Aggregate Planning in Services and Diverse Operating Environments

Introduction

Aggregate production planning (APP) determines how an organisation will balance forecast demand, capacity, workforce, inventory, subcontracting and timing over a medium-term horizon, commonly 3–18 months. Traditional APP is often presented through the image of a factory producing a relatively standard product. That image is useful but incomplete. A hospital cannot inventory an unused operating-room hour; a construction firm cannot easily transfer a crane from one project to another; a hotel loses revenue from an unsold room tonight; and a high-variety manufacturer must plan families of products whose processing requirements differ substantially.

The central managerial problem is therefore not simply “how many units should be produced?” It is: how should a system configure and deploy scarce, time-dependent capacity to meet uncertain and heterogeneous demand while protecting cost, quality, responsiveness and strategic capability?

Aggregate planning remains valuable in all these settings, but its variables, constraints, performance measures and decision horizons must be adapted to the operating environment.

Learning objectives

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

1. Distinguish aggregate planning issues in services, process manufacturing, discrete manufacturing, seasonal operations, projects and perishable/high-variety environments.
2. Select suitable demand and capacity levers for each environment.
3. Formulate and interpret aggregate planning relationships, including capacity, workforce and flow-balance equations.
4. Evaluate chase, level, hybrid, yield-management and project-capacity strategies.
5. Apply a structured planning process to a complex operating case.
6. Critically assess technology, uncertainty, ethics and implementation risks.

1. The aggregate-planning logic

An aggregate plan uses product or service families, rather than individual stock-keeping units or individual customer transactions. A family may be defined by a common labour group, a shared process, similar material requirements or a common market segment.

A basic period-by-period flow balance is:

where:

• I_t = ending inventory in period t (units),
• P_t = internally produced or completed units,
• R_t = receipts from subcontracting or external supply,
• D_t = demand fulfilled in period t, and
• B_t = demand backlogged or deferred at the end of period t.

The relationship assumes that all quantities are expressed in a common equivalent unit and that inventory and backlog are measured consistently. In a pure service system, physical inventory may be zero, while I_t can be interpreted as available capacity, appointments, cases in process or another appropriate flow measure.

A simplified cost objective is:

Here, c^p_t is regular production cost, c^h_t holding cost, c^b_t backlog or lateness cost, c^o_t overtime cost, c^u_t subcontracting or purchased-capacity cost, c^s_t hiring or activation cost, and c^f_t layoff, closure or deactivation cost. Variables O_t, U_t, S_t and F_t represent the corresponding decisions. The objective is a planning approximation: quality, safety, contractual service levels and strategic constraints must also be included.

A plan is feasible only if capacity, workforce, material, cash, labour-law and service constraints are respected. For example, if each worker supplies a_t productive hours and the available workforce is W_t, regular capacity is:

where u_t is standard labour time per unit. Units are consistent when W_t a_t is measured in labour-hours and u_t in labour-hours per equivalent unit. The equation assumes stable productivity and ignores setup, absenteeism and mix effects; these should be modelled separately when material.

2. Classifying operating environments

The first planning decision is to identify the system’s dominant characteristics. The following classification is more useful than a simple service/manufacturing distinction.

This diagnosis determines what “inventory,” “capacity” and “output” mean. Treating all systems as factories can produce mathematically feasible but operationally impossible plans.

3. Services: planning capacity that cannot be stored

3.1 The service-capacity problem

Services are often produced and consumed simultaneously. An empty airline seat, hotel room or consultation slot usually cannot be carried forward for sale tomorrow. Consequently, the classic level strategy—produce steadily and store output—has limited applicability. The principal buffers are:

• time: waiting, appointments and reservations;
• labour: overtime, part-time work, cross-training and temporary staff;
• space and equipment: flexible rooms, vehicles or servers;
• demand: prices, promotions, access rules and customer self-service; and
• queues: deliberate short-term accumulation of customers or work.

For a service centre with m identical agents, H available hours per agent, utilisation ρ and average service time s, approximate service capacity is:

Capacity is measured in customers when H and s use the same time unit. The utilisation factor should not be interpreted as permission to operate indefinitely at 100%; variability causes queues and service deterioration as utilisation approaches one.

3.2 Service demand and workforce levers

Managers may smooth demand by differential pricing, reservations, promotions, off-peak incentives, appointment rules, customer education and digital self-service. They may adjust capacity through shift redesign, split shifts, part-time employees, cross-training, temporary labour, overtime, outsourcing, tele-service and automation.

