Resource-Constrained Scheduling and Critical Chain

Introduction

Traditional Critical Path Method (CPM) assumes that the resources required by activities are available whenever they are needed. In practice, this assumption is frequently false. A project may have a single specialist engineer, limited construction equipment, a constrained testing laboratory, or a fixed budget for concurrent work. Consequently, activities that are logically independent may compete for the same resource, and the schedule may become infeasible even though its network logic is correct.

Resource-constrained scheduling extends network analysis by recognising that project duration is determined by both precedence relationships and resource availability. Critical Chain Project Management (CCPM) goes further by integrating resource dependencies, uncertainty management, and strategically placed time buffers. This chapter examines resource loading, allocation, levelling, smoothing, resource-critical schedules, heuristic prioritisation, and the design and management of project, feeding, and resource buffers.

Learning Objectives

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

1. Explain resource loading, allocation, levelling, and smoothing.
2. Distinguish logic-critical activities from resource-critical activities.
3. Identify resource conflicts and construct feasible resource-constrained schedules.
4. Apply practical prioritisation heuristics to resolve competing activities.
5. Explain the assumptions and limitations of Critical Chain Project Management.
6. Design and interpret project, feeding, and resource buffers.
7. Evaluate the managerial, technological, and governance implications of resource-constrained scheduling.

1. From Logic-Driven to Resource-Feasible Scheduling

A precedence network specifies which activities must precede others. If activity A must finish before activity B begins, the network represents a technological or managerial dependency. CPM calculates early and late dates from these dependencies and identifies the logic-critical path—the longest path through the network.

A resource-constrained schedule adds resource limits. Let an activity i require an amount rᵢ of a resource during time period t, and let the available capacity be Rₜ. A schedule is resource-feasible when:

for every resource type and every time period in which the activities are performed. The summation includes only activities active during period t. The inequality must be checked separately for each resource category, such as labour, equipment, facilities, or specialist capability. Resource units must be consistent: for example, engineers per day must be compared with engineers available per day.

A feasible schedule must therefore satisfy two conditions:

• Precedence feasibility: no activity starts before all required predecessors are complete.
• Capacity feasibility: resource demand does not exceed available capacity.

The shortest schedule satisfying both conditions is often called a resource-constrained project schedule. Unlike the ordinary CPM problem, it may require delaying an activity that has zero calculated float in the original logic-only schedule, or delaying an activity with positive float because another activity has higher priority.

2. Resource Terminology and Data Requirements

Resource loading

Resource loading is the process of assigning resource demand to time periods. A resource-loaded schedule shows not merely when an activity occurs, but how much labour, equipment, money, or capacity it consumes while occurring. A common representation is a histogram, with time on the horizontal axis and resource demand on the vertical axis.

For a resource with constant demand rᵢ over activity duration dᵢ, total resource effort is:

where Wᵢ is work in resource-days, rᵢ is resource units, and dᵢ is duration in days. This equation assumes constant resource intensity and a linear relationship between assigned resources and effort. It does not imply that adding resources will proportionally shorten the activity.

Resource allocation

Resource allocation assigns available resources to activities. Allocation decisions include who performs work, which equipment is used, and when a scarce capability is committed. Allocation may be constrained by skill, location, working time, contractual rules, safety requirements, or authorisation limits.

Resource levelling

Resource levelling modifies activity dates, and sometimes project duration, to remove over-allocation. It prioritises feasibility over preservation of the original completion date. Levelling may delay activities beyond their original late dates and therefore extend the project.

Resource smoothing

Resource smoothing attempts to reduce fluctuations in resource demand while keeping the project completion date and, usually, activity logic within available float. Smoothing is therefore less disruptive than levelling but has less power. If the resource profile cannot be made feasible without exceeding float, levelling is required.

3. Resource Profiles and Over-Allocation

Suppose three activities are scheduled concurrently:

The demand is 4 analysts during days 1–2 and 5 analysts during days 3–4 if the stated periods overlap. If only three analysts are available, the schedule is infeasible. The problem is not a precedence error; it is a capacity violation.

