Demand Forecasting and Integrated Planning

Introduction

Supply chains are coordinated systems that transform inputs into products and services for customers. Their performance depends heavily on how well organisations anticipate demand and align purchasing, production, inventory, transport and capacity with that demand. Demand forecasting is not the same as guessing sales. It is a structured process of using information, assumptions and methods to estimate future customer requirements. Integrated planning then converts those estimates into coordinated decisions across functions and business partners.

Poor forecasts can create two costly conditions. If demand is underestimated, the organisation may experience stockouts, lost sales, emergency transport, overtime and dissatisfied customers. If demand is overestimated, it may hold excessive inventory, experience obsolescence, tie up working capital and operate unused capacity. Forecasts will never be perfectly accurate; therefore, professional supply chain management combines forecasting with safety stock, capacity flexibility, scenario planning and regular review.

This chapter explains qualitative and basic quantitative forecasting methods, demand patterns, forecast error and bias. It then examines sales and operations planning (S&OP), integrated business planning (IBP), aggregate planning and the balancing of demand with available capacity.

Learning Objectives

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

1. Explain the purpose, principles and process of demand forecasting.
2. Distinguish qualitative and quantitative forecasting methods.
3. Calculate moving-average forecasts and basic measures of forecast error.
4. Recognise level, trend, seasonal, cyclical and irregular demand patterns.
5. Explain forecast bias and identify causes of systematic error.
6. Describe S&OP, IBP and aggregate planning.
7. Evaluate alternatives for balancing demand and capacity.
8. Apply forecasting and integrated planning concepts to a supply-chain case.

1. The Nature and Purpose of Demand Forecasting

A forecast is an estimate of future demand for a specified product, service, location and time period. The forecast must therefore identify its time horizon, level of aggregation, unit of measure and purpose. A forecast of monthly national demand for a product family is suitable for capacity and workforce planning, whereas a daily forecast for a specific stock-keeping unit (SKU) supports replenishment.

Forecasts are commonly classified as:

• Short-term forecasts: hours to several months; used for scheduling, replenishment and transport.
• Medium-term forecasts: several months to approximately two years; used for workforce, inventory and production planning.
• Long-term forecasts: several years; used for facilities, technology, supplier networks and strategic investment.

Forecast accuracy normally decreases as the time horizon increases. This is not a reason to ignore long-term forecasting; rather, long-term decisions should use ranges, scenarios and assumptions instead of false precision.

Forecasting is also affected by the forecast hierarchy. Demand may be forecast at corporate, market, region, channel, product family, SKU or customer level. A forecast at one level need not equal the sum of independently produced forecasts at another level. Organisations therefore use forecast reconciliation, in which forecasts are adjusted so that values across levels are logically consistent.

A demand forecast is different from a demand plan. A forecast describes what is expected to happen under stated assumptions. A plan describes what the organisation intends to do, including decisions about prices, promotions, capacity and inventory. Confusing these concepts can conceal risks: a sales target is not automatically an unbiased forecast.

2. Forecasting as a Management Process

A sound forecasting process usually follows these stages:

1. Define the decision, horizon, product scope and unit of measure.
2. Collect and clean historical, market and operational data.
3. Examine demand patterns, outliers, promotions and structural changes.
4. Select an appropriate forecasting method.
5. Produce a statistical or judgemental baseline forecast.
6. Obtain informed input from sales, marketing, operations and customers.
7. Document assumptions and approve a consensus forecast.
8. Measure error and bias after actual demand is known.
9. Investigate exceptions and improve the process.

The process is iterative rather than a one-time calculation. A forecast should be time-stamped and its assumptions recorded. For example, “forecast of 10,000 units” is incomplete unless it states the period, product definition, market and assumptions about price, promotion and supply availability.

Data quality is fundamental. Common problems include missing observations, duplicate transactions, returns recorded as negative demand, stockouts that make sales appear lower than true customer demand, changes in product codes and calendar effects. Observed sales are not always equal to unconstrained demand. If a store sells 100 units because only 100 were available, actual customer demand may have been higher.

