Demand Forecasting and Integrated Planning

Introduction

Entrepreneurial firms operate under uncertainty. Customers do not purchase at a constant rate, suppliers face capacity and lead-time constraints, and competitors, regulations, technologies and economic conditions change demand. Operations and supply chain planning therefore begins with a disciplined estimate of future requirements. Demand forecasting does not eliminate uncertainty; it makes uncertainty visible and allows managers to choose appropriate capacity, inventory, sourcing, workforce and financial responses.

Forecasting is also a cross-functional activity. Marketing may anticipate a promotion, sales may have information about a major customer, finance may impose a budget constraint, and operations may know that a proposed volume cannot be produced within the available capacity. Integrated planning connects these perspectives through a common set of assumptions and decisions.

This chapter examines qualitative, time-series and causal forecasting; forecast accuracy and bias; uncertain and intermittent demand; aggregate planning; sales and operations planning (S&OP); integrated business planning (IBP); and scenario-based demand–supply balancing.

Learning Objectives

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

1. Select an appropriate qualitative, time-series or causal forecasting method.
2. Calculate and interpret forecast error, MAD, MSE, RMSE, MAPE and bias.
3. Recognize seasonality, trend, uncertainty and intermittent demand.
4. Develop an aggregate plan using capacity, inventory, workforce and subcontracting decisions.
5. Explain the stages and governance of S&OP and IBP.
6. Construct scenarios that balance demand, supply, service and financial objectives.
7. Evaluate the managerial, ethical and technological implications of forecasting decisions.

1. The Nature of Demand and Forecasting

Demand is the quantity customers are willing and able to purchase during a specified period. It differs from sales, which are the transactions actually completed. Sales can be lower than demand because of stockouts, delivery failures or credit restrictions. A forecast based only on historical sales may therefore underestimate true demand.

A useful decomposition is:

• Level: the underlying average demand.
• Trend: a persistent upward or downward movement.
• Seasonality: a repeated pattern linked to calendar or operating cycles.
• Cyclical movement: longer fluctuations associated with economic conditions.
• Random variation: irregular movement that cannot be predicted reliably.
• Exceptional events: promotions, strikes, product launches, disasters or regulatory changes.

The forecast horizon should match the decision. Strategic capacity and facility decisions may require forecasts several years ahead; workforce and procurement decisions may require monthly or quarterly forecasts; replenishment decisions may require daily or weekly forecasts. Forecast uncertainty normally increases as the horizon lengthens.

A forecast is not a target. A target expresses what management wants to achieve, whereas a forecast expresses what is expected under stated assumptions. Confusing these concepts encourages managers to alter an analytical forecast to satisfy a budget, thereby concealing risk.

Forecasting workflow

A professional forecasting process normally includes:

1. Define the item, customer segment, horizon, time bucket and purpose.
2. Clean and reconcile historical demand data.
3. Identify causal events, censored demand and structural changes.
4. Select a baseline method and one or more alternatives.
5. Produce the statistical forecast.
6. Obtain informed judgment through a controlled process.
7. Measure accuracy and bias.
8. Translate the forecast into capacity, inventory, cash and service decisions.
9. Monitor assumptions and revise when evidence changes.

Forecasting at the correct level of aggregation is important. Aggregate demand is often more stable than individual-item demand, a phenomenon known as risk pooling. However, an aggregate forecast cannot directly replace forecasts needed for item-level replenishment.

2. Forecasting Methods

2.1 Qualitative methods

Qualitative methods are appropriate when historical data are limited, when the market is new, or when structural change makes the past a poor guide.

Expert judgment uses the experience of managers, engineers or industry specialists. It is quick but vulnerable to anchoring, overconfidence and political influence.

Delphi forecasting obtains several anonymous rounds of expert estimates. A facilitator summarizes the responses between rounds. Anonymity reduces status effects and encourages revision, although the result still depends on expert selection and question design.

Sales-force composite forecasting collects estimates from sales personnel. Sales staff possess customer knowledge but may understate demand to obtain easy quotas or overstate it to secure inventory.

Market research and customer surveys explore purchase intentions, product preferences and willingness to pay. Stated intentions may not equal actual purchases, especially for innovative products.

Analogy forecasting compares a new product or market with a similar historical product. The analogy must be adjusted for price, distribution, technology, competition and adoption rate.

