Demand Forecasting and Sales and Operations Planning

Introduction

Entrepreneurial ventures rarely fail because demand is completely unknowable. They fail because managers treat uncertain demand as certain, use weak information, or fail to connect the forecast with purchasing, production, staffing, cash flow, and customer commitments. Demand forecasting is the disciplined process of estimating future customer requirements. Sales and operations planning (S&OP) is the cross-functional process that converts those estimates into an agreed, feasible plan for demand, supply, inventory, capacity, and finance.

Forecasts do not eliminate uncertainty. Their purpose is to make uncertainty visible, quantify its likely effects, and support better decisions. A forecast for a new food-delivery venture, for example, affects courier capacity, packaging purchases, working capital, promotional timing, and service promises. The forecast is therefore not merely a statistical output; it is an input to managerial judgment and coordinated action.

Learning Objectives

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

1. Explain the role, scope, and limitations of demand forecasting.
2. Select suitable qualitative and quantitative forecasting methods.
3. Calculate and interpret forecast error measures.
4. Recognize demand uncertainty and use it in operational decisions.
5. Describe the S&OP cycle and its relationship to marketing, operations, and finance.
6. Explain collaborative forecasting and the responsible use of digital and AI tools.
7. Apply forecasting and S&OP principles to an entrepreneurial business case.

1. The Nature and Purpose of Demand Forecasting

Demand is the quantity customers are willing and able to purchase under specified conditions. Sales are the transactions actually recorded. The two may differ because of stockouts, pricing, distribution limitations, unavailable service capacity, or inaccurate order capture. Forecasting should therefore distinguish between observed sales and underlying demand whenever possible.

A forecast is an estimate of a future value or set of future values made using information available at the time. A forecast horizon is the period into the future covered by the estimate. Short-term forecasts support replenishment and scheduling; medium-term forecasts support workforce, inventory, and capacity decisions; long-term forecasts support facility, technology, and market-entry decisions.

Forecasts can be generated at several levels:

• Aggregate level: total demand for a product family, region, or business.
• Disaggregate level: demand for a specific stock-keeping unit (SKU), channel, or customer.
• Independent demand: demand originating from external customers.
• Dependent demand: demand derived from another item or activity, such as packaging required for finished units.

A useful forecast is not necessarily the one with the smallest historical error. It must also be timely, understandable, stable enough for planning, and appropriate to the decision. Forecasting a new product using a highly complex model may create false confidence when little relevant data exist.

Demand patterns

Four patterns are especially important:

• Level: demand fluctuates around a relatively stable average.
• Trend: demand rises or falls over time.
• Seasonality: a repeating pattern occurs at known intervals, such as higher beverage demand in summer.
• Cyclical or irregular movement: demand responds to economic conditions, competitors, unusual events, or random variation.

Demand may also contain promotions, price effects, cannibalization between products, and structural breaks. A temporary advertising campaign should not automatically be interpreted as a permanent trend.

2. Qualitative Forecasting Methods

Qualitative methods rely primarily on informed judgment rather than a long numerical history. They are particularly valuable for new products, market disruptions, technological change, and situations in which historical data are unavailable or no longer representative.

Executive opinion and expert judgment

Managers with knowledge of customers, competitors, and operations estimate future demand. The method is quick and inexpensive, but it may suffer from hierarchy effects, optimism, anchoring, and deliberate manipulation of estimates. A documented assumption log and independent estimates before group discussion can reduce these weaknesses.

Sales-force composite

Sales representatives estimate demand in their territories or customer segments. Because sales staff interact with customers, they may detect emerging needs early. However, they may understate demand to create an easier target or overstate it to obtain inventory. Training, historical calibration, and review by customer segment improve reliability.

Customer surveys and market research

Surveys, interviews, concept tests, pilot sales, and conjoint studies can estimate customer interest and willingness to pay. Stated intention is not identical to actual purchase behavior, so research should be combined with behavioral evidence where possible.