3. Demand Patterns

Understanding the pattern is a prerequisite for selecting a method.

Seasonality must be distinguished from a one-off event. A festival that occurs every year may create seasonality; a once-only supply disruption is irregular. Trend and seasonality can occur together. For example, online grocery demand may rise over several years while also peaking during holidays.

A useful conceptual relationship is:

This additive representation is appropriate when seasonal fluctuations are approximately constant in units. A multiplicative representation is more suitable when seasonal variation grows with the level of demand:

These are conceptual decompositions, not universal laws. The choice depends on data behaviour and model assumptions.

4. Qualitative Forecasting Methods

Qualitative methods are based primarily on judgement, experience and structured information rather than a numerical time series. They are especially useful when historical data are unavailable, when a new product is introduced or when the market is undergoing structural change.

Executive opinion and expert judgement

Managers estimate future demand using experience and knowledge of markets, customers, competitors and technology. This approach is quick but may be affected by optimism, anchoring, hierarchy or incentives. A documented assumption log and independent estimates can reduce these risks.

Sales-force composite

Sales representatives estimate demand in their territories, and the estimates are combined. Sales staff often possess valuable customer information, but they may understate demand to make targets easier or overstate it to secure inventory. Training, calibration and comparison with actual results are important.

Customer or market research

Surveys, interviews, pilot sales and test markets can reveal purchase intentions and preferences. Stated intention is not identical to actual behaviour, so research should be combined with observed transactions where possible.

Delphi method

The Delphi method gathers anonymous estimates from a panel of experts through several rounds. A facilitator summarises the results between rounds without identifying individuals. Participants revise their views in light of the group response. Anonymity reduces pressure to conform, although results still depend on panel selection and question design.

Scenario analysis

Scenario analysis develops internally consistent alternatives such as high growth, base case and disruption. It is valuable for long-term decisions where a single-point forecast would be misleading. Scenarios should include trigger indicators and corresponding actions.

5. Basic Quantitative Forecasting Methods

Quantitative methods use historical observations and, in some cases, explanatory variables. They work best when the future relationship is expected to resemble the past and when data are sufficiently reliable.

Naïve forecast

A simple naïve forecast sets the next period equal to the latest actual demand:

Here, F_t+1 is the forecast for period t+1, and A_t is actual demand in period t. The method is easy to understand and provides a benchmark. For seasonal demand, a seasonal naïve forecast may use the demand from the same season in the previous cycle:

where s is the number of periods in a seasonal cycle. The seasonal form assumes the corresponding prior season is informative.

Moving average

A moving average uses the mean of the most recent observations:

where n is the number of observations included. All observations have equal weight. A larger n smooths random variation but responds more slowly to changes; a smaller n responds faster but is more sensitive to noise.

Worked example. Weekly demand for the last three weeks was 120, 135 and 150 units. The three-week moving-average forecast for week 4 is:

The forecast is expressed in units per week, consistent with the input data.

Weighted moving average

A weighted moving average assigns greater importance to selected observations:

The weights must normally be non-negative and sum to 1. If the weights are 0.5, 0.3 and 0.2 for the most recent three weeks, using demands of 150, 135 and 120 units gives:

The result may be rounded to a practical whole-unit decision, but the unrounded value should be retained for analysis.

Exponential smoothing

Simple exponential smoothing updates the previous forecast using the latest error:

where 0 ≤ α ≤ 1 is the smoothing constant, A_t is actual demand and F_t is the forecast for period t. A high α gives more weight to recent actual demand and makes the forecast more responsive. Simple smoothing is appropriate mainly for level demand without a strong trend or seasonality.