A useful service plan separates three quantities:

1. Demand arrivals: customers requesting service.
2. Demand served: customers actually processed.
3. Work in process or waiting: customers, calls, cases or documents awaiting completion.

A bank may not store a completed transaction, but it can carry a queue of loan applications. A hospital may not store treatment, but it can manage an appointment backlog—subject to clinical and ethical limits.

Service plans should therefore include response-time targets, abandonment, first-time quality and customer experience, not only labour cost. A call centre that minimises staffing cost by creating unacceptable abandonment has not achieved an effective aggregate plan.

3.3 Revenue management as aggregate planning

When demand is time-sensitive and customer segments have different willingness to pay, revenue management complements APP. The organisation allocates finite capacity among fare classes or booking periods. A simplified expected contribution from an accepted demand class is:

where p is price per unit, q accepted quantity, c variable fulfilment cost and v expected cost of displacing a more valuable future customer. The displacement term is essential: accepting a low-price booking is not automatically beneficial when capacity is scarce.

4. Process manufacturing

Process industries—chemicals, food, cement, paper and pharmaceuticals—often use continuous or batch processes. Their aggregate decisions concern production rates, campaign sizes, yields, inventories, maintenance windows, quality release and tank or warehouse capacity.

Unlike discrete production, a process may be difficult or impossible to stop and restart economically. A level production rate can be attractive because it avoids thermal, chemical or cleaning transients. However, inventory may be constrained by shelf life, hazardous-material rules or quality-release time.

Effective planning must model:

• material balances and yield losses;
• bottleneck rates and line availability;
• sequence-dependent cleaning or changeover time;
• minimum run lengths and campaign constraints;
• maintenance and regulatory shutdowns; and
• quarantine or release inventory.

If input quantity is X_t and yield is y_t, expected good output is:

The equation assumes yield is expressed as a fraction and that rework and scrap are not double-counted. When yield is uncertain, planning with only the average can understate shortage risk; scenario or safety-capacity analysis is preferable.

5. Discrete manufacturing and high-variety products

Discrete manufacturers produce distinguishable units through routings involving machines, labour, materials and setups. Aggregate planning is usually performed at product-family level, but family plans must be translated into a master production schedule and material requirements plan.

The most important adaptation is to use capacity hours rather than units alone. If family j requires a_jt bottleneck hours per unit and planned output is x_jt, then:

where C_t is available bottleneck capacity in hours. A plan based only on unit totals can appear feasible while overloading a critical machine.

High variety increases setup frequency, forecasting error and coordination cost. Appropriate levers include postponement, modular design, common components, product-family planning, setup reduction, flexible cells, finite-capacity scheduling and selective subcontracting. Postponement delays final differentiation until demand is clearer; it is effective when products share common upstream content and final customisation is rapid.

A product mix should be evaluated by its contribution per constrained resource, not merely contribution per unit. If contribution is g_j and bottleneck time is a_j, the ranking index is:

This is a short-run prioritisation aid, not a complete strategic rule. It can mislead when demand commitments, quality, setups, materials or multiple bottlenecks dominate.

6. Seasonal operations

Seasonality creates a predictable mismatch between average capacity and peak demand. Examples include apparel, tourism, agriculture, education, tax services and festive food production. A seasonal plan may combine:

• prebuilding inventory before the peak;
• temporary or seasonal employees;
• overtime and additional shifts;
• subcontracting;
• reservations and advance sales;
• peak-period pricing; and
• product or service postponement.

Prebuilding is sensible only when the expected cost of carrying and potentially obsoleting output is lower than the cost of peak capacity. For a unit produced one period early, a simple comparison is:

where c^p is early production cost, c^h holding cost, P(obsolete) probability of obsolescence and c^o_b the loss associated with an obsolete unit. This is a simplified decision rule; shortage penalties, salvage value and service commitments should also be included.

Seasonal plans should be tested against high-demand scenarios. A plan that is least costly at the mean forecast may be fragile during a peak. Capacity reservation, contingent labour and supplier agreements can provide real options at a premium.

7. Project-based and engineer-to-order environments

Projects produce unique outputs and consume resources over a network of activities. Aggregate planning is less about inventory and more about the portfolio of bids, project starts, resource loading and completion commitments.