The project manager must decide whether to delay B or C, split an activity, use another resource, change the working calendar, outsource work, or negotiate additional capacity. Each option affects cost, risk, quality, and stakeholder commitments.

A useful resource-loading analysis includes:

• activity identifiers and precedence relationships;
• duration and planned dates;
• resource types and quantities;
• calendars, shifts, holidays, and availability;
• skill or substitution rules;
• maximum capacity by period;
• assumptions about productivity and interruption;
• approved constraints and required milestones.

Resource data must be realistic. Treating a part-time specialist as a full-time resource creates a mathematically neat but operationally impossible schedule.

4. Logic-Critical and Resource-Critical Schedules

In a logic-only CPM schedule, the critical path is determined by precedence and duration. Its total float is commonly computed as:

where ES and EF are early start and early finish, LS and LF are late start and late finish, and TF is total float. The relationship assumes a consistent CPM network and a defined project completion date.

A logic-critical activity has zero or near-zero float because of network relationships. A resource-critical activity is critical because resource competition, capacity limits, or a resource-dependent sequence determines project completion. An activity may be logic-critical, resource-critical, both, or neither.

Resource criticality is dynamic. When one activity is delayed to resolve a conflict, the delay can create a new longest path. Thus, the critical path in a resource-constrained schedule should be recalculated after major scheduling changes. It is unsafe to assume that the original CPM path remains the controlling path.

Resource dependencies can be represented as additional precedence relationships. If two activities require the same unique specialist and cannot overlap, the scheduler must sequence them. Choosing A before B creates an effective dependency A → B; choosing B before A creates B → A. The network has thereby acquired a decision-dependent relationship.

5. Heuristic Prioritisation Under Scarcity

The general resource-constrained project scheduling problem is computationally difficult for realistic networks. Exact optimisation may be impractical when there are many activities, multiple resources, calendars, and alternative modes. Project organisations therefore use heuristics: rules that produce good and explainable solutions without guaranteeing mathematical optimality.

At each decision point, the scheduler identifies eligible activities—those whose predecessors are complete—and ranks them. Common priority rules include:

A practical composite score can support transparent judgement. For example, a team may rank activities using a weighted score:

where Sᵢ is the priority score for activity i; Pᵢ represents precedence or downstream importance; Uᵢ represents urgency; Rᵢ represents risk or contractual consequence; Vᵢ represents value; and w₁ through w₄ are agreed weights. The terms must be normalised to comparable scales before being combined, and the weights should be documented. This is a decision framework, not a universal optimisation formula.

A robust scheduling procedure is:

1. Calculate the logic-only schedule.
2. Identify resource over-allocations by period and resource type.
3. List eligible activities at each conflict point.
4. Apply an agreed priority rule or composite decision rule.
5. Delay, split, substitute, or re-mode lower-priority work.
6. Recalculate dates, floats, resource profiles, and completion risk.
7. Conduct a management review of cost, quality, safety, and stakeholder effects.

Heuristics should not be applied mechanically. A low-float activity may be less important than a safety inspection, regulatory deliverable, or activity protecting a major contractual milestone.

6. Worked Example: Levelling a Specialist Resource

Consider the following project network. Activities A and B begin at project start. C follows A, D follows B, and E follows both C and D.

Assume only one specialist is available. In a logic-only schedule, A and B could start together, but that would require two specialists. Thus, the schedule is infeasible.

Two principal sequencing alternatives exist.

Alternative 1: A before B. A occupies days 1–3, B days 4–5, C days 6–7, and D may begin only after B, occupying days 6–9. E follows both C and D and occurs on day 10. Project duration is 10 working days.

Alternative 2: B before A. B occupies days 1–2, A days 3–5, D can occur during days 3–6, C occurs days 6–7, and E occurs on day 8. Project duration is 8 working days.

The second alternative is shorter because sequencing B first allows D to progress while A and C are completed. The original logic-only critical path does not determine the answer; the scarce specialist creates a new sequencing decision. If A had a regulatory or contractual priority, however, Alternative 1 might remain preferable despite its longer duration.