Worked example. If α = 0.30, actual demand is 150 units and the previous forecast is 135 units:

Trend and causal methods

Trend-adjusted methods are appropriate when demand has a persistent direction. Causal forecasting relates demand to explanatory variables such as price, advertising, income, weather or construction activity. A simple linear relationship may be written as:

where Y is predicted demand, X is an explanatory variable, a is the intercept and b is the estimated change in demand for a one-unit change in X. A relationship is useful only when it is logically plausible, supported by data and stable enough for the planning horizon. Correlation alone does not establish causation.

6. Forecast Error, Accuracy and Bias

A forecast error compares actual demand with forecast demand. A common signed definition is:

Positive error means actual demand exceeded the forecast; negative error means the forecast exceeded actual demand. Some organisations use the reverse sign, so the definition must always be stated.

Mean absolute deviation

MAD is expressed in demand units and measures the average absolute error. It is easy to interpret and does not allow positive and negative errors to cancel.

Mean squared error and root mean squared error

MSE is in squared units, whereas RMSE returns to the original demand units. Squaring gives greater influence to large errors.

Mean absolute percentage error

MAPE is interpretable as a percentage, but it is unsuitable or unstable when actual demand is zero or very close to zero. In such cases, MAD, RMSE or a scale-independent measure should be preferred.

Worked error example. Actual demand for three periods was 100, 120 and 80 units; forecasts were 90, 130 and 100 units.

Thus:

Bias

Bias is systematic error rather than random variation. Using the signed definition above, the mean error is:

A positive mean error indicates consistent under-forecasting; a negative mean error indicates consistent over-forecasting. In the example, mean error is −20 ÷ 3 = −6.67 units, showing over-forecasting on average. Bias can arise from sales incentives, outdated assumptions, failure to incorporate promotions, optimism, pessimism or a method that reacts too slowly to trend.

Accuracy metrics should not be used mechanically. A low average error can conceal large errors in critical periods, and an accurate aggregate forecast may hide serious SKU-level inaccuracies. Forecast performance should therefore be examined by product, customer, time horizon, location and business impact.

7. Sales and Operations Planning

S&OP is a cross-functional, usually monthly, process that aligns demand, supply, inventory, capacity and financial objectives. It operates at an aggregate level, commonly by product family and time period, rather than scheduling every individual order.

A typical S&OP cycle includes:

1. Data and performance review: actual sales, inventory, capacity, service and forecast performance are examined.
2. Demand review: sales and marketing develop a consensus demand view, including promotions and market intelligence.
3. Supply review: operations assess production, procurement, labour, inventory, logistics and supplier constraints.
4. Pre-S&OP reconciliation: alternative plans and gaps are evaluated financially and operationally.
5. Executive S&OP: senior managers approve priorities, trade-offs, scenarios and actions.
6. Implementation and monitoring: decisions are communicated and exceptions tracked.

S&OP is not merely a forecasting meeting. It is a decision process with accountability. A useful output identifies the demand plan, supply plan, assumptions, unresolved gaps, owners and due dates.

8. Integrated Business Planning and Aggregate Planning

Integrated business planning extends S&OP by strengthening strategic, financial and commercial integration. It may connect the operating plan with budgets, product portfolios, capital investment, customer profitability and strategic scenarios. Terminology varies among organisations, so IBP should be defined by its governance and decision rights rather than by its name.

Aggregate planning determines how a firm will meet expected demand over a medium-term horizon using broad resources. It may plan total units, labour hours, machine groups, inventory and subcontracting rather than individual orders. The central challenge is to balance service, cost, workforce stability, flexibility and risk.

A simplified aggregate-planning cost expression is:

This is a planning identity for included cost categories, not a universal complete formula. Organisations should add relevant costs such as transport, setup, idle capacity and obsolescence. The plan must also satisfy constraints such as available labour, machine hours, storage space, supplier limits and legal requirements.

Common strategies include:

• Level strategy: maintain relatively stable production and workforce, using inventory or backorders to absorb demand variation.
• Chase strategy: vary production or workforce to follow demand more closely.
• Mixed strategy: combine regular production, overtime, inventory, temporary labour and subcontracting.
• Demand management: use pricing, promotions, reservations, product substitution or delivery-date commitments to shift or shape demand.

