Demand Forecasting and Materials Planning

Introduction and Context

Materials management connects market requirements with the resources needed to produce and distribute goods and services. Its central planning problem is deceptively simple: determine what is required, how much is required, when it is required, and how the organisation can provide it at acceptable cost and service levels. In practice, the problem is difficult because demand is uncertain, products have multilevel structures, suppliers have variable lead times, capacity is finite, and decisions made at one level affect the entire supply network.

Demand forecasting provides an informed estimate of future requirements. Materials planning converts that estimate, together with confirmed customer orders, inventory records, product structures, lead times and capacity information, into time-phased procurement and production decisions. Forecasting is therefore not a prediction of the future in a deterministic sense; it is a structured basis for making decisions under uncertainty.

The planning system normally operates at several levels. Sales and operations planning (S&OP) balances market demand and aggregate supply. The master production schedule (MPS) converts the aggregate plan into specific end-item quantities and dates. Material requirements planning (MRP) explodes the MPS through bills of materials (BOMs) to calculate component and raw-material requirements. Distribution requirements planning (DRP) performs a similar time-phased calculation for distribution locations. Capacity planning tests whether the resulting plan is feasible.

Learning Objectives

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

1. Distinguish qualitative and quantitative forecasting methods and select an appropriate method.
2. Calculate and interpret moving averages, exponential smoothing, forecast error and tracking measures.
3. Explain the structure and governance of a bill of materials.
4. Develop a basic master production schedule and MRP record.
5. Interpret gross requirements, net requirements, planned receipts and planned-order releases.
6. Assess capacity implications and understand the role of DRP and S&OP.
7. Evaluate planning assumptions, data quality, technology, risks and managerial trade-offs.

1. Demand Forecasting as a Planning Activity

A forecast is an estimate of future demand based on available information, assumptions and a specified time horizon. Forecasts may concern independent demand, such as demand for finished products, or dependent demand, such as demand for components derived from the production plan.

Important characteristics of demand include:

• Level: the average magnitude of demand.
• Trend: a persistent upward or downward movement.
• Seasonality: a repeating pattern linked to a calendar or other cycle.
• Cyclical variation: longer-term fluctuations associated with economic or industry conditions.
• Random variation: irregular movement that cannot be explained reliably.
• Intermittency: demand occurring sporadically, often for service parts.

The forecast horizon should match the decision. Strategic capacity decisions may require forecasts several years ahead; aggregate production planning may use monthly forecasts for 12–18 months; MPS decisions often use weekly forecasts over a shorter frozen and flexible horizon. Forecast accuracy generally decreases as the horizon increases.

A useful distinction is between forecast accuracy and forecast usefulness. A forecast can be statistically accurate but operationally unhelpful if it is not available at the right product, location or time level. Conversely, a forecast with moderate error may be valuable if it supports timely procurement and exposes uncertainty clearly.

2. Qualitative Forecasting

Qualitative methods are appropriate when historical data are limited, the product is new, structural change is expected, or expert knowledge contains information not represented in the data. They are also useful as a complement to statistical forecasts.

2.1 Major qualitative methods

Executive opinion combines the judgements of senior managers. It is fast but may be affected by hierarchy, optimism or anchoring.

Sales-force composite aggregates estimates from sales personnel. Sales staff possess customer knowledge but may deliberately understate or overstate demand because of incentives.

Customer survey and market research collect information directly from current or prospective customers. Stated intention, however, may differ from actual purchasing behaviour.

Delphi method obtains independent estimates from a panel of experts through several anonymous rounds. A facilitator summarises the responses between rounds until convergence or a documented range is achieved. Anonymity reduces dominance effects, but the method can be time-consuming.

Analogy method forecasts a new product by examining the demand pattern of a comparable product or market. The quality of the analogy must be tested rather than assumed.

Qualitative forecasts should document the evidence, assumptions, participants, date and degree of confidence. A forecast presented as a single precise number can create false certainty; a range or scenario set is often more responsible.

3. Quantitative Forecasting

Quantitative methods use historical observations and mathematical relationships. They are not automatically superior to judgement: their validity depends on data quality, stability of relationships and appropriate method selection.