Qualitative forecasts should be documented with assumptions, evidence and confidence ranges. Combining independent judgments is generally safer than allowing a single influential person to determine the forecast.

2.2 Naïve and time-series methods

Time-series methods use the historical sequence of demand. They are useful when demand patterns are reasonably persistent and causal information is weak or unavailable.

The naïve forecast for the next period equals the most recent actual demand:

Here, A_t is actual demand in period t and F_t+1 is the forecast for period t + 1. The method is appropriate as a simple benchmark, especially for highly volatile or random-walk-like demand.

A seasonal naïve forecast uses demand from the corresponding previous season:

where s is the number of periods in one seasonal cycle. For monthly data with annual seasonality, s = 12.

A moving average with window m is:

The method smooths random variation but responds slowly to turning points. A larger m produces more smoothing and less responsiveness.

A weighted moving average assigns weights to the latest observations:

The weights are dimensionless, and forecast units are the same as demand units. Greater weight on recent observations increases responsiveness.

Simple exponential smoothing is:

where α is the smoothing constant. A high α reacts rapidly to new demand; a low α produces greater smoothing. Simple exponential smoothing assumes no systematic trend or seasonality. An equivalent error-correction form is:

For data with trend, Holt’s method estimates level and trend. One common additive form is:

Here, L_t is the estimated level, T_t is the trend, h is the forecast horizon, and α and β are smoothing constants between 0 and 1. The additive form is most suitable when the trend is approximately constant in absolute units.

When seasonal variation changes proportionally with the level, multiplicative seasonal models may be more appropriate. Seasonal indices should be estimated carefully and reviewed after major changes in product mix or distribution.

2.3 Causal and explanatory forecasting

Causal methods relate demand to explanatory variables such as price, advertising, income, weather, holidays, competitor actions or installed equipment. A simple linear model is:

where Y_t is demand, X_t is an explanatory variable, β₀ is the intercept, β₁ is the estimated change in demand per unit change in X, and ε_t is the error term. The units of β₁ are demand units per unit of X.

With several drivers:

A causal forecast is only useful if the explanatory variables themselves can be forecast or controlled. Correlation does not establish causation. For example, sales and advertising may rise together because management increases advertising when demand is already expected to rise. Omitted variables, multicollinearity, reverse causality and structural breaks can make coefficients misleading.

Regression models should be evaluated using out-of-sample validation rather than fit to historical data alone. A model that explains the past very closely may overfit noise and perform poorly in future periods.

2.4 Choosing and combining methods

No method dominates in every context. Selection should consider data availability, demand pattern, horizon, cost of error, interpretability and response time. A naïve forecast is a mandatory benchmark; a complex model should demonstrate improvement over it.

Forecast combinations often improve robustness. For example, a weighted combination of a time-series and causal forecast can reduce dependence on one model. Weights should be determined using historical validation, not political preference. Human overrides should be recorded, categorized and evaluated after the outcome is known.

3. Forecast Accuracy, Error and Bias

Let actual demand in period t be A_t and forecast demand be F_t. Define forecast error as:

Under this convention, a positive error means actual demand exceeded the forecast; a negative error means the forecast was too high.

Common measures

Here, n is the number of evaluated periods. These measures assume comparable periods and a clearly defined error convention. MAPE becomes undefined when actual demand is zero and unstable when actual demand is close to zero. Weighted MAPE, scale-dependent measures, or mean absolute scaled error may be preferable for intermittent demand.

A useful inventory-oriented measure is forecast error standard deviation. When errors are approximately independent and normally distributed, a sample estimate is:

where ē is mean error. This assumption should not be used automatically: autocorrelation, changing variance and outliers can invalidate safety-stock calculations.

Tracking signal can flag persistent bias:

where RSFE = ∑e_t is the running sum of forecast errors. A large positive tracking signal indicates repeated underforecasting under the stated convention. Control limits must be set by the organization; a universal threshold is not valid for every process.

Worked example: evaluating two forecasts

Suppose actual weekly demand is 100, 120, 80 and 100 units. Forecast A is 90, 130, 90 and 110; Forecast B is 105, 115, 85 and 95.

Errors for A are 10, −10, −10 and −10. Thus, ME = (−20) ÷ 4 = −5 units, MAD = 40 ÷ 4 = 10 units, and RMSE = √[(100 + 100 + 100 + 100) ÷ 4] = 10 units.