Delphi method

The Delphi method obtains forecasts from a panel of experts through multiple anonymous rounds. A coordinator summarizes responses between rounds, allowing participants to revise their views without pressure from dominant individuals. Delphi is useful when uncertainty is high, but results depend on panel composition, question design, and the quality of information supplied.

Analogy and scenario analysis

Analogy uses the experience of a similar product, market, or technology. Scenario analysis develops coherent alternative futures, such as conservative, expected, and high-growth cases. Scenarios are not probabilities unless probabilities are explicitly assigned; they are structured ways of examining assumptions and preparedness.

3. Quantitative Forecasting Methods

Quantitative methods use historical observations and mathematical relationships. They are most useful when data are sufficiently reliable and the future is expected to retain some relationship with the past.

Naïve forecast

The simplest method sets the next-period forecast equal to the most recent actual observation:

Here, F_t+1 is the forecast for period t+1, and A_t is actual demand in period t. The method is often a useful benchmark. For strongly seasonal data, a seasonal naïve method uses the observation from the same season in the previous cycle:

where s is the number of periods in a seasonal cycle, such as 12 for monthly annual seasonality. These methods assume that recent or corresponding-period demand is informative and do not correct for structural change.

Moving average

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

All observations receive equal weight. A larger n smooths random variation but responds more slowly to changes; a smaller n responds faster but is less smooth.

Example. Demand for the last three weeks was 120, 135, and 150 units. The three-week forecast is:

The result is in units because every input is in units and division is by a dimensionless count.

Weighted moving average

A weighted moving average gives different importance to observations:

where w_i is the weight assigned to an observation and the weights must satisfy w_i ≥ 0 and ∑w_i = 1. More recent observations commonly receive larger weights. The method is appropriate when recency matters but a transparent calculation is preferred.

Exponential smoothing

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

Here, α is the smoothing constant, with 0 < α < 1. A high value responds quickly to new observations; a low value produces more smoothing. Basic exponential smoothing is suitable for data with no systematic trend or seasonality. It should not be used alone when persistent trend or seasonal effects are evident.

Example. If actual demand is 200 units, the previous forecast is 180 units, and α = 0.30, then:

The new forecast moves 6 units toward the latest actual value.

Trend and seasonal models

When trend exists, trend-adjusted exponential smoothing or regression may be appropriate. A simple linear trend model is:

where y_t is demand at time t, a is the intercept, b is estimated demand change per period, and ε_t is random error. The model assumes a reasonably stable linear relationship over the relevant range. Extrapolation far beyond the observed data is risky.

A seasonal index compares typical demand in a season with average demand. If a quarterly seasonal index is 1.20, demand in that quarter is estimated to be 20% above the nonseasonal level. Seasonal indices must be calculated consistently and adjusted carefully when trends or promotions distort historical observations.

Causal forecasting

Causal models relate demand to explanatory variables such as price, advertising expenditure, income, weather, or distribution reach. A simple model may be expressed as:

where D is demand, P is price, A is advertising expenditure, β₀, β₁, and β₂ are estimated coefficients, and ε captures unobserved influences. Association is not automatically causation. Managers should consider reverse causality, omitted variables, multicollinearity, and whether future values of the explanatory variables are themselves known.

4. Measuring Forecast Error and Bias

Let the forecast error for period t be:

A positive error means actual demand exceeded the forecast; a negative error means the forecast exceeded actual demand. This sign convention must be stated because some organizations use the reverse convention.

Common measures

Here, n is the number of evaluated periods. MAPE is undefined when A_t = 0 and can become unstable when actual demand is near zero. It may also overemphasize errors on low-volume items. Weighted measures or scale-dependent measures such as MAD may be preferable.

Worked example. Actual demand over three periods was 100, 120, and 80 units; forecasts were 90, 130, and 100 units. Errors are 10, −10, and −20 units. Therefore:

The negative ME indicates overforecasting overall, while MAD describes the typical absolute miss. One measure alone is insufficient.

