service levels but create financial risk when demand is uncertain.

Integrated planning treats the facility and supply network as a system rather than as a collection of isolated projects. The central managerial problem is to design a configuration that performs acceptably across a range of possible futures. This requires attention to demand variability, capacity flexibility, inventory positioning, expansion timing, disruption risk, business continuity, and resilience.

Uncertainty is not simply a lack of information. It may arise from random variation, incomplete knowledge, structural change, or rare disruptive events. Different forms of uncertainty require different responses. Statistical forecasting can help with routine demand variation, whereas redundancy, postponement, and alternate sourcing are more appropriate for disruption risk.

Learning Objectives

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

1. Explain why location, layout, capacity, inventory, and network decisions must be integrated.
2. Distinguish risk, uncertainty, variability, and ambiguity.
3. Apply scenario-based and expected-value approaches to strategic facility decisions.
4. Evaluate expansion options, flexibility, redundancy, and resilience.
5. Design an integrated response to demand and disruption uncertainty.
6. Critically assess the limitations, governance requirements, and ethical implications of integrated planning.

1. The Integrated Planning Problem

A facility network contains several decision layers:

The layers interact. Location influences transport time, labor availability, utility reliability, tax exposure, and disruption risk. Layout influences material-handling distance, safety, throughput, and the ease of installing additional equipment. Capacity influences inventory and outsourcing requirements. Inventory can compensate for transportation or production uncertainty, but it consumes space and capital. Network design determines whether a disruption at one facility can be absorbed elsewhere.

A useful integrated objective is to minimize total relevant cost while satisfying service and risk requirements:

Here, C_F is fixed facility and location cost, C_T transportation cost, C_P production or operating cost, C_I inventory cost, C_L shortage, delay, or lost-sales cost, and C_R risk-mitigation cost. All terms must be measured over the same planning horizon and expressed in comparable monetary units. The formulation is a decision model, not a universal law: regulatory, environmental, service, and safety constraints may be more important than cost minimization.

A constrained version can be stated as:

The variable vector x represents location, layout, capacity, inventory, sourcing, and contingency decisions. S_min is the minimum service requirement, E_max an environmental limit, and R_max an acceptable risk threshold. The feasible set X includes engineering, legal, labor, and operational restrictions.

2. Understanding Uncertainty

2.1 Variability, risk, uncertainty, and ambiguity

Variability is the natural fluctuation of a measurable process, such as daily demand or delivery time. Risk refers to uncertain outcomes for which probabilities can reasonably be estimated. Uncertainty, in a narrower sense, concerns outcomes for which probabilities are incomplete or unreliable. Ambiguity exists when decision makers disagree about how a situation should be interpreted—for example, whether a new technology will replace or complement an existing process.

These distinctions matter because a safety-stock calculation may address demand variability, while a second supplier or geographically separate facility may address disruption risk. Neither response is sufficient for every type of uncertainty.

2.2 Sources of uncertainty in facility planning

Important sources include:

• demand volume, mix, seasonality, and geographical distribution;
• product life-cycle changes and technological substitution;
• supplier reliability, lead time, quality, and financial health;
• labor availability, wage changes, and industrial relations;
• energy prices, utilities, transport access, and infrastructure reliability;
• exchange rates, tariffs, taxation, and regulatory change;
• natural hazards, cyberattacks, pandemics, conflict, and political instability;
• construction delays, permitting uncertainty, and equipment obsolescence.

Uncertainty should be documented in an assumptions register. Each assumption should identify its source, confidence level, decision relevance, owner, and review date.

2.3 Scenario thinking

Scenario planning does not attempt to predict one precise future. It constructs a small number of internally consistent futures and tests whether a decision remains effective across them. Scenarios should vary the drivers that matter most, such as demand growth, geographic demand, energy cost, disruption frequency, and technology adoption.

A practical scenario set might include:

Scenarios must not be arbitrary stories. They should be supported by historical data, expert judgment, macroeconomic analysis, hazard maps, supplier information, and sensitivity testing.

3. Integrated Design Logic

3.1 Location and network configuration

Location analysis should include both expected operating conditions and exposure to adverse events. Common criteria include market access, supplier proximity, labor, utilities, land, tax and regulation, transport reliability, environmental impact, security, and community acceptance.