Errors for B are −5, 5, −5 and 5. Thus, ME = 0, MAD = 20 ÷ 4 = 5 units, and RMSE = 5 units. Forecast B is both more accurate on these measures and unbiased over the sample. However, four observations are insufficient to establish permanent superiority. Evaluation should use a longer rolling-origin test and consider the cost of underforecasting versus overforecasting.

Accuracy metrics must not be used mechanically. A forecast with lower average error may be worse when stockouts are extremely costly, when demand is lumpy, or when the forecast is used for a different decision horizon.

4. Uncertain, Volatile and Intermittent Demand

Uncertainty has several sources: random customer behavior, forecast-model error, lead-time variability, promotions, new products, supply disruption and incomplete information. Managers should represent uncertainty with ranges or probability distributions rather than a single point estimate whenever decisions are sensitive to risk.

For a single-period decision with underage cost C_u and overage cost C_o, the critical fractile is:

Here, D is demand, Q* is the optimal order quantity under the standard newsvendor assumptions, and the costs are measured per unit. The result means that the chosen quantity should correspond to a demand percentile equal to the ratio of underage cost to total mismatch cost. This model assumes one ordering opportunity, known demand distribution, no replenishment during the selling period and linear mismatch costs.

For replenishment systems, safety stock is often linked to demand and lead-time uncertainty. If demand per period has standard deviation σ_d, lead time is constant L periods, and independent demand is assumed, lead-time demand standard deviation is:

A cycle-service safety stock is:

where z is the standard-normal value associated with the desired cycle-service probability. This formula does not apply unchanged when lead time is variable, demand is autocorrelated, distributions are highly skewed, or the service measure is fill rate rather than cycle service.

Intermittent demand contains many zero-demand periods and occasional non-zero demand. Conventional moving averages and MAPE can perform poorly. Croston-type methods separately estimate the size of a non-zero demand and the interval between non-zero demands. Variants such as SBA and TSB address some bias and obsolescence issues. These methods should be tested against simple benchmarks and evaluated with inventory-relevant measures.

Practical responses to uncertainty include postponement, flexible suppliers, modular product design, capacity reservations, substitution rules, smaller replenishment batches, customer prioritization and explicit service-level segmentation. A forecast alone cannot solve a structural flexibility problem.

5. Aggregate Planning

Aggregate planning converts medium-term demand expectations into a feasible plan for product families, usually over 3–18 months. It balances demand, regular capacity, overtime, inventory, workforce, subcontracting, backlog and sometimes lost sales.

The plan is aggregate because it uses product families and common capacity units rather than detailed stock-keeping units. It should be reconciled later with the master production schedule and material requirements plan.

Typical strategies are:

• Level strategy: maintain a stable workforce and production rate, using inventory or backlog to absorb demand variation.
• Chase strategy: vary production or workforce to follow demand.
• Hybrid strategy: combine stable core capacity with overtime, temporary labor, inventory or subcontracting.

A simplified planning objective can be written as:

subject to the flow balance:

and capacity and non-negativity constraints. Here, P is regular production, I is ending inventory, B is backlog, O is overtime, S is subcontracting, W represents workforce decisions, D is demand, and c terms are unit or period costs. The exact objective depends on the business; costs may include hiring, layoffs, setup, transport, quality, service penalties and cash constraints.

The flow equation states that beginning net inventory plus production and subcontracting, less demand, becomes ending net inventory. Units must be consistent: if demand is units per month, production and inventory must also be units per month or units.

Illustrative aggregate plan

A firm expects demand of 800, 1,000 and 1,200 units in three months. It begins with 100 units of inventory and wants 100 units at the end. Regular capacity is 900 units per month. If the firm uses only regular production, total available units are 100 + 2,700 = 2,800, while total demand plus desired ending inventory is 3,100. The plan therefore has a shortfall of 300 units. Management must use overtime, subcontracting, inventory reduction below the target, backlog or demand-shaping actions. The arithmetic demonstrates why a forecast must be translated into capacity requirements rather than merely circulated as a report.

Aggregate planning is not simply a cost-minimization exercise. Excessive overtime may increase accidents and quality failures; subcontracting may expose intellectual property; layoffs may damage capability; and deliberate backlog may violate customer commitments. A feasible plan must satisfy operational, financial, legal and ethical constraints.