3.1 Naïve forecasting

For a stable series, the next-period forecast may equal the latest actual demand:

where F_t+1 is the forecast for period t+1 and A_t is actual demand in period t. The method is simple and provides a benchmark. A more seasonal naïve method uses demand from the corresponding prior season.

3.2 Moving average

A simple moving average of the previous n observations is:

All observations receive equal weight. A larger n smooths random variation but responds more slowly to change. The observations must be measured in consistent units and periods.

A weighted moving average gives different weights:

The weights w_i are non-negative unless a justified alternative specification is used. More recent observations commonly receive greater weight.

Example. Demand during the last three months was 420, 450 and 480 units. The three-month moving-average forecast is:

F_t+1 = (420 + 450 + 480) ÷ 3 = 450 units.

3.3 Exponential smoothing

Single exponential smoothing is expressed as:

where 0 ≤ α ≤ 1 is the smoothing constant, A_t is actual demand and F_t is the current forecast. A high α responds quickly to recent changes; a low value produces greater smoothing. The method is most suitable when there is no strong trend or seasonality.

Example. If α = 0.30, actual demand is 500 units and the previous forecast is 460 units:

F_t+1 = 0.30(500) + 0.70(460) = 472 units.

For trend, a method such as Holt’s double exponential smoothing is appropriate. For trend and seasonality, Holt–Winters methods may be used. The precise additive or multiplicative form should reflect whether seasonal variation is approximately constant or proportional to the level of demand.

3.4 Regression and causal forecasting

Regression relates demand to explanatory variables such as price, advertising, income, weather or installed equipment. A simple linear model is:

where Y_t is demand, X_t is an explanatory variable, a is the intercept, b is the estimated change in demand per unit change in X_t, and ε_t is the unexplained error. The relationship should not be interpreted as causal merely because two variables are correlated. Managers should test plausibility, timing, omitted variables and structural breaks.

4. Forecast Evaluation and Governance

Forecast evaluation should use measures appropriate to the decision and data. Let the forecast error be:

Mean absolute deviation (MAD) is:

Mean squared error (MSE) is:

Root mean squared error is RMSE = √MSE. Mean absolute percentage error (MAPE) is:

MAPE is undefined when actual demand is zero and can overemphasise errors on small denominators. MAD and RMSE retain the original unit or its square-root equivalent and are often easier to interpret operationally.

Forecast bias is indicated by cumulative error:

A positive CFE under the definition above indicates that actual demand has generally exceeded the forecast. A tracking signal may be calculated as:

Its use requires a defined control limit and a stable measurement policy; no universal threshold should be applied without considering the organisation and demand process.

Forecast governance includes version control, ownership, exception review, parameter approval, documented overrides and separation of forecast accuracy from individual performance appraisal where such linkage would encourage manipulation. Bias may also arise from promotions, stock-outs, product substitutions or data censoring. Observed sales are not always equal to unconstrained demand.

5. Bills of Materials

A bill of materials is a structured record of the components, quantities and relationships required to make one unit of a parent item. A BOM may include raw materials, purchased parts, subassemblies, packaging and phantom items. It should identify item numbers, descriptions, quantities per, units of measure, revision or effectivity dates, scrap factors and engineering or manufacturing status.

A single-level BOM lists immediate components. A multilevel BOM represents the complete product structure. A product tree is useful for explanation, but operational systems normally store parent–component relationships and low-level codes.

If a parent requires q units of a component per parent and planned production is P, gross component requirement is:

If a scrap rate s is expected and the requirement is grossed up to compensate for scrap, one common convention is:

where 0 ≤ s < 1. The convention must be stated because some systems apply scrap to planned usage rather than dividing by the yield factor.

BOM accuracy is a governance issue. Engineering changes, substitute parts, obsolete components, co-products and configurable options can cause serious planning errors. Effective dates must align with the MPS, and units of measure must be consistent.

6. Master Production Scheduling

The MPS specifies which end items will be completed, in what quantities and in which time periods. It translates aggregate plans into a feasible statement of production intent. The MPS is not simply a sales forecast: it considers confirmed orders, forecast demand, inventory policy, lot sizes, due dates, capacity and management priorities.