A location with the lowest nominal cost may not be the best choice if a flood, port closure, or political event can interrupt the entire network. A more robust assessment uses risk-adjusted total cost and service consequences.

For scenario s, the discounted total cost can be represented as:

where C_s,t is cost in period t, B_s,t is benefit or avoided loss, r is the discount rate, and T is the planning horizon. The units are monetary units at the chosen price basis. The discount rate, inflation treatment, tax treatment, and salvage value must be consistent.

If scenario probabilities p_s are credible, expected value may be calculated as:

Expected value is useful for comparing repeated or aggregatable risks, but it can be misleading for catastrophic events, where probability estimates are weak and risk tolerance is asymmetric. In such cases, decision makers should also examine worst-case loss, conditional loss, recovery time, and service failure.

3.2 Layout as a flexibility decision

A layout is not merely a drawing of departments. It determines travel distance, congestion, visibility, safety, storage density, utility access, and the ability to modify the process. Under uncertainty, layout quality includes option value—the value of retaining feasible future choices.

Design principles include:

• place high-frequency, high-volume flows close together where safety permits;
• separate incompatible materials and processes;
• provide clear emergency access and evacuation routes;
• reserve utility capacity and connection points for expansion;
• use modular workstations, movable equipment, and standardized interfaces;
• avoid bottlenecks that become critical under peak demand;
• distinguish permanent infrastructure from temporary or scalable resources.

A layout that minimizes current distance may be inferior to one with slightly higher current cost but better expansion speed and lower disruption consequence.

3.3 Capacity and expansion

Capacity decisions should distinguish base capacity, flexible capacity, and surge capacity. Base capacity supports normal demand. Flexible capacity can be adjusted through shifts, labor arrangements, overtime, cross-training, or subcontracting. Surge capacity is reserved for peaks or emergencies.

If demand is uncertain, a staged expansion strategy may dominate a single large investment. The organization first installs capacity that supports credible near-term demand, while preserving land, utilities, layout space, financing, and permits for later expansion. Expansion triggers should be defined in advance, such as sustained utilization above a threshold, confirmed customer contracts, or a forecast distribution exceeding available capacity.

Utilization is commonly expressed as:

where D is required output per period and C is effective capacity per period. Effective capacity is the realistic output after planned maintenance, breaks, changeovers, quality losses, and staffing constraints. Utilization is dimensionless. High utilization is not automatically desirable: excessive utilization reduces recovery capacity and increases queueing, overtime, and vulnerability.

3.4 Inventory positioning

Inventory can buffer demand, production, transportation, and supply uncertainty. However, inventory requires space, handling, capital, insurance, and control. A facility strategy should therefore decide not only how much inventory to hold but also where to hold it.

For a simple continuous-review item with approximately normal demand during lead time, the reorder point is:

where μ_L is mean demand during lead time, σ_L is the standard deviation of demand during lead time, and z is the standard-normal safety factor corresponding to the desired cycle-service probability. Units of μ_L and σ_L must be inventory units. This expression assumes that the demand distribution and lead-time behavior are reasonably stable and that replenishment quantity, review policy, and lost-sales or back-order treatment are appropriately specified.

If daily demand is independent with mean μ_d and standard deviation σ_d, and lead time is a fixed L days, then:

If both demand and lead time vary independently, with lead-time mean μ_L and standard deviation σ_L, a commonly used approximation is:

This approximation excludes the interaction term and assumes independence. If demand and lead time are correlated, or distributions are highly skewed, simulation or an empirically estimated lead-time-demand distribution is preferable.

3.5 Risk pooling and postponement

Pooling demand across locations can reduce aggregate variability when demands are not perfectly correlated. If independent demand streams have standard deviations σ₁ and σ₂, the standard deviation of their sum is:

For correlated demands with correlation coefficient ρ:

The expression is valid when the variances and correlation are appropriately estimated. Lower correlation generally increases the benefit of pooling. A centralized inventory may therefore require less safety stock than two fully independent inventories, although transportation time, service differentiation, and disruption exposure may offset the benefit.

Postponement delays product differentiation until demand information is clearer. For example, a facility may hold generic components centrally and complete final configuration near the customer. This reduces finished-goods variety and can improve flexibility, but it requires compatible product architecture, reliable information, and suitable finishing capability.