The example demonstrates three principles:

• a resource conflict can create an effective precedence relationship;
• the order selected at a conflict can change total project duration;
• a shorter schedule is not automatically the best schedule when risk, value, or compliance is considered.

7. Critical Chain Project Management

Critical Chain Project Management, associated particularly with Eliyahu M. Goldratt’s Theory of Constraints, defines the critical chain as the longest sequence of dependent activities considering both precedence and resource dependencies. It differs from conventional CPM by treating resource contention as a fundamental determinant of the controlling sequence.

CCPM also challenges the practice of embedding large safety allowances in every activity estimate. Such allowances may be consumed through student syndrome—delaying work until a deadline approaches—Parkinson’s law—allowing work to expand to the available time—and frequent multitasking. CCPM commonly uses shorter, more focused activity estimates and aggregates protection into buffers.

The central managerial logic is:

1. identify the chain constrained by precedence and resources;
2. protect the promised project completion date with a project buffer;
3. protect the critical chain from non-critical paths with feeding buffers;
4. protect scarce resources from becoming unavailable through resource buffers or alerts;
5. manage buffer consumption rather than demanding that every activity finish exactly to its estimate.

CCPM is not simply a different calculation of float. It is a scheduling and control philosophy involving estimates, behavioural rules, buffer monitoring, and prioritised execution.

Project buffer

The project buffer is placed at the end of the critical chain, between the last chain activity and the committed project completion date. It absorbs variation in critical-chain work and protects the external commitment.

Feeding buffer

A feeding buffer is placed where a non-critical chain joins the critical chain. It protects the critical chain from delays on that feeding path. A project can have several feeding buffers, because several non-critical paths may merge into the critical chain.

Resource buffer

A resource buffer is not normally a block of time inserted into the network. It is an advance warning, readiness mechanism, or mobilisation signal intended to ensure that a critical-chain resource will be available when needed. Examples include a notification before a specialist is required, a pre-approved equipment reservation, or a readiness checklist.

Figure: A critical chain protected by a feeding buffer and a project buffer; the resource buffer is a readiness mechanism rather than necessarily a time block.

8. Buffer Sizing and Monitoring

There is no universally valid buffer-sizing formula. Buffer size depends on estimate reliability, activity variability, correlation of risks, resource stability, and the organisation’s tolerance for lateness. Several practices are used, including aggregation of safety removed from activities, statistical analysis of uncertainty, and simple proportional rules.

A frequently used heuristic is the square-root aggregation of independent safety allowances:

where B is the aggregate buffer and sᵢ is the safety allowance associated with activity or uncertainty source i. This expression is appropriate only as a heuristic when the components are treated as approximately independent and represented on compatible time scales. If uncertainties are correlated, the simple formula can understate risk. It is not a law of CCPM and should not be presented as a guarantee.

Buffer management tracks actual buffer consumption against work completed. A simple penetration measure is:

where BP is buffer penetration, Cᵦ is buffer consumed, and B is total buffer size. Both quantities must use the same time units. A high BP with little corresponding progress indicates deteriorating schedule health. Organisations may define green, amber, and red response zones, but thresholds should be calibrated using historical performance rather than adopted uncritically.

Buffer monitoring should trigger action, not merely reporting. Actions may include removing multitasking, expediting inputs, reallocating specialists, resolving decisions, or protecting the remaining chain from additional work.

9. Uncertainty, Estimation, and Behavioural Assumptions

CCPM’s effectiveness depends heavily on the quality of estimates and execution discipline. Shorter estimates are not automatically more accurate. If estimates are unrealistically compressed, the project buffer becomes a concealed contingency for systematic underestimation rather than a protection against natural variation.

Important assumptions include:

• activities can be started and finished with focused effort;
• work is not continuously interrupted by unrelated priorities;
• resources can be coordinated across the chain;
• management will avoid using buffers as hidden milestones for individual blame;
• task estimates represent achievable work durations under agreed conditions;
• dependencies and acceptance criteria are sufficiently clear.