Typical MPS elements include:

• forecast and customer orders;
• projected available balance;
• safety stock;
• available-to-promise quantity;
• MPS receipts and production dates;
• a time fence separating frozen, slushy and liquid planning zones.

A basic projected available balance can be calculated as:

where PAB_t is projected available balance at the end of period t, MPS_t is the scheduled receipt, F_t is forecast demand and O_t is customer orders. The use of max(F_t,O_t) is a common rule when orders consume the forecast rather than being added to it. Organisations may use alternative consumption rules, which must be documented.

The MPS should be stable enough for purchasing and production control but responsive enough to reflect important changes. Excessive changes create nervousness, expedite costs and supplier disruption.

7. Material Requirements Planning

MRP is a time-phased planning technique for dependent demand. Its logic uses three primary inputs:

1. MPS: what finished items are required and when.
2. BOM: which components and quantities are required.
3. Inventory and status records: on-hand balances, scheduled receipts, lead times, lot sizes, safety stocks and open orders.

MRP calculates gross requirements, scheduled receipts, projected available balances, net requirements, planned-order receipts and planned-order releases. The process is often described as BOM explosion, netting and time offsetting.

For a simplified period-by-period record:

where SR_t is scheduled receipt, POR_t is planned-order receipt and GR_t is gross requirement. If projected availability would fall below safety stock, net requirement is:

where SS is safety stock. The planned receipt is then determined by the lot-sizing rule. With lot-for-lot sizing, planned receipt equals net requirement. With a fixed order quantity, the receipt is the smallest allowed multiple that satisfies the net requirement.

A planned-order release is offset backwards by the item’s lead time. If a receipt is needed in period t and lead time is L periods:

where PORC_t is planned-order receipt and PORL_t−L is planned-order release. This assumes lead time is known, measured in the same planning periods and unaffected by capacity or queue variability.

Worked MRP example

A finished product requires two units of component X. The MPS calls for 50 finished units in week 3. Component X has 20 units on hand, no scheduled receipts, a safety stock of 10 units, a one-week lead time and lot-for-lot ordering.

Gross requirement for X in week 3 is 50 × 2 = 100 units. Available inventory before the receipt is 20 units. To preserve safety stock, the net requirement is 100 + 10 − 20 = 90 units. Therefore, the planned receipt is 90 units in week 3, and the planned release is 90 units in week 2. After the receipt and issue, the projected balance is 10 units, equal to safety stock.

This example assumes no other demand, no scheduled receipt, perfect BOM accuracy, no scrap and a reliable one-week lead time. Real MRP records must include all relevant orders and planning policies.

Lot sizing and planning trade-offs

Lot-for-lot minimises planned inventory but may create frequent orders and setup costs. Fixed-order-quantity rules exploit economies of scale but create residual inventory. Period-order quantity groups requirements over a selected number of periods. More sophisticated methods may include economic order quantity, part-period balancing or dynamic programming, although the best method depends on setup cost, carrying cost, service requirements, supplier constraints and demand variability.

MRP is not an optimisation of the entire supply chain by itself. It is a planning logic whose outputs require review, capacity validation and execution discipline.

8. Capacity Considerations

A material plan is feasible only if required resources are available. Capacity includes machines, labour, tooling, storage, supplier capability, transport and critical skills. Capacity requirements planning translates planned orders into hours or resource loads.

For a work centre, required load may be estimated as:

If available productive capacity is C hours and required load is L hours, a simple utilisation measure is:

The measure must use comparable hours. If C represents gross clock time but L excludes setups, the comparison is invalid. Capacity should account for planned downtime, breaks, maintenance, yield losses and realistic efficiency.

When capacity is insufficient, managers may revise the MPS, adjust lot sizes, subcontract, add shifts, reschedule orders, use alternate resources or negotiate due dates. A mathematically balanced material plan that cannot be produced is not a valid plan.

9. Distribution Requirements Planning

DRP applies time-phased planning to distribution networks. It begins with demand at regional warehouses or customer-facing locations and calculates replenishment requirements at upstream distribution centres and the plant. The logic resembles MRP but includes transportation lead times, shipment calendars, order cycles, safety stock and network relationships.