4. Resilience, Flexibility, and Business Continuity

4.1 Resilience dimensions

Resilience is the capability to prepare for, absorb, recover from, and adapt to disruptive events. It is broader than redundancy. Important dimensions include:

• robustness: performance remains acceptable during disturbance;
• redundancy: alternative capacity, suppliers, routes, or facilities exist;
• resourcefulness: the organization can mobilize people, information, and funds;
• rapidity: recovery occurs within an acceptable time;
• adaptability: the system learns and changes after disruption.

A resilient network does not necessarily duplicate every asset. It combines selective redundancy with visibility, flexible contracts, modular design, emergency procedures, and collaborative relationships.

4.2 Business continuity planning

A continuity program connects risk analysis to operational action. It should identify critical products and processes, maximum tolerable downtime, recovery time objectives, recovery point objectives for information, dependencies, alternate resources, communication protocols, and authority levels.

A business impact analysis asks:

1. Which processes are essential to customers, safety, revenue, and compliance?
2. How long can each process be unavailable?
3. Which utilities, suppliers, systems, people, and transport links are dependencies?
4. What minimum operating mode is acceptable?
5. What resources are required to restore normal performance?

Continuity plans must be exercised. A plan that exists only as a document may fail because contact data, assumptions, equipment, or responsibilities have changed.

4.3 Recovery and service metrics

Useful measures include recovery time, fill rate, order cycle time, time to restore capacity, percentage of demand served during disruption, and cumulative lost output. A simple resilience loss measure is:

where P₀ is normal performance, P(t) is performance at time t, and T_R is recovery time. The integral represents lost performance over the disruption interval; its units depend on the performance measure, such as unit-hours or revenue-hours. In practice, performance may be recorded at discrete intervals and summed.

5. Decision Models for Integrated Planning

5.1 Multi-criteria evaluation

Strategic facility decisions include factors that cannot be reduced reliably to money. A multi-criteria model may score alternatives on cost, service, resilience, sustainability, workforce, and regulatory fit. A weighted score is:

where v_jk is the normalized performance of alternative j on criterion k, and w_k is the criterion weight with Σw_k = 1. Scores are useful for structuring judgment, not for disguising subjective assumptions. Sensitivity analysis should test whether rankings change when weights or scores change.

5.2 Robust and stochastic approaches

A stochastic model represents uncertain parameters with probability distributions and optimizes expected or risk-adjusted performance. A robust model seeks a solution that performs acceptably across a defined uncertainty set, even without reliable probabilities. Robustness can be expressed through constraints such as ensuring capacity under every selected scenario:

This may produce a more expensive design than an expected-value solution, but it protects against unfavorable conditions. The correct approach depends on the consequences of failure, the quality of data, and the organization's risk appetite.

5.3 Real-options perspective

An expansion option has value because it preserves the right, but not the obligation, to invest later. A company may pay more initially for land, utilities, or a modular building in order to avoid irreversible commitment. The value of an option is higher when uncertainty is high, future investment can be staged, and waiting provides useful information.

Real-options reasoning does not mean delaying every decision. Delay can lose market share, increase construction cost, or create service failure. The managerial task is to compare the value of flexibility with the cost of preserving it.

6. Worked Example: A Regional Distribution Network

An illustrative company is evaluating two alternatives for a regional distribution operation. Alternative A is a large centralized facility. Alternative B consists of two smaller facilities. The alternatives have the following annual equivalent costs under three scenarios.

The expected costs are:

On expected cost alone, B is preferred by 0.25 million annually. Its worst-case cost is also lower. However, the conclusion depends on the credibility of the probabilities and the assumptions behind the disruption costs. If the disruption probability were much lower, or if coordination and inventory duplication were underestimated, A might be preferable.

The company should therefore conduct sensitivity analysis, verify transport and labor assumptions, model inventory positioning, test service levels by customer region, and examine whether the two sites are genuinely independent. Two facilities located in the same floodplain or dependent on the same supplier are not full redundancy.

A staged solution may also be considered: establish one main facility with a leased overflow site, preserve land or building rights for a second node, and contract emergency capacity with a qualified third-party provider. This solution may have lower initial commitment while preserving response options.