These assumptions may be violated in knowledge work, research, software development, or environments with frequent reprioritisation. A critical-chain schedule should therefore be treated as a control model whose assumptions must be tested, not as a promise that uncertainty has been eliminated.

10. Applied Case: Laboratory Modernisation Project

Background

The following is an illustrative case. A university laboratory plans to install a new analytical instrument. The project requires one specialist commissioning engineer, one laboratory technologist, and a shared validation facility. The sponsor has committed to operational readiness by the beginning of the next academic term.

Problem

The initial CPM schedule placed equipment installation, software configuration, and method preparation in parallel. However, installation and configuration both required the same commissioning engineer, while method validation required the shared facility. The logic-only schedule predicted 18 working days, but the resource profile showed two engineer units demanded during several periods and an unavailable validation facility during the planned validation window.

Analysis

The team identified the following chains:

• procurement and delivery → installation → commissioning;
• software configuration → commissioning;
• method preparation → validation → approval.

Commissioning could not begin until installation and software configuration were complete. Validation could not begin until method preparation and commissioning were complete. The commissioning engineer therefore created a resource dependency not visible in the original network.

The team ranked activities using three criteria: protection of the academic-term milestone, regulatory consequence, and effect on downstream work. It decided to complete installation before software configuration because installation released the facility and enabled early detection of physical defects. A feeding buffer was placed before validation, and a project buffer was placed before the operational-readiness date. A resource buffer consisted of a readiness confirmation two working days before commissioning.

Approach and solution

The schedule was rebuilt with resource dependencies and a single-engineer calendar. Non-critical documentation work was moved away from the commissioning window. The validation facility was reserved in advance, and the laboratory technologist was assigned continuously during method preparation. Daily meetings focused on decisions that could consume the feeding buffer.

Outcome

The resource-feasible plan required 20 working days rather than the original 18, but it eliminated the engineer over-allocation and protected the validation window. During execution, a two-day delay in software configuration consumed much of the feeding buffer but did not affect the committed readiness date. The project buffer remained partly intact.

Learning points

1. A shorter logic-only schedule can be operationally impossible.
2. A resource dependency may change the controlling chain.
3. Buffer placement makes uncertainty visible and actionable.
4. Resource readiness must be actively managed; a nominal assignment is insufficient.
5. A two-day increase in planned duration can reduce greater delivery risk.

11. Technology, Data, and Artificial Intelligence

Contemporary project-management platforms can produce resource histograms, identify over-allocation, calculate alternative schedules, and display buffer penetration. Integration with enterprise resource planning, timesheets, procurement systems, and workforce calendars improves data currency.

Artificial intelligence and optimisation tools can generate candidate schedules, detect recurring bottlenecks, forecast resource demand, and learn from historical productivity. Their outputs require professional review. Risks include biased historical data, false precision, inappropriate assumptions about worker productivity, privacy concerns, and opaque prioritisation. A responsible implementation should preserve human accountability, document input data, expose constraints and objective functions, test for discriminatory effects, and permit audit of schedule decisions.

Digital tools should not obscure basic questions: Is the resource truly available? Does the person possess the required competence? Are safety and fatigue limits respected? Is the activity estimate credible? Technology accelerates analysis, but it does not replace governance.

12. Governance, Ethics, and Professional Practice

Resource-constrained scheduling raises ethical and governance issues. Managers may be tempted to overload staff, hide capacity violations, or compress estimates to satisfy an announced date. Such practices can create fatigue, quality defects, unsafe work, and misleading stakeholder communication.

Professional practice requires that:

• capacity assumptions be explicit and reviewable;
• safety-critical and legally required work receive appropriate priority;
• staff workload and wellbeing be considered;
• schedule compression not bypass quality assurance;
• changes to resource availability be documented;
• stakeholders understand the difference between a feasible plan and a target;
• buffers not be manipulated to conceal poor performance;
• confidential workforce data be protected.

Where resources are shared across projects, portfolio governance is essential. Local optimisation of one project can damage the portfolio if it consumes a scarce specialist needed for a higher-value or regulatory initiative.