A DRP record may include forecast, customer orders, projected available balance, planned shipment receipts and planned shipment releases. The plan must distinguish demand at a location from transfers arriving from another location. Failure to do so can double-count requirements.

DRP supports deployment decisions, but it does not eliminate uncertainty. Transport disruption, allocation rules, minimum shipment quantities and inventory imbalance can require managerial intervention. In multi-echelon systems, safety stock should be positioned based on total network cost and service objectives rather than independently maximised at every location.

10. Sales and Operations Planning

S&OP is a cross-functional process that reconciles demand, supply, inventory, capacity, financial objectives and strategic priorities. It normally operates at an aggregate product-family level and a monthly horizon. A typical cycle includes data gathering, statistical forecasting, demand review, supply review, pre-S&OP reconciliation and executive approval.

The process should produce one approved plan, explicit assumptions and documented decisions. Common aggregate strategies include:

• Level production: relatively stable production with inventory or backlog absorbing demand variation.
• Chase demand: production varies with demand, reducing inventory but increasing workforce and capacity volatility.
• Mixed strategy: combines inventory, overtime, subcontracting, workforce changes and demand management.

S&OP is not merely a meeting. It is a governance mechanism that links commercial commitments with operational feasibility. Its effectiveness depends on common data, cross-functional accountability, scenario analysis and executive decisions when trade-offs cannot be resolved operationally.

11. Integrated Planning Architecture

The planning hierarchy can be represented as follows:

Figure: Relationship among forecasting, S&OP, MPS, MRP, DRP and capacity validation.

An enterprise resource planning system can integrate these activities, but integration does not guarantee correctness. Poor master data, inappropriate parameters or undisciplined overrides can produce highly automated errors. The planner remains responsible for interpreting exceptions and testing assumptions.

12. Technology, AI and Contemporary Practice

Modern planning platforms combine ERP, advanced planning systems, warehouse data, point-of-sale information, supplier portals and transport data. Machine-learning models can identify nonlinear relationships, detect anomalies and generate forecasts at scale. They are most useful when demand signals are rich and data processes are controlled.

AI does not remove the need for causal reasoning. A model may learn stock-out patterns as if they were low demand, confuse promotions with permanent growth, or reproduce historical bias. Governance should include data lineage, explainability appropriate to the decision, human review, drift monitoring, access controls and documented override reasons. Scenario-based planning remains essential when historical relationships are unlikely to hold.

Contemporary practice also emphasises demand sensing, postponement, collaborative forecasting, supplier collaboration, digital twins and probabilistic planning. These approaches can improve responsiveness, but their benefits depend on data timeliness, process maturity and organisational willingness to share information.

13. Illustrative Case: Apex Home Appliances

Background. Apex Home Appliances assembles a countertop mixer. Demand has become seasonal, with high sales before a major holiday. The company uses a two-level BOM: one mixer requires one motor, one housing and two control knobs. Motors have a two-week supplier lead time. The plant can assemble 1,000 mixers per week but has only 700 hours of motor-test capacity, while each mixer requires 0.8 test hours.

Problem. The sales department proposes an MPS of 1,000 mixers per week for the next four weeks. The forecast is based on a moving average that includes two stock-out months, so it understates unconstrained demand. Purchasing argues that motors should be ordered immediately, while operations warns that testing is the bottleneck.

Analysis. The MPS creates a test load of 1,000 × 0.8 = 800 hours per week, exceeding the 700-hour capacity by 100 hours. The plan is therefore infeasible without overtime, outsourcing or a revised schedule. The forecast also requires adjustment because stock-outs suppressed recorded sales. A cross-functional demand review should combine lost-sales estimates, customer orders, promotion plans and market information.

Approach. S&OP approves a four-week aggregate plan of 3,600 mixers rather than 4,000, with 800 units in week 1, 900 in week 2, 950 in week 3 and 950 in week 4. The MPS is then checked against test capacity. Week 3 and week 4 still require 760 hours, so 60 hours of approved overtime are arranged in each week. MRP explodes the schedule into motors, housings and knobs. Because motors have a two-week lead time, releases are brought forward and supplier confirmation is obtained. DRP allocates finished mixers to two regional warehouses according to orders, safety stock and transport lead times.