Important decisions include:

• which projects to accept and when to start them;
• how much scarce engineering, equipment and specialist labour to reserve;
• whether to subcontract work packages;
• how to sequence projects with shared resources; and
• how to protect contractual milestones.

The critical path is the longest-duration path through a precedence network. If activity durations are fixed, project completion time equals the duration of this path. However, aggregate planning must also consider resource-constrained critical paths: a noncritical activity can become schedule-critical when its specialist resource is unavailable.

A project portfolio may be evaluated using expected contribution per scarce resource-hour, adjusted for risk, strategic learning and contractual penalties. Starting every available project can overload the organisation and increase work-in-process, engineering changes and late delivery. Capacity reservation and explicit work-in-process limits are often more valuable than aggressive utilisation targets.

8. Perishable capacity and products

Perishability has two forms. Perishable capacity expires when time passes, as with a hotel room or aircraft seat. Perishable product loses value through spoilage, fashion, technological change or regulatory expiry.

Planning must distinguish lost sales from backlog. A missed hotel booking tonight cannot normally be fulfilled tomorrow, whereas an uncompleted machine order may be delayed. For a single-period product, an order-up-to decision often balances underage and overage costs. If C_u is the cost of ordering too little and C_o the cost of ordering too much, the critical fractile is:

where D is random demand and Q* the target quantity. This result assumes a single selling period, known demand distribution, linear underage/overage costs and no replenishment during the period. Salvage value should be included in C_o; it reduces the effective overage cost.

For capacity allocation, controls such as booking limits, overbooking, dynamic prices and allocation by customer class may be used. Overbooking requires a careful estimate of no-shows and a mitigation plan; otherwise it can create serious fairness, compensation and reputation problems.

9. A unified planning process

A robust cross-environment process contains the following stages:

Stage 1: Define the planning object

Select product families, customer segments, projects or service categories. Define the unit of measure: physical units, labour-hours, cases, room-nights, machine-hours or revenue-equivalent units.

Stage 2: Develop demand scenarios

Use a baseline forecast plus high and low cases. Record assumptions about promotions, cancellations, seasonality, economic conditions and customer behaviour. Forecast accuracy should be measured separately by family and horizon; aggregation can reduce random error but may hide mix errors.

Stage 3: Map capacity and constraints

Identify bottlenecks, labour skills, maintenance, supplier limits, regulatory restrictions, minimum run lengths, setup times, storage, service-level requirements and project precedence.

Stage 4: Generate alternative strategies

Compare level, chase and hybrid policies. Add system-specific options such as pricing, postponement, reservations, yield management, prebuild, subcontracting, cross-training and capacity reservation.

Stage 5: Optimise and stress-test

Linear programming is suitable when relationships are linear and data are sufficiently stable. Mixed-integer models are needed for binary decisions such as opening a shift, activating a supplier or starting a campaign. Simulation is useful when queues, variability, routing and human behaviour are central. Robust or stochastic optimisation is appropriate when uncertainty is significant and failure is costly.

Stage 6: Translate and govern

Convert the aggregate plan into a master schedule, staffing plan, procurement plan, project portfolio and financial plan. Establish ownership, exception thresholds, revision frequency and escalation rules. Sales and operations planning should be a cross-functional process rather than a production-only meeting.

10. Worked application: Northstar Outdoor Equipment

The following is an illustrative case. Northstar makes two product families: standard camping stoves and customised expedition stoves. The assembly bottleneck has 1,000 hours available per month. A standard unit requires 1 hour and contributes ₹600; a customised unit requires 2 hours and contributes ₹1,000. Forecast demand for the month is 700 standard units and 300 customised units.

The unconstrained requirement is:

Thus, demand exceeds bottleneck capacity by 300 hours. Producing all demand is infeasible. Contribution per bottleneck hour is:

• standard: ₹600 ÷ 1 = ₹600 per hour;
• customised: ₹1,000 ÷ 2 = ₹500 per hour.

If demand is equally firm and no strategic or contractual condition changes the decision, the short-run contribution rule ranks standard units first. Producing 700 standard units uses 700 hours, leaving 300 hours for 150 customised units. Contribution is:

A naïve manager might prefer customised products because their contribution per unit is higher. The bottleneck analysis shows why that conclusion is wrong for this month. Nevertheless, the final plan should examine customer penalties, future demand, learning effects, supplier minimums and the possibility of overtime or subcontracting. If 300 hours of overtime cost less than the contribution lost from rejecting customised demand, overtime may be justified. If custom orders create strategic relationships, a weighted multi-period model may deliberately accept fewer standard units.