13. Limitations and Critical Evaluation

Resource levelling and CCPM are powerful but not universal solutions.

First, resource demand may be uncertain or non-linear. Two people do not always complete work in half the time; communication overhead may increase duration. Second, resources may be substitutable only in theory: a junior analyst cannot necessarily replace a certified engineer. Third, activities may be interruptible, partially transferable, or dependent on learning effects, making simple fixed-duration models inadequate.

Fourth, CCPM’s behavioural assumptions are difficult to sustain where individuals are evaluated on local task completion or assigned to many projects. Fifth, buffer penetration alone cannot explain root causes; a project may consume little time buffer while accumulating severe quality or scope risk. Finally, heuristic schedules may be good but not optimal, and the criteria used may favour schedule duration over value, resilience, or fairness.

For advanced practice, managers should combine resource-constrained scheduling with risk analysis, scenario planning, portfolio prioritisation, cost analysis, and stakeholder negotiation. Sensitivity analysis is especially valuable: test what happens if the scarce resource is unavailable, an activity takes longer, or a milestone becomes fixed.

Best-Practice Checklist

Before approving a resource-constrained schedule, ask:

1. Are all resource types, calendars, skills, and availability represented?
2. Does every activity satisfy both precedence and capacity constraints?
3. Have resource-induced dependencies been identified?
4. Is the controlling chain different from the original CPM critical path?
5. Are priority rules agreed, documented, and aligned with project value?
6. Are buffers based on explicit uncertainty assumptions?
7. Is the project buffer protected from routine management interference?
8. Are feeding and resource buffers monitored through defined actions?
9. Have safety, quality, ethical, and contractual constraints been included?
10. Has the schedule been tested under plausible scenarios?

Chapter Summary

Resource-constrained scheduling recognises that a project must be both logically valid and feasible within available capacity. Resource loading displays demand over time; allocation assigns resources; smoothing reduces variation within available float; and levelling resolves over-allocation, possibly by extending the project. Logic-critical activities derive their importance from precedence relationships, whereas resource-critical activities become controlling because scarce resources impose additional dependencies.

Because realistic scheduling problems can be computationally difficult, practitioners use heuristics such as minimum float, urgency, successor count, risk, and value. These rules must be applied transparently and reviewed against strategic, safety, and stakeholder considerations.

Critical Chain Project Management defines the controlling sequence using precedence and resource dependencies and protects delivery through project buffers, feeding buffers, and resource-readiness mechanisms. Buffer sizing is context dependent; simple aggregation rules are heuristics, not universal laws. Effective implementation requires credible estimates, focused execution, disciplined prioritisation, and active buffer management. Digital and AI tools can improve analysis, but human judgement, governance, ethics, and professional accountability remain essential.

Key Terms

Resource loading: Mapping resource demand to activities and time periods.

Resource allocation: Assigning available resources to project activities.

Resource levelling: Adjusting the schedule to remove resource over-allocation, potentially extending duration.

Resource smoothing: Reducing resource fluctuations while generally preserving the completion date.

Resource profile: The pattern of resource demand over time.

Capacity constraint: A limit on the amount of a resource available in a period.

Resource dependency: A sequencing relationship created by competition for a scarce resource.

Logic-critical activity: An activity critical because of precedence and duration relationships.

Resource-critical activity: An activity critical because resource limitations control its timing.

Heuristic: A practical decision rule that seeks a good solution without guaranteeing optimality.

Critical Chain: The longest precedence- and resource-dependent sequence in CCPM.

Project buffer: Time placed at the end of the critical chain to protect the committed completion date.

Feeding buffer: Time protecting the critical chain from delays on a non-critical feeding path.

Resource buffer: A readiness alert or mobilisation mechanism for a scarce resource.

Buffer penetration: The proportion of a buffer consumed.

Student syndrome: Delaying work until available time is nearly exhausted.

Parkinson’s law: The tendency for work to expand to fill the time available.

Multitasking: Switching among activities, often reducing effective throughput.

Resource-constrained project schedule: A schedule satisfying precedence and resource-capacity constraints.