Outcome. The revised plan reduces the capacity violation, aligns motor releases with actual requirements and makes the trade-off visible to executives. It does not guarantee the forecast, but it creates a controlled response plan with explicit assumptions and exception ownership.

Learning points. First, recorded sales may not represent true demand. Second, MPS quantities must be capacity checked. Third, MRP depends on accurate BOMs, lead times and inventory records. Fourth, S&OP is valuable because no single function can resolve the demand–capacity conflict independently.

14. Challenges, Limitations and Best Practices

Forecasting and materials planning are limited by uncertainty, data latency, master-data errors, unstable suppliers, changing product designs, long lead times and behavioural responses. The bullwhip effect can amplify small changes in end demand as each stage adds safety stock or reacts to incomplete information.

Best practices include:

1. Segment items by value, variability, criticality and supply risk rather than applying one policy to all items.
2. Use a simple benchmark before adopting a complex model.
3. Separate demand history from stock-out, promotion and substitution effects.
4. Maintain BOM, lead-time, lot-size and inventory accuracy through formal ownership.
5. Establish frozen and flexible planning time fences.
6. Use exception management rather than reviewing every item with equal intensity.
7. Validate MRP outputs against supplier, labour, machine and logistics capacity.
8. Measure forecast bias as well as average error.
9. Conduct scenario analysis for supply disruption, demand surges and engineering changes.
10. Record assumptions and decisions so that forecasts and plans are auditable.

Ethically, planners should not manipulate forecasts to conceal capacity problems or meet inappropriate performance targets. Safety stock and allocation rules should be transparent when shortages affect customers or business units. Data privacy and cybersecurity are relevant when customer-level demand data are integrated into planning systems. Governance should ensure that automated recommendations remain reviewable and contestable.

Chapter Summary

Demand forecasting provides structured estimates of future independent demand, while materials planning translates those estimates and confirmed orders into feasible time-phased actions. Qualitative methods are valuable when data are scarce or structural change is expected; quantitative methods include naïve forecasts, moving averages, exponential smoothing and regression. Forecast accuracy, bias and usefulness must be monitored with clearly defined measures.

The BOM expresses product structure and enables dependent-demand calculation. The MPS converts aggregate plans into end-item schedules. MRP uses the MPS, BOM and inventory records to calculate gross requirements, net requirements, planned receipts and releases. Capacity planning tests whether the material plan can be executed. DRP extends time-phased planning across distribution locations, while S&OP reconciles demand, supply, inventory, capacity and financial priorities at an aggregate level.

Effective planning is not achieved by software alone. It requires accurate master data, explicit assumptions, cross-functional governance, disciplined exception management and ethical use of information. AI and advanced analytics can improve responsiveness, but human judgement remains necessary when relationships change, data are biased or capacity constraints require strategic choices.

Key Terms

Forecast: An estimate of future demand or another future condition.

Independent demand: Demand determined by the market rather than by the requirement for another item.

Dependent demand: Demand calculated from the planned production of a parent item.

Seasonality: A repeating demand pattern associated with a calendar or cycle.

Moving average: A forecast based on the average of recent observations.

Exponential smoothing: A forecasting method that weights recent observations through a smoothing constant.

Forecast bias: A persistent tendency for forecasts to overestimate or underestimate actual demand.

Bill of materials: A structured list of components and quantities required for a parent item.

Low-level code: The lowest product-structure level at which an item appears for MRP processing.

Master production schedule: A time-phased schedule of end-item production.

Time fence: A boundary defining how freely a plan may be changed.

Material requirements planning: A system for calculating time-phased dependent material requirements.

Gross requirement: Total requirement before considering available inventory or receipts.

Net requirement: Requirement remaining after available supply and policy stocks are considered.

Planned-order release: The date an order should be initiated to meet a planned receipt.

Lot sizing: The rule used to determine the quantity ordered or produced.

Capacity requirements planning: The assessment of resource loads created by planned orders.