Bias and tracking signal

Bias is a persistent directional error, not merely a large random error. One common tracking signal is:

where RSFE = ∑e_t is the running sum of forecast errors. A large positive or negative value may signal systematic underforecasting or overforecasting. Control limits should be defined by the organization; there is no universally correct threshold for every context. Managers should investigate the cause rather than mechanically changing the model.

Forecast evaluation must use a time-ordered holdout period or rolling-origin testing. Evaluating a model on the same data used to fit it can produce overly optimistic results. Comparisons should also reflect business consequences: a forecast error that causes a stockout may be more costly than one that causes modest excess inventory.

5. Managing Demand Uncertainty

A point forecast is only one estimate. A responsible plan also considers a range of outcomes. Uncertainty arises from random variation, imperfect information, forecast-model limitations, competitor actions, economic change, and managerial decisions such as promotions.

Useful practices include:

• producing low, expected, and high scenarios;
• identifying assumptions and their owners;
• separating baseline demand from promotional uplift;
• using probability distributions or forecast intervals when justified;
• postponing product differentiation when possible;
• maintaining flexible capacity and supplier options;
• holding inventory or capacity protection where the cost of shortage is high;
• monitoring leading indicators such as web traffic, quotations, reservations, and cancellations.

Safety stock is a buffer against uncertainty, but its calculation depends on the replenishment system, service objective, and assumptions about variability. For a simple independent-demand case with a known lead time, a commonly used expression is:

where SS is safety stock in units, z is the standard-normal factor associated with the target cycle-service probability, and σ_L is the standard deviation of demand during lead time. This expression assumes that lead-time demand is approximately normally distributed and that σ_L is already measured over the lead-time horizon. If both demand and lead time vary, a different variance model is required; the simple expression should not be applied automatically.

Risk pooling can reduce aggregate variability when demands are not perfectly positively correlated. However, centralization may increase delivery distance or response time. The correct decision balances statistical benefits with service and logistics costs.

6. Sales and Operations Planning

S&OP is a recurring, cross-functional process through which an organization reconciles market demand with operational and financial capability. It is usually conducted at an aggregate product-family or business-unit level over a medium-term horizon, often several months to two years. The exact frequency and horizon should fit the business.

A typical S&OP cycle includes:

1. Data gathering and portfolio review: collect actual sales, inventory, orders, forecast performance, capacity, supplier information, and financial results.
2. Demand review: develop an unconstrained demand plan incorporating statistical forecasts, sales intelligence, marketing campaigns, customer commitments, and assumptions.
3. Supply review: test capacity, labor, materials, inventory, supplier constraints, lead times, and logistics against the demand plan.
4. Reconciliation and scenario analysis: compare alternatives such as overtime, subcontracting, inventory building, delayed promotion, price changes, or selective allocation.
5. Executive review and decision: approve one plan, document trade-offs, assign actions, and escalate unresolved issues.
6. Implementation and monitoring: translate the aggregate plan into detailed schedules, budgets, purchasing decisions, and performance reviews.

S&OP is not simply a meeting and is not identical to a production schedule. It is a decision process that aligns three views of the business:

• Marketing and sales: customers, channels, campaigns, prices, market growth, and competitive moves.
• Operations: capacity, labor, inventory, procurement, quality, and service feasibility.
• Finance: revenue, margin, cash, working capital, investment, and risk.

Typical planning strategies

A level strategy maintains relatively stable production and uses inventory or backlog to absorb demand changes. It supports workforce stability but may increase inventory or shortages. A chase strategy adjusts production or workforce to follow demand. It reduces inventory but may increase hiring, overtime, training, and quality risks. A mixed strategy combines inventory, flexible labor, subcontracting, pricing, and demand management.

Plans should be evaluated using relevant costs and constraints, not only forecast accuracy. A simple aggregate balance is:

If backorders are permitted, ending inventory may be negative under the chosen convention. Units must be consistent across all terms. The equation is an accounting flow relationship, not a complete optimization model; it does not by itself include capacity, costs, service targets, or lead times.