7. Applied Case: NorthStar Medical Devices

This is an illustrative case developed for educational purposes.

Background

NorthStar Medical Devices produces temperature-sensitive diagnostic kits. Demand is growing, but product mix is uncertain because hospitals are adopting different testing platforms. The existing plant is located near a major airport and has an efficient functional layout. Its sterilization supplier, however, is concentrated in a region exposed to severe storms. The plant operates at 82% effective capacity during normal periods and has limited space for additional equipment.

Problem

Management must decide whether to build a larger single plant, add a smaller regional finishing site, or expand inventory and outsource selected activities. The decision must maintain regulatory compliance, protect product quality, and provide continuity during supplier or facility disruption.

Analysis

The team maps the network from critical suppliers to final customers and classifies activities by postponability. Sterilization and core component production are difficult to relocate quickly, while labeling and final configuration can be performed at qualified regional sites. Demand scenarios are developed using hospital-contract information, historical variability, and expert judgment.

The analysis identifies four vulnerabilities:

1. the single sterilization source;
2. high utilization during seasonal peaks;
3. insufficient protected space for new equipment;
4. dependence on one airport route for urgent shipments.

A purely cost-focused expansion would increase capacity but leave the critical supplier and transport dependencies unchanged.

Approach and solution

NorthStar adopts a portfolio solution:

• qualify a second sterilization supplier in a different geographic risk zone;
• install modular finishing cells in the existing plant;
• lease a small regional finishing facility rather than immediately constructing a second full plant;
• hold generic components centrally and postpone final labeling where feasible;
• reserve floor space, utilities, and clean-room interfaces for expansion;
• maintain a controlled emergency inventory for critical components;
• establish disruption triggers and a quarterly continuity exercise.

The layout is redesigned around product families and material segregation. Movable cells allow the company to shift labor and equipment between product types. A digital dashboard monitors supplier lead time, capacity utilization, inventory coverage, weather alerts, and quality deviations.

Outcome

The solution costs more than a single-site expansion under stable demand, but it reduces dependence on one supplier, lowers the time needed to add finishing capacity, and improves service during regional disruption. The company also avoids prematurely building a second full-scale plant before product mix uncertainty is resolved.

Learning points

The case demonstrates that resilience is usually a system property. A new building alone cannot solve supplier concentration, layout rigidity, transport dependence, or poor information. The best solution combines network diversification, modular capacity, postponement, inventory positioning, and governance.

8. Technology, Data, and AI

Digital tools can improve integrated planning when they are connected to sound operational reasoning. Network optimization models can compare facility, transport, inventory, and capacity alternatives. Discrete-event simulation can test congestion, queueing, staffing, and recovery sequences. Digital twins can represent changing demand, flows, equipment states, and disruption scenarios.

AI and machine learning may support demand sensing, anomaly detection, supplier-risk monitoring, and predictive maintenance. Their value depends on data quality, model validation, interpretability, and human oversight. A forecast model trained on stable historical conditions may fail during structural change. Automated recommendations should therefore show assumptions, confidence ranges, and scenario consequences rather than presenting one number as fact.

Data governance is essential. Facility and supply-chain data may include commercially sensitive information, worker information, security details, and regulated product data. Access controls, audit trails, cybersecurity, retention rules, and human accountability must be designed into the system.

9. Governance, Ethics, and Sustainability

Integrated planning has consequences beyond the firm. Location decisions affect employment, traffic, land use, emissions, water demand, and community risk. A facility may be economically attractive because environmental or social costs are shifted to local residents. Responsible decisions include stakeholder consultation, environmental impact assessment, worker safety, accessibility, fair labor standards, and transparent evaluation of externalities.

Resilience investments also raise distributional questions. Should scarce continuity capacity be allocated to the most profitable customers, the most vulnerable communities, or critical public services? In healthcare and essential goods, service priorities should be established ethically and transparently before a crisis occurs.

Sustainability should be integrated rather than treated as a final score. Relevant considerations include life-cycle carbon emissions, energy resilience, renewable integration, building adaptability, packaging, reverse logistics, water, waste, and the carbon effects of inventory and redundancy. A second facility may improve continuity but increase embodied carbon; the appropriate evaluation is comparative and life-cycle based.