No strategy is universally best. A level strategy may protect workforce stability but increase inventory. A chase strategy may reduce inventory but increase hiring, training and morale costs. Subcontracting adds flexibility but may reduce control over quality, confidentiality or delivery.

9. Demand–Capacity Balancing

Capacity is the maximum feasible output under specified resources and conditions. It should not be confused with theoretical capacity. Practical capacity allows for maintenance, breaks, changeovers, quality checks, absenteeism and variability.

A simple capacity-utilisation measure is:

For example, producing 800 units against a design capacity of 1,000 units gives 80% utilisation. This does not prove that 20% capacity is safely available, because practical capacity may be lower and demand may arrive unevenly.

When demand exceeds capacity, managers may prioritise customers, use overtime, subcontract, add shifts, postpone maintenance, redesign products, transfer demand to another facility or adjust promised dates. When capacity exceeds demand, they may reduce shifts, perform preventive maintenance, build selected inventory, accept additional business or redeploy employees.

Capacity decisions must consider the bottleneck. Increasing non-bottleneck capacity may not increase total system output. The effective plan should compare time-phased demand with time-phased capacity and identify when and where shortages occur. Aggregation can hide local constraints: a national capacity surplus does not solve a shortage at a particular warehouse or supplier.

Figure: The integrated planning cycle links forecast, demand review, supply review, decisions and feedback.

10. Illustrative Case: FreshBite Foods

Background

FreshBite Foods, a fictional regional producer of chilled prepared meals, supplies supermarkets and workplace cafeterias. Demand for its family meal range rises before weekends and during school holidays. The company has experienced stockouts on popular meals while holding excess inventory of slow-moving varieties.

Problem

The sales team provides optimistic monthly estimates, while production uses a three-month moving average without adjusting for promotions or holidays. The result is an over-forecast of some products and an under-forecast of promotional items. Chilled products also have limited shelf life, so excess inventory creates waste.

Analysis

The supply-chain team first separated ordinary demand from promotional sales and identified stockout periods. It calculated MAD and signed mean error by product family. The family meal range showed negative mean error under the definition A−F, indicating systematic over-forecasting. Promotional products showed positive error, indicating under-forecasting. The team also mapped practical capacity and found that packaging, rather than cooking, was the bottleneck during the two days before the weekend.

Approach and solution

FreshBite introduced a monthly S&OP cycle. Marketing supplied a documented promotion calendar; sales estimates were reviewed against historical conversion rates; and operations translated the consensus forecast into packaging-hour requirements. A weighted moving average was used for stable products, while a seasonal-naïve benchmark and management judgement were used for products with strong holiday effects. The company also created high, base and low scenarios for uncertain customer contracts.

To address the bottleneck, FreshBite moved selected preparation activities earlier in the week, trained employees across packaging stations and negotiated limited external packaging support for peak weeks. It reduced forecast overrides by requiring an explanation and expected commercial effect for each adjustment. Forecast error was reviewed monthly, but decisions also considered waste, service level and contribution margin.

Outcome

In this illustrative case, the process reduced avoidable stockouts and waste over subsequent planning cycles, although it did not eliminate uncertainty. The most important improvement was not a single forecasting formula. It was the integration of market information, capacity constraints, financial consequences and accountability in one planning process.

Learning points

1. Forecasting must distinguish unconstrained demand from observed sales.
2. Product-specific patterns require different methods.
3. A forecast should be judged using both statistical error and business impact.
4. The bottleneck determines whether additional capacity will improve service.
5. S&OP creates a forum for resolving cross-functional trade-offs.

11. Technology, Governance and Contemporary Practice

Enterprise resource planning and advanced planning systems can consolidate sales, inventory, orders, capacity and supplier data. Machine-learning tools may identify nonlinear relationships and process large numbers of variables. However, technology does not correct poor master data, biased inputs or unclear decision rights. A complex model can be less useful than a transparent method if users cannot understand or maintain it.