A feasible plan should also satisfy capacity constraints, for example:

Available capacity may depend on labor hours, equipment hours, maintenance, skills, supplier limits, and planned downtime.

7. Collaborative Forecasting and Information Governance

Collaborative forecasting shares information and planning responsibilities across supply-chain partners. Examples include collaborative planning, forecasting, and replenishment (CPFR), vendor-managed inventory, joint business planning, and supplier forecasts based on point-of-sale data.

Collaboration can reduce the bullwhip effect, in which small changes in consumer demand create increasingly large order fluctuations upstream. Sharing actual consumption, inventory positions, promotions, planned product launches, and replenishment rules helps partners distinguish real demand from ordering behavior.

Effective collaboration requires:

• a common definition of demand and forecast horizon;
• agreed data ownership, frequency, format, and quality rules;
• documented promotion and product-change assumptions;
• exception thresholds and escalation procedures;
• clear responsibility for forecast overrides;
• confidentiality, cybersecurity, and appropriate access controls;
• performance measures that discourage local optimization.

Information sharing does not guarantee cooperation. Partners may withhold information, inflate forecasts, or use shared data to strengthen bargaining positions. Governance should therefore specify permitted uses, auditability, dispute resolution, and protection of commercially sensitive information.

8. Technology, Analytics, and AI

Enterprise resource planning systems, point-of-sale platforms, customer relationship management systems, cloud collaboration tools, and business-intelligence dashboards support forecasting and S&OP. Machine-learning models can identify nonlinear relationships, interactions, and large numbers of signals. They may be useful when data volume and quality are sufficient.

Nevertheless, predictive performance is not the same as managerial usefulness. AI systems can reproduce biased historical decisions, fail during structural breaks, obscure assumptions, or generate apparently precise outputs from poor data. Responsible use requires data validation, version control, back-testing, monitoring for drift, explainable inputs where decisions affect stakeholders, and human review of exceptional situations. Forecast overrides should be logged with reasons so that judgment can be evaluated rather than treated as unquestionable expertise.

A practical architecture often combines methods: a statistical baseline, documented commercial adjustments, scenario analysis, and an approval workflow. The best model is the one that improves decisions at an acceptable total cost, not necessarily the most sophisticated algorithm.

9. Illustrative Case: FreshBite Subscription Meals

Background

FreshBite is a fictional early-stage venture delivering refrigerated subscription meals to university students. It offers three product families: standard meals, vegetarian meals, and high-protein meals. Demand is seasonal because enrollment, examinations, holidays, and summer residence patterns affect purchasing.

Problem

FreshBite’s founders used sales-force judgment and increased production whenever orders rose. This caused frequent stockouts of vegetarian meals, excess waste in standard meals, and cash pressure from buying ingredients too early. Marketing planned a discount campaign without confirming kitchen capacity or supplier lead times.

Analysis

The operations manager separated recorded sales from lost sales caused by stockouts and created a weekly demand history by product family. A three-week moving average provided a transparent baseline, while a seasonal review identified lower demand during university holidays. Marketing supplied the campaign dates and expected customer acquisition. Finance translated alternative plans into cash and contribution-margin effects.

For the next week, observed demand for vegetarian meals in the last three weeks was 180, 210, and 240 units. The moving-average baseline was:

Because a promotion was expected to add 15% to baseline demand, the unconstrained promotional estimate was:

The kitchen could produce only 230 vegetarian meals after allowing for labor and maintenance. The S&OP team considered four alternatives: limit the promotion, shift labor from standard meals, outsource a qualified component, or accept a service shortfall. It selected a smaller promotion and reserved supplier capacity for the following week. The team also created low, expected, and high scenarios and a trigger based on preorders.