10. Implementation Framework

A disciplined implementation process can follow these stages:

1. Define the decision boundary. Specify facilities, products, time horizon, service requirements, and irreversible choices.
2. Map the system. Document suppliers, flows, customers, dependencies, information, utilities, and critical assets.
3. Establish the uncertainty register. Separate variability, measurable risk, deep uncertainty, and ambiguity.
4. Develop scenarios. Use evidence and test combinations of demand, disruption, cost, and technology conditions.
5. Generate alternatives. Include location, layout, capacity, inventory, sourcing, postponement, and continuity options.
6. Evaluate total performance. Compare cost, service, resilience, environmental impact, safety, and implementability.
7. Test sensitivity and failure modes. Identify assumptions that could reverse the recommendation.
8. Select triggers and contingencies. Define expansion thresholds, emergency actions, decision rights, and escalation paths.
9. Pilot and validate. Use simulation, layout trials, supplier qualification, and continuity exercises.
10. Monitor and learn. Review assumptions, leading indicators, near misses, and actual performance.

Leading indicators include forecast error, supplier concentration, capacity utilization, inventory coverage, recovery time in exercises, percentage of critical suppliers qualified, and the number of expansion decisions that remain feasible.

Challenges and Limitations

Integrated models can create false precision. A result stated to two decimal places may depend on uncertain probabilities, subjective weights, or incomplete cost data. Models may also overlook organizational behavior, political constraints, learning curves, and the difficulty of coordinating multiple facilities.

Risk pooling assumptions can fail when disruptions are correlated. Two geographically separate sites may share the same power grid, cloud provider, raw-material source, or transport corridor. Redundancy can create complexity and quality variation. Excessive inventory can hide process problems and become obsolete. Flexible labor arrangements may create fatigue or inequity if poorly governed.

The remedy is not to abandon modeling. It is to combine quantitative analysis with field observation, expert review, stress testing, transparent assumptions, and post-decision learning. Decision makers should distinguish what the model calculates from what management must judge.

Chapter Summary

Integrated planning aligns location, network, layout, capacity, inventory, and continuity decisions over multiple time horizons. Uncertainty may arise from ordinary variability, measurable risk, deep uncertainty, or ambiguity; each requires different managerial responses. Scenario planning, sensitivity analysis, stochastic models, robust constraints, and real-options reasoning help decision makers avoid overcommitment to one forecast.

Resilient systems combine robustness, redundancy, resourcefulness, rapid recovery, and adaptability. Modular layouts, staged expansion, postponement, inventory positioning, supplier diversification, alternate routes, and continuity exercises are practical mechanisms for creating flexibility. Expected cost is useful but insufficient when probabilities are weak or consequences are catastrophic. Service, safety, sustainability, ethics, and regulatory requirements must be included explicitly.

Effective implementation requires a clear decision boundary, an uncertainty register, evidence-based scenarios, integrated alternatives, stress testing, predefined triggers, governance, and continuous learning. Technology and AI can strengthen analysis, but only when data quality, cybersecurity, interpretability, and human accountability are maintained.

Key Terms

Ambiguity: Uncertainty caused by competing interpretations or meanings.

Base capacity: Capacity intended to support normal operating demand.

Business continuity: Capability to maintain or restore critical operations after disruption.

Capacity flexibility: Ability to adjust available output through shifts, labor, equipment, outsourcing, or process changes.

Demand variability: Natural fluctuation in demand over time or across customers.

Disruption: An event that materially interrupts normal operations or flows.

Effective capacity: Realistic output after allowing for operational losses and constraints.

Inventory positioning: The decision about how much inventory to hold and where to locate it.

Layout flexibility: Ability to modify physical arrangements, equipment, and flows as conditions change.

Network resilience: Capability of a network to absorb, recover from, and adapt to disruption.

Option value: Value of preserving a future choice without being obliged to exercise it.

Postponement: Delaying final product differentiation until better demand information is available.

Risk pooling: Reduction of aggregate variability by combining demands or resources.

Robustness: Ability to maintain acceptable performance under adverse conditions.

Scenario planning: Evaluation of decisions across internally consistent possible futures.

Sensitivity analysis: Examination of how results change when assumptions or parameters vary.

Stochastic model: A model that represents uncertain inputs or outcomes probabilistically.