6. S&OP and Integrated Business Planning

Sales and operations planning is a recurring cross-functional process that aligns an unconstrained demand view with supply, inventory, capacity and financial consequences. A typical monthly cycle includes:

1. Data gathering and statistical baseline: collect actuals, inventory, orders, capacity and forecast performance.
2. Demand review: incorporate market intelligence, promotions, product launches and customer information.
3. Supply review: assess capacity, materials, suppliers, labor, inventory and constraints.
4. Pre-S&OP reconciliation: evaluate alternatives, gaps, risks and financial effects.
5. Executive S&OP: senior leaders approve one plan, decisions, owners and escalation items.
6. Execution and learning: communicate decisions, monitor assumptions and measure outcomes.

The process should use one set of numbers, a documented calendar, clear decision rights and exception-based discussion. It is not a meeting in which each function defends its own spreadsheet.

Integrated business planning extends S&OP by explicitly integrating strategic initiatives, product and portfolio decisions, financial planning, cash, risk, sustainability and enterprise scenarios. It is especially valuable for entrepreneurial firms where a capacity choice may affect funding, hiring, customer promises and investor expectations simultaneously.

A mature process distinguishes:

• statistical forecast from consensus forecast;
• demand plan from demand target;
• unconstrained demand from constrained shipment plan;
• assumptions from decisions;
• risks from issues;
• accountability from mere participation.

Key governance questions include: Who can override the model? What evidence is required? How are forecast overrides audited? Which customers receive scarce supply? How are service, margin, working capital and sustainability balanced?

7. Scenario-Based Demand–Supply Balancing

A single forecast hides alternative futures. Scenario planning develops coherent combinations of assumptions, such as base, high-demand and disruption scenarios. Scenarios are not arbitrary optimistic and pessimistic numbers; each should specify drivers, probabilities where defensible, operational implications and trigger indicators.

A useful scenario table contains:

Scenario analysis should test decisions, not merely produce narratives. For each scenario, calculate demand, available capacity, inventory trajectory, service effect, cash requirement and contribution margin. Managers can then distinguish no-regret actions—useful across scenarios—from contingent actions that should be activated only when trigger conditions occur.

Decision trees and sensitivity analysis are useful when probabilities or financial consequences can be estimated. Robust planning may prefer a slightly more expensive plan that performs acceptably across several futures instead of a low-cost plan that fails catastrophically in one scenario.

8. Technology, AI and Data Governance

Modern forecasting platforms combine enterprise-resource-planning data, point-of-sale information, web behavior, promotions, weather, macroeconomic indicators and machine-learning models. Machine learning can identify nonlinear relationships and high-dimensional patterns, but it does not remove data problems or causal ambiguity.

Good practice includes:

• establishing data ownership and common definitions;
• separating training, validation and test periods;
• using rolling-origin evaluation for time-series data;
• monitoring model drift after market or product changes;
• retaining a transparent benchmark model;
• explaining material overrides and automated recommendations;
• protecting customer and supplier data;
• testing for systematic bias across regions, customer groups or access conditions.

AI-generated forecasts may reproduce historical allocation bias. If a historically underserved region received less inventory, a model trained on sales may interpret low sales as low demand. Managers must distinguish observed sales from unconstrained demand and provide governance for overrides, escalation and human accountability.

9. Case Study: FreshBite’s Capacity and Service Decision

The following is an illustrative case.

Background

FreshBite, an entrepreneurial producer of chilled meal kits, sells through supermarkets and an online channel. Demand is seasonal and promotional. The company has 8,000 units of weekly packaging capacity, while a retailer promotion could raise demand to 10,000 units for three weeks. Chilled products have short shelf lives, so excess production cannot be held indefinitely.

Problem

The sales team proposes a forecast of 10,000 units per week. The statistical baseline, based on comparable promotions, is 8,700 units. The operations team can add 1,000 units of overtime capacity, but this increases labor cost and fatigue. A contract packer can provide 1,500 units, but requires a minimum commitment and has a higher defect rate. The retailer imposes a service penalty for short shipment.

Analysis

The team first separates demand scenarios from targets:

• Base: 8,700 units per week.
• Upside: 10,000 units per week if the promotion achieves expected customer traffic.
• Downside: 7,500 units if the retailer’s placement is delayed.