Schedule robustness: The ability of a schedule to withstand foreseeable variation.

Review Questions

Short Answer Questions

1. What is the difference between resource levelling and resource smoothing?
2. Why can a logic-only CPM schedule be infeasible?
3. Define a resource-critical activity.
4. What is the purpose of a feeding buffer?
5. Why is a resource buffer generally different from a time buffer?

Long Answer Questions

1. Explain how resource constraints create additional precedence relationships, using an example.
2. Compare at least four heuristic priority rules used in resource-constrained scheduling.
3. Explain the differences between CPM and Critical Chain Project Management.
4. Discuss the assumptions and limitations of buffer-based schedule control.
5. Evaluate the ethical and governance issues involved in scheduling scarce human resources.

Analytical / Application Questions

1. A project has one specialist and two activities that are both eligible to start. Design a decision framework for choosing their order, and explain how you would assess the resulting schedule.
2. A project buffer is 10 working days and 6 days have been consumed while only 35% of critical-chain work is complete. Interpret this situation and recommend management actions.
3. Construct a resource-feasible alternative for a project in which two parallel activities require the same unique engineer. Compare the alternatives using duration, risk, downstream impact, and stakeholder value.

Multiple Choice Questions with Answers

Question 1

Which condition must a resource-constrained schedule satisfy in addition to precedence feasibility?

A. Every activity must have zero float

B. Resource demand must not exceed available capacity

C. All activities must be performed in parallel

D. The original CPM critical path must remain unchanged

Correct Answer: B. Resource demand must not exceed available capacity

Question 2

Resource smoothing normally attempts to achieve which outcome?

A. Reduce resource fluctuations while preserving the completion date

B. Eliminate all project buffers

C. Extend the project regardless of available float

D. Replace precedence relationships with resource assignments

Correct Answer: A. Reduce resource fluctuations while preserving the completion date

Question 3

What makes an activity resource-critical?

A. It has the lowest cost

B. It is always on the original CPM path

C. Resource limitations control its timing or sequencing

D. It uses the greatest number of calendar days

Correct Answer: C. Resource limitations control its timing or sequencing

Question 4

A project has a resource demand of 3 analysts for 4 days. Assuming constant demand, what is the total effort?

A. 7 analyst-days

B. 12 analyst-days

C. 1.33 analyst-days

D. 16 analyst-days

Correct Answer: B. 12 analyst-days

Question 5

Which priority rule gives preference to activities with the least total float?

A. Shortest processing time

B. Greatest successor count

C. Minimum total float

D. Highest resource substitution

Correct Answer: C. Minimum total float

Question 6

Where is a project buffer placed in Critical Chain Project Management?

A. Before every activity

B. At the end of the critical chain before the committed completion date

C. Only at the start of every feeding path

D. Inside each resource calendar

Correct Answer: B. At the end of the critical chain before the committed completion date

Question 7

What is the principal purpose of a feeding buffer?

A. Protect the critical chain from delays on a non-critical path

B. Increase the demand for scarce resources

C. Replace the project completion milestone

D. Guarantee that every activity finishes early

Correct Answer: A. Protect the critical chain from delays on a non-critical path

Question 8

Which statement best describes a resource buffer?

A. It is always a fixed block of time added to project duration

B. It is a readiness or mobilisation mechanism for a scarce resource

C. It is the same as total float

D. It removes the need for resource calendars

Correct Answer: B. It is a readiness or mobilisation mechanism for a scarce resource

Question 9

Using BP = Cᵦ ÷ B, what is buffer penetration when 6 days of a 10-day buffer have been consumed?

A. 6%

B. 40%

C. 60%

D. 166.7%

Correct Answer: C. 60%

Question 10

Why should AI-generated resource schedules be professionally reviewed?

A. AI cannot display a resource histogram

B. AI outputs may reflect biased data or inappropriate assumptions

C. AI always produces schedules longer than CPM schedules

D. AI removes all precedence relationships

Correct Answer: B. AI outputs may reflect biased data or inappropriate assumptions