Good governance requires version control, documented assumptions, role-based access, audit trails and clear ownership of overrides. Forecasts may influence production, employment and supplier commitments, so consequential decisions should not be delegated blindly to automated systems. Human review is particularly important when data represent historical inequities, when market conditions change suddenly or when a model uses sensitive customer information.

Contemporary supply chains increasingly use demand sensing, real-time point-of-sale data, Internet-of-Things signals and collaborative forecasts with suppliers and customers. These tools can shorten reaction time, but faster data does not necessarily mean better decisions. Organisations should test whether the signal is reliable, whether the update changes an action and whether frequent changes create operational instability—the so-called nervousness of the plan.

12. Best Practices and Limitations

Effective practice includes:

• define forecast purpose and level clearly;
• segment products by value, volatility, lifecycle and service requirement;
• use simple benchmarks before adopting complex models;
• separate promotions, stockouts and exceptional events from normal demand;
• measure both error and bias;
• reconcile forecasts across organisational levels;
• record assumptions and reasons for overrides;
• connect demand plans to capacity, inventory and financial consequences;
• use scenarios when uncertainty is material;
• review method performance and retire unsuitable methods;
• involve customers and suppliers where collaboration improves visibility.

Forecasting has inherent limitations. Historical data may not represent a changed market. New products have little history. Customers may respond to competitor actions, prices or shortages in ways that are difficult to predict. Forecast accuracy can also be improved statistically while the business outcome worsens if the metric encourages excessive inventory or ignores service priorities. Forecasting should therefore support judgement and structured decisions, not replace responsibility.

Chapter Summary

Demand forecasting estimates future customer requirements for a specified product, location, horizon and unit of measure. Forecasts support replenishment, production, workforce, inventory, capacity and strategic decisions, but they are uncertain and must be combined with buffers, flexibility and scenarios. Demand may display level, trend, seasonality, cycles, irregular variation or structural change.

Qualitative methods include expert judgement, sales-force composite, market research, Delphi and scenario analysis. Basic quantitative methods include naïve forecasting, moving averages, weighted moving averages and exponential smoothing. Method choice depends on data quality, demand pattern, horizon and decision purpose. Forecast errors can be measured using MAD, MSE, RMSE and MAPE, while signed error reveals bias. MAPE requires caution when actual demand is zero or very small.

S&OP integrates demand, supply, inventory, capacity and financial considerations through a recurring cross-functional process. IBP extends this integration toward strategic and financial planning. Aggregate planning determines broad resource and production decisions over the medium term. Level, chase, mixed and demand-management strategies involve different cost, service, workforce and risk trade-offs. Demand–capacity balancing requires attention to practical capacity, bottlenecks, time-phased constraints and the consequences of excess or insufficient capacity.

Key Terms

Forecast: A structured estimate of future demand under stated assumptions.

Demand plan: An agreed view of expected demand and commercial assumptions.

Forecast horizon: The future period covered by a forecast.

Forecast hierarchy: Organisational levels at which demand is forecast, such as region, family or SKU.

Qualitative forecasting: Forecasting based mainly on judgement, expertise or market research.

Quantitative forecasting: Forecasting based on numerical data and formal methods.

Moving average: The mean of a specified number of recent demand observations.

Weighted moving average: A moving average in which observations receive different weights.

Exponential smoothing: A method that updates a forecast using a smoothing constant and recent demand.

Seasonality: Repeating demand variation associated with a calendar or regular cycle.

Forecast error: The difference between actual and forecast demand.

MAD: Mean absolute deviation; average absolute forecast error in demand units.

MAPE: Mean absolute percentage error, subject to problems with zero actual demand.

Forecast bias: Persistent tendency to over-forecast or under-forecast.

S&OP: Sales and operations planning, a cross-functional process aligning demand and supply.

IBP: Integrated business planning, extending operational planning into strategic and financial integration.

Aggregate planning: Medium-term planning of broad resources, output, inventory and workforce.