Distribution requirements planning: Time-phased replenishment planning for distribution locations.

Sales and operations planning: Cross-functional reconciliation of aggregate demand and supply.

Demand nervousness: Frequent and costly changes to planned orders caused by instability or inappropriate planning parameters.

Review Questions

5 Short Answer Questions

1. Distinguish independent demand from dependent demand.
2. State two advantages and two limitations of qualitative forecasting.
3. What information is required as the primary input to MRP?
4. Define a planned-order release.
5. Why should forecast bias be monitored separately from forecast accuracy?

5 Long Answer Questions

1. Explain the role of S&OP, MPS, MRP and DRP within an integrated planning hierarchy.
2. Compare moving averages, exponential smoothing and regression forecasting, including their assumptions and suitable applications.
3. Explain how BOM accuracy, lead times and lot-sizing rules influence MRP results.
4. Discuss the relationship between material requirements and capacity requirements, using an industrial example.
5. Evaluate the opportunities and governance risks associated with AI-enabled demand forecasting.

3 Analytical / Application Questions

1. Monthly demand is 700, 760 and 820 units. Calculate the three-month moving-average forecast. If the previous forecast was 780 units, actual demand was 840 units and α = 0.25, calculate the exponential-smoothing forecast and compare the two results.
2. A product requires three units of component A. Planned production is 240 units, on-hand A is 300 units, safety stock is 30 units and there are no scheduled receipts. Calculate the gross requirement and net requirement, assuming no scrap and lot-for-lot ordering.
3. A proposed schedule requires 1,200 units per week and each unit uses 0.6 machine hours. Available productive capacity is 650 hours. Determine utilisation, identify the constraint and recommend two managerial responses. Explain the assumptions behind your calculation.

Multiple Choice Questions with Answers

Question 1

Which forecasting method gives equal weight to each observation in the selected historical window?

A. Simple moving average

B. Exponential smoothing

C. Delphi method

D. Causal regression

Correct Answer: A. Simple moving average

Question 2

If actual demand is 500 units, the previous forecast is 460 units and α = 0.30, what is the exponential-smoothing forecast?

A. 448 units

B. 472 units

C. 488 units

D. 500 units

Correct Answer: B. 472 units

Question 3

Which item is a primary input to material requirements planning?

A. Employee appraisal records

B. Bill of materials

C. Corporate tax return

D. Advertising slogan

Correct Answer: B. Bill of materials

Question 4

A parent requires two units of component X and the MPS requires 50 parents. What is the component's gross requirement before inventory adjustments?

A. 25 units

B. 50 units

C. 100 units

D. 150 units

Correct Answer: C. 100 units

Question 5

A component has a gross requirement of 100 units, 20 units available, and safety stock of 10 units. Under lot-for-lot planning with no scheduled receipt, what is the net requirement?

A. 70 units

B. 80 units

C. 90 units

D. 130 units

Correct Answer: C. 90 units

Question 6

What is the purpose of offsetting a planned-order release in MRP?

A. To eliminate the BOM

B. To account for lead time

C. To increase forecast bias

D. To measure supplier quality

Correct Answer: B. To account for lead time

Question 7

A work centre requires 480 hours and has 600 hours of available productive capacity. What is its utilisation?

A. 20%

B. 60%

C. 80%

D. 125%

Correct Answer: C. 80%

Question 8

Which planning process primarily reconciles aggregate demand, supply, inventory and capacity?

A. S&OP

B. BOM explosion

C. Cycle counting

D. Goods receiving

Correct Answer: A. S&OP

Question 9

Which statement best describes DRP?

A. It plans employee recruitment by department.

B. It calculates replenishment requirements across distribution locations.

C. It replaces all supplier contracts.

D. It measures only machine utilisation.

Correct Answer: B. It calculates replenishment requirements across distribution locations.

Question 10

Why can recorded sales understate true demand?

A. Because a stock-out can prevent otherwise possible sales from being observed

B. Because all forecasts are always higher than demand

C. Because BOM quantities are always fixed

D. Because capacity is unlimited

Correct Answer: A. Because a stock-out can prevent otherwise possible sales from being observed