Outcome

Stockouts fell, food waste declined, and the promotion was rescheduled to a period with greater capacity. The founders gained a weekly cross-functional review rather than relying on informal arguments. Forecast error was tracked using MAD and bias, while the business monitored preorder conversion as a leading indicator.

Learning points

1. Sales history may understate true demand when stockouts exist.
2. Marketing actions must be represented explicitly rather than hidden in a baseline forecast.
3. A demand plan is not feasible until capacity, materials, and finance are considered.
4. Transparent simple methods can be valuable for a young venture.
5. Forecasts should trigger decisions and contingencies, not merely populate a spreadsheet.

10. Challenges, Ethics, and Best Practices

Forecasting can be distorted by poor master data, changing product definitions, missing sales, stockouts, one-time events, and inconsistent calendars. Aggregating too early can hide SKU-specific shortages; disaggregating too far can produce noisy estimates and excessive administrative work. Managers should match forecast detail to the decision and review data quality before model selection.

Ethically, organizations should avoid deliberately biased forecasts intended to manipulate budgets, suppliers, employees, or investors. Customer data must be collected and used lawfully, transparently, and only for legitimate purposes. Sensitive commercial information shared with partners requires appropriate security. Forecasts that affect staffing or customer access should be reviewed for discriminatory effects and unjustified assumptions.

Best practice is to:

• define the forecasting purpose, horizon, unit, and owner;
• establish a clean, consistent history and document missing or censored demand;
• compare a baseline with more advanced alternatives;
• use time-based validation and monitor error by segment;
• measure both accuracy and bias;
• distinguish statistical forecasts from commercial overrides;
• quantify uncertainty and prepare contingencies;
• conduct S&OP with accountable representatives from sales, marketing, operations, and finance;
• review assumptions after promotions, disruptions, and structural changes;
• keep a decision record and learn from forecast misses.

Chapter Summary

Demand forecasting estimates future customer requirements so that entrepreneurs can coordinate resources, capacity, inventory, finance, and customer service. Qualitative approaches such as expert judgment, surveys, Delphi, analogy, and scenarios are especially useful when data are limited or structural change is substantial. Quantitative approaches include naïve methods, moving averages, exponential smoothing, trend models, seasonal methods, and causal models. Method selection must reflect demand patterns, data quality, decision horizon, and the consequences of error.

Forecast error can be measured through ME, MAD, MSE, RMSE, and MAPE, but each measure has limitations. Bias indicates systematic over- or underforecasting and should be investigated. Demand uncertainty should be represented through scenarios, ranges, forecast intervals, flexible capacity, and appropriate inventory or capacity buffers.

S&OP integrates the demand plan with operational and financial capability through data gathering, demand review, supply review, reconciliation, executive decisions, and implementation. Collaborative forecasting improves visibility across supply-chain partners but requires governance, common definitions, data quality, and trust. Technology and AI can strengthen forecasting, but they do not replace human accountability, ethical judgment, or disciplined process design.

Key Terms

Demand forecasting: Estimating future customer requirements.

Forecast horizon: The future period covered by a forecast.

Independent demand: Demand generated by external customers.

Dependent demand: Demand derived from another item or activity.

Seasonality: A recurring demand pattern at a known interval.

Naïve forecast: A forecast based on a recent or corresponding-period actual value.

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

Exponential smoothing: A method that updates forecasts using a smoothing constant.

Causal forecasting: Forecasting demand from explanatory variables.

Forecast error: The difference between actual demand and forecast demand.

Bias: Persistent directional forecast error.

MAD: Mean absolute deviation, expressed in demand units.

MAPE: Mean absolute percentage error.

Safety stock: Inventory held to protect against uncertainty.

S&OP: A cross-functional process aligning demand, supply, and financial plans.

Level strategy: A plan maintaining relatively stable production.

Chase strategy: A plan adjusting capacity or output to follow demand.

CPFR: Collaborative planning, forecasting, and replenishment.

Bullwhip effect: Amplification of demand variability upstream in a supply chain.