The unconstrained upside demand exceeds internal capacity by 2,000 units. Overtime reduces the gap to 1,000 units; subcontracting can eliminate the remaining gap but increases quality and minimum-volume risk. The team estimates contribution margin, shortage penalty, overtime cost, subcontracting cost and expected waste under each option. It also verifies that the contract packer meets food-safety requirements and that product specifications are controlled.

Approach and solution

FreshBite adopts a hybrid plan: produce 8,000 regular units, reserve 1,000 overtime units, and negotiate an option—not an unconditional commitment—for up to 1,000 subcontracted units. It postpones final labeling for part of production so units can be allocated between supermarket and online orders. A trigger is set: if retailer pre-orders exceed 9,200 units ten days before launch, the subcontracting option is activated. If pre-orders remain below 8,000, overtime is cancelled.

The S&OP decision records the forecast assumptions, the trigger, financial exposure, food-safety approval and customer allocation rule. After the promotion, the company evaluates forecast error, waste, fill rate, labor incidents, margin and the value of the capacity option.

Learning points

1. A forecast should be expressed as scenarios when promotion response is uncertain.
2. Flexible capacity can be more valuable than maximum nominal capacity.
3. Postponement reduces the risk of producing the wrong configuration.
4. Quality and food safety cannot be traded away for forecast accuracy.
5. A documented trigger converts uncertainty into a manageable decision rule.

Challenges and Limitations

Forecasting fails when historical data are inaccurate, definitions change, stockouts censor demand, or organizational incentives reward optimistic promises. Structural breaks caused by pandemics, new competitors, platform changes or regulation cannot be solved by simply increasing model complexity. Small samples make apparent patterns unreliable. Models can also create false precision when managers treat a point estimate as a fact.

Integrated planning can fail through functional conflict, excessive meeting time, weak executive sponsorship, unowned assumptions or plans that are financially impossible. The remedy is not always a better algorithm. It may require better master data, clearer decision rights, improved supplier collaboration, product rationalization or investment in flexible capacity.

Best-Practice Checklist

• Define demand, sales, forecast horizon and units precisely.
• Use a naïve benchmark before adopting complexity.
• Validate methods out of sample and over multiple horizons.
• Report both accuracy and directional bias.
• Avoid MAPE when actual demand is zero or near zero.
• Separate baseline forecasts, judgmental overrides and targets.
• Identify lost sales and stockout-censored demand.
• Aggregate where risk pooling helps, but disaggregate where decisions require it.
• Link forecasts to capacity, inventory, cash and service decisions.
• Use scenarios, triggers and options for high uncertainty.
• Audit overrides, data access, model drift and allocation fairness.
• Review forecast performance and planning assumptions after each cycle.

Chapter Summary

Demand forecasting provides an evidence-based view of future requirements, but it is not a promise or a substitute for managerial judgment. Qualitative methods are valuable when data are limited or structural change is substantial; time-series methods exploit level, trend and seasonality; causal models relate demand to explanatory drivers. A naïve benchmark and out-of-sample validation help prevent unjustified complexity.

Forecast error should be defined consistently as actual minus forecast. ME identifies directional bias, MAD and RMSE measure error magnitude, and MAPE requires caution when actual demand is zero or small. Intermittent demand requires specialized methods and inventory-relevant evaluation. Uncertainty can be managed through probability-based decisions, safety stock, flexible capacity, postponement and scenario planning.

Aggregate planning translates family-level demand into production, workforce, inventory, overtime and sourcing decisions. S&OP aligns demand, supply and finance through a structured cycle, while IBP extends the logic to strategy, risk, cash and sustainability. Effective scenario-based planning combines coherent assumptions, quantified consequences, trigger points and documented governance. Technology and AI can improve scale and responsiveness, but data quality, model drift, transparency, privacy and fairness remain managerial responsibilities.

Key Terms

Forecast: An estimate of future demand under stated assumptions.

Demand: Quantity customers are willing and able to purchase.

Sales: Transactions actually completed.

Time series: Ordered observations recorded over time.

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

Trend: Persistent upward or downward movement in a series.

Naïve forecast: Forecast equal to the most recent actual value.

Moving average: Average of a specified number of recent observations.

Exponential smoothing: Forecasting method that weights observations with geometrically declining weights.

Causal forecast: Forecast based on explanatory variables.

Forecast error: Actual demand minus forecast demand.

Bias: Persistent tendency to