Learning points: aggregate planning must use the constrained resource; product-family plans require mix analysis; and a local contribution rule is informative but not sufficient for a strategic decision.

11. Technology, data and contemporary practice

Advanced planning and scheduling systems integrate forecasts, orders, inventory, labour, capacity and supplier data. Digital twins and discrete-event simulation can test alternative policies before implementation. Machine-learning forecasts may improve nonlinear or granular demand prediction, but forecast accuracy does not remove capacity constraints or guarantee unbiased decisions.

AI-assisted planning should therefore include:

• traceable data lineage and versioned assumptions;
• human review of recommendations;
• monitoring for forecast and allocation bias;
• protection of employee and customer data;
• explainable exception rules; and
• controls against automated decisions that violate contracts, safety or equality obligations.

A mathematically optimal schedule based on inaccurate master data is operationally inferior to a simpler, trusted plan. Data governance—standard units, valid calendars, reliable routings and disciplined change control—is a prerequisite for advanced analytics.

12. Challenges, ethics and limitations

Aggregate plans are abstractions. They may conceal individual customer priorities, skill fatigue, setup realities and local bottlenecks. Forecasts are uncertain, cost coefficients are estimates, and organisational incentives can distort information. Common failure modes include:

• treating average demand as certain;
• ignoring mix and bottleneck hours;
• assuming labour is perfectly flexible;
• using utilisation as the sole performance target;
• shifting instability to suppliers or employees;
• counting backlog as harmless in a safety-critical service; and
• optimising cost while degrading quality or resilience.

Ethically, workforce flexibility must respect employment law, fair scheduling, health and safety, consultation and non-discrimination. In hospitals, allocation cannot be based only on revenue. In food and pharmaceutical operations, expiry and quality decisions must comply with regulation. In customer allocation, opaque dynamic pricing or overbooking may create unacceptable inequity. Resilience may justify a higher-cost plan with backup suppliers, spare capacity or strategic inventory.

The best plan is consequently a feasible, transparent and resilient compromise, not necessarily the numerical minimum of a narrow cost function.

Chapter Summary

Aggregate planning balances medium-term demand with capacity across product families, services and projects. Its core flow balance remains useful, but the meaning of inventory, output and capacity changes by environment. Services rely on staffing, appointments, queues, pricing and revenue management because capacity cannot generally be stored. Process manufacturing requires attention to yield, campaigns, maintenance and continuous-flow constraints. Discrete and high-variety manufacturing require bottleneck-hour, setup and mix analysis. Seasonal systems combine prebuild, temporary capacity, pricing and scenario planning. Project environments focus on portfolio selection, resource loading, precedence and milestone risk. Perishable systems require allocation, salvage, booking and lost-sales analysis.

Effective practice defines the planning unit, develops scenarios, maps constraints, compares alternative levers, stress-tests the plan and translates it into accountable operating decisions. Optimisation, simulation and AI can support this work, but reliable data, human governance, ethics, quality and resilience remain essential.

Key Terms

Aggregate planning: Medium-term coordination of demand, capacity and resources at family or portfolio level.

Backlog: Demand accepted but not yet fulfilled.

Bottleneck: Resource whose limited capacity constrains system output.

Capacity reservation: Advance commitment of capacity for uncertain future demand.

Chase strategy: Policy that adjusts capacity or workforce to follow demand.

Contribution per bottleneck hour: Contribution earned for each unit of constrained-resource time.

Demand smoothing: Actions that shift demand across time.

Finite-capacity scheduling: Scheduling that explicitly respects available resource capacity.

Level strategy: Policy that maintains relatively stable production or staffing.

Postponement: Delaying final product differentiation until demand information improves.

Product family: Group of products sharing relevant resources or planning characteristics.

Revenue management: Allocation and pricing of constrained capacity among demand segments.

Seasonality: Systematic demand variation associated with time of year or recurring periods.

Service capacity: Available ability to process customers, cases or transactions.

Setup time: Time required to prepare a resource for a different product or service type.

Single-period model: Decision model for an item or capacity opportunity with one selling period.

Yield: Proportion of input converted into acceptable output.