Forecast override: A deliberate adjustment to a statistical forecast.

Review Questions

Short Answer Questions

1. Distinguish demand from recorded sales and explain why the distinction matters.
2. What is the difference between a level strategy and a chase strategy?
3. State two limitations of MAPE.
4. Why is a forecast baseline often useful in collaborative planning?
5. Identify the principal functional areas involved in S&OP.

Long Answer Questions

1. Compare qualitative and quantitative forecasting methods, including their assumptions, strengths, and limitations.
2. Explain the typical S&OP cycle and show how marketing, operations, and finance contribute to it.
3. Discuss how a firm can manage demand uncertainty beyond producing a single point forecast.
4. Explain forecast error, bias, MAD, RMSE, and MAPE, and discuss when each may be useful.
5. Evaluate the opportunities and risks of using AI in demand forecasting and S&OP.

Analytical / Application Questions

1. A venture records demand of 90, 110, and 130 units in three successive weeks. Calculate the three-week moving-average forecast for week four. Explain one reason the result might be inappropriate.
2. A retailer’s actual demand is 100, 120, and 80 units, while forecasts are 90, 130, and 100 units. Calculate ME and MAD using e_t = A_t − F_t, and interpret the signs.
3. A promotion is expected to increase a baseline forecast by 20%, but available capacity is only 95% of the promotional estimate. Recommend an S&OP response and identify the information needed before making the final decision.

Multiple Choice Questions with Answers

Question 1

Which statement best describes the purpose of demand forecasting?

A. To eliminate all uncertainty from operations

B. To estimate future requirements and support decisions under uncertainty

C. To guarantee that actual sales equal planned sales

D. To replace financial planning with a statistical model

Correct Answer: B. To estimate future requirements and support decisions under uncertainty

Question 2

A three-period moving average uses demands of 120, 135, and 150 units. What is the next forecast?

A. 125 units

B. 130 units

C. 135 units

D. 145 units

Correct Answer: C. 135 units

Question 3

If actual demand is 200 units, the previous forecast is 180 units, and α = 0.30, what is the exponential-smoothing forecast?

A. 180 units

B. 186 units

C. 194 units

D. 200 units

Correct Answer: B. 186 units

Question 4

Using e_t = A_t − F_t, what does a negative mean error generally indicate?

A. Systematic underforecasting

B. Systematic overforecasting

C. No forecast error

D. Increasing seasonal demand

Correct Answer: B. Systematic overforecasting

Question 5

Which measure expresses average absolute forecast error in the original demand units?

A. MAD

B. MAPE

C. MSE

D. Tracking signal

Correct Answer: A. MAD

Question 6

Which forecasting method is particularly appropriate when expert views are needed anonymously through multiple rounds?

A. Naïve forecasting

B. Delphi method

C. Simple moving average

D. Seasonal index

Correct Answer: B. Delphi method

Question 7

What is the primary purpose of S&OP?

A. To create only the daily production schedule

B. To reconcile demand, operational capability, and financial objectives

C. To allow marketing to set production independently

D. To measure employee attendance

Correct Answer: B. To reconcile demand, operational capability, and financial objectives

Question 8

If beginning inventory is 500 units, production is 800 units, and demand is 900 units, what is ending inventory?

A. 300 units

B. 400 units

C. 600 units

D. 1,200 units

Correct Answer: B. 400 units

Question 9

Which situation makes ordinary MAPE problematic?

A. All actual demands are large and positive

B. Actual demand is zero or very close to zero

C. Forecasts are generated monthly

D. The forecast horizon is short

Correct Answer: B. Actual demand is zero or very close to zero

Question 10

Which practice most directly supports collaborative forecasting?

A. Keeping promotion plans confidential from supply partners

B. Using different definitions of demand across organizations

C. Sharing agreed data, assumptions, and exception rules

D. Allowing every partner to change the forecast without a record

Correct Answer: C. Sharing agreed data, assumptions, and exception rules