Utilization: Required output divided by effective capacity.

Review Questions

Short Answer Questions

1. Why should location and layout decisions be evaluated together?
2. Distinguish variability, risk, uncertainty, and ambiguity.
3. What is the purpose of staged capacity expansion?
4. Explain how postponement can reduce exposure to product-mix uncertainty.
5. Identify four dimensions of supply-network resilience.

Long Answer Questions

1. Discuss how facility location, capacity, inventory, and transportation decisions interact in an uncertain market.
2. Compare expected-value, robust, and scenario-based approaches to strategic facility planning.
3. Explain how layout design can create or destroy future flexibility.
4. Evaluate the role of redundancy, risk pooling, and supplier diversification in business continuity.
5. Discuss the ethical, environmental, and governance issues associated with integrated facility planning.

Analytical / Application Questions

1. A firm is considering one centralized facility versus two regional facilities. Design a scenario table and identify the cost, service, and resilience measures required before making the decision.
2. Daily demand has a mean of 100 units and a standard deviation of 20 units. Lead time is fixed at 4 days, and the desired cycle-service factor is z = 1.645. Using the simple fixed-lead-time approximation, calculate the reorder point and state the assumptions.
3. Develop an integrated continuity plan for a plant operating at 85% utilization with one critical supplier, one major transport route, and limited expansion space. Specify layout, capacity, inventory, sourcing, and governance actions.

Multiple Choice Questions with Answers

Question 1

Which statement best describes integrated facility planning?

A. Selecting the lowest-rent site independently of operations

B. Coordinating location, layout, capacity, inventory, and network decisions

C. Maximizing utilization regardless of service consequences

D. Replacing all physical facilities with suppliers

Correct Answer: B. Coordinating location, layout, capacity, inventory, and network decisions

Question 2

Which uncertainty response is most directly associated with routine demand variability?

A. Safety stock based on lead-time demand

B. Geographic duplication of every facility

C. Political lobbying

D. Permanent closure of excess sites

Correct Answer: A. Safety stock based on lead-time demand

Question 3

In the expression u = D/C, what does C represent?

A. Customer count

B. Carrying cost

C. Effective capacity per period

D. Correlation coefficient

Correct Answer: C. Effective capacity per period

Question 4

What is the main purpose of postponement?

A. To increase finished-goods variety everywhere

B. To delay differentiation until demand information improves

C. To eliminate all inventory

D. To maximize early customization

Correct Answer: B. To delay differentiation until demand information improves

Question 5

If two independent demand streams have standard deviations 3 and 4 units, their aggregate standard deviation is:

A. 1 unit

B. 5 units

C. 7 units

D. 12 units

Correct Answer: B. 5 units

Question 6

Why can expected-value analysis be inadequate for catastrophic disruption risk?

A. It cannot include any costs

B. It always selects the most expensive alternative

C. Probabilities may be unreliable and losses may be highly asymmetric

D. It applies only to layout decisions

Correct Answer: C. Probabilities may be unreliable and losses may be highly asymmetric

Question 7

Which feature most directly creates layout flexibility?

A. Fixed equipment with no spare utility connections

B. Modular cells and reserved expansion interfaces

C. Maximum storage density in every area

D. Single-purpose corridors that cannot be changed

Correct Answer: B. Modular cells and reserved expansion interfaces

Question 8

A robust capacity constraint across scenarios requires capacity to be:

A. At least demand in every selected scenario

B. Equal to average demand only

C. Less than demand during disruption

D. Independent of all scenarios

Correct Answer: A. At least demand in every selected scenario

Question 9

Which is a business-continuity activity?

A. Ignoring recovery time until a disruption occurs

B. Identifying critical processes and maximum tolerable downtime

C. Removing all emergency procedures after implementation

D. Using one untested supplier for efficiency

Correct Answer: B. Identifying critical processes and maximum tolerable downtime

Question 10

What is a key governance risk when using AI for facility planning?

A. Models can never process operational data

B. Automated outputs may conceal assumptions and produce poorly monitored decisions

C. AI eliminates the need for scenario analysis

D. AI makes cybersecurity irrelevant

Correct Answer: B. Automated outputs may conceal assumptions and produce poorly monitored decisions