Work in process: Jobs or customers started but not yet completed.

Review Questions

Short Answer Questions

1. Why is aggregate planning usually performed by product family rather than by individual stock-keeping unit?
2. Identify four capacity levers available to a labour-intensive service organisation.
3. Why can a discrete manufacturer not rely only on total unit capacity?
4. What is meant by postponement, and when is it useful?
5. Distinguish perishable capacity from perishable product.

Long Answer Questions

1. Explain how aggregate planning should be adapted for a hospital or call centre. Include capacity, demand and service-quality considerations.
2. Compare level, chase and hybrid strategies in a seasonal manufacturing environment.
3. Discuss the distinctive planning requirements of process manufacturing and discrete manufacturing.
4. Explain how project portfolio planning differs from inventory-based aggregate planning.
5. Critically evaluate the role of AI and optimisation in aggregate planning, including governance risks.

Analytical / Application Questions

1. A service centre has 8 agents, each available 150 productive hours per month. Average service time is 0.5 hour and sustainable utilisation is 85%. Estimate monthly capacity and explain two limitations of the estimate.
2. A factory has 900 bottleneck hours. Product A contributes ₹500 per unit and uses 1 hour; Product B contributes ₹900 per unit and uses 2 hours. Demand is 600 units of each. Construct a short-run mix plan using contribution per bottleneck hour, then state two conditions that could change the recommendation.
3. A hotel faces a predictable festival peak but has limited rooms and uncertain cancellations. Design an aggregate plan using reservations, pricing, overbooking controls and a contingency capacity policy. Identify the principal ethical risk.

Multiple Choice Questions with Answers

Question 1

Which characteristic most directly distinguishes service aggregate planning from traditional inventory-based manufacturing planning?

A. Services always have lower demand variability

B. Service capacity often cannot be stored for future periods

C. Services never use forecasts

D. Service organisations do not face bottlenecks

Correct Answer: B. Service capacity often cannot be stored for future periods

Question 2

A process has 1,200 input units and a yield of 0.80. What is the expected good output?

A. 240 units

B. 960 units

C. 1,200 units

D. 1,500 units

Correct Answer: B. 960 units

Question 3

A product requires 3 bottleneck hours and contributes ₹1,200. Its contribution per bottleneck hour is:

A. ₹300

B. ₹400

C. ₹900

D. ₹3,600

Correct Answer: B. ₹400

Question 4

Which is a typical demand lever in a service operation?

A. Finished-goods warehousing

B. Differential pricing by time period

C. Increasing material yield

D. Reducing component count only

Correct Answer: B. Differential pricing by time period

Question 5

What is the principal purpose of postponement?

A. To complete all customisation before demand is known

B. To delay differentiation until better demand information is available

C. To eliminate all inventory

D. To increase setup time deliberately

Correct Answer: B. To delay differentiation until better demand information is available

Question 6

In project-based aggregate planning, a resource-constrained critical path arises mainly because:

A. All projects have identical activities

B. Resource limitations can delay activities that were not initially time-critical

C. Projects never have precedence relationships

D. Inventory is always the primary constraint

Correct Answer: B. Resource limitations can delay activities that were not initially time-critical

Question 7

If a service centre has 10 agents, each with 100 available hours, utilisation is 0.8, and average service time is 0.5 hour, approximate capacity is:

A. 160 customers

B. 800 customers

C. 1,600 customers

D. 2,000 customers

Correct Answer: C. 1,600 customers

Question 8

The critical fractile in a single-period inventory model is based on:

A. Holding cost divided by production cost

B. Underage cost divided by the sum of underage and overage costs

C. Demand divided by capacity

D. Price divided by lead time

Correct Answer: B. Underage cost divided by the sum of underage and overage costs

Question 9

Which planning approach is especially useful when queues, variability and routing behaviour are central?

A. Discrete-event simulation

B. Simple unit averaging only

C. Ignoring stochastic effects

D. Financial statement analysis alone

Correct Answer: A. Discrete-event simulation

Question 10

Why can a plan with the lowest forecast-based cost be strategically weak?

A. Cost is never relevant to planning

B. It may be fragile under demand uncertainty and ignore resilience or service commitments

C. Low-cost plans always increase inventory

D. Forecasts cannot be used in aggregate planning

Correct Answer: B. It may be fragile under demand uncertainty and ignore resilience or service commitments