Practical capacity: Feasible capacity after allowing for operational losses and constraints.

Bottleneck: A resource whose limited capacity restricts system throughput.

Demand–capacity balancing: Coordinating expected requirements with available resources over time.

Review Questions

Short Answer Questions

1. Why is a forecast different from a sales target?
2. What is the difference between seasonality and irregular variation?
3. State two advantages and one limitation of a moving average.
4. What does a negative mean forecast error indicate when error is defined as A−F?
5. Why should practical capacity be distinguished from design capacity?

Long Answer Questions

1. Explain the stages of a professional demand-forecasting process and the role of data quality.
2. Compare qualitative forecasting methods and discuss when each may be appropriate.
3. Explain MAD, RMSE, MAPE and bias, including one limitation of each relevant measure.
4. Describe the S&OP cycle and explain how it differs from simply producing a forecast.
5. Compare level, chase and mixed aggregate-planning strategies, including their operational trade-offs.

Analytical / Application Questions

1. Weekly demand for a product was 200, 220 and 240 units. Calculate the three-week moving-average forecast for the next week. Explain how the forecast would respond if a one-week promotion caused demand of 400 units.
2. A company has actual demand of 500, 600 and 400 units and forecasts of 450, 650 and 500 units. Calculate the signed mean error and MAD. Determine whether the forecasts show over-forecasting or under-forecasting under the definition A−F.
3. A distribution centre has demand exceeding picking capacity every Friday but unused capacity on Monday. Propose an integrated plan using demand management, workforce or scheduling changes, and explain the risks of your proposal.

Multiple Choice Questions with Answers

Question 1

Which statement best describes a demand forecast?

A. A guaranteed sales result

B. An estimate of future demand under stated assumptions

C. A record of historical inventory

D. A legally binding supplier contract

Correct Answer: B. An estimate of future demand under stated assumptions

Question 2

A three-period moving average uses demands of 90, 120 and 150 units. What is the forecast for the next period?

A. 110 units

B. 120 units

C. 130 units

D. 150 units

Correct Answer: B. 120 units

Question 3

Using error defined as A−F, a negative mean error generally indicates:

A. Systematic under-forecasting

B. Systematic over-forecasting

C. No forecast error

D. Increasing capacity utilisation

Correct Answer: B. Systematic over-forecasting

Question 4

Which measure is expressed in the original demand units and uses absolute errors?

A. MAD

B. MAPE

C. MSE

D. Correlation

Correct Answer: A. MAD

Question 5

Which forecasting method gives greater weight to recent observations through a smoothing constant?

A. Delphi method

B. Simple exponential smoothing

C. Customer survey

D. Seasonal index only

Correct Answer: B. Simple exponential smoothing

Question 6

Which demand pattern is associated with repeating calendar-related changes?

A. Trend

B. Seasonality

C. Structural change

D. Random variation

Correct Answer: B. Seasonality

Question 7

The primary purpose of S&OP is to:

A. Replace all operational schedules

B. Align demand, supply, capacity, inventory and financial objectives

C. Maximise production regardless of demand

D. Eliminate all forecast uncertainty

Correct Answer: B. Align demand, supply, capacity, inventory and financial objectives

Question 8

Which strategy normally maintains a relatively stable workforce and uses inventory or backorders to absorb demand variation?

A. Level strategy

B. Chase strategy

C. Pure subcontracting strategy

D. Emergency purchasing strategy

Correct Answer: A. Level strategy

Question 9

If actual output is 800 units and design capacity is 1,000 units, utilisation is:

A. 20%

B. 60%

C. 80%

D. 125%

Correct Answer: C. 80%

Question 10

Why can MAPE be unsuitable for a product with zero actual demand in some periods?

A. It requires capacity data

B. It cannot be calculated meaningfully when the denominator is zero

C. It always produces a negative result

D. It measures only bias

Correct Answer: B. It cannot be calculated meaningfully when the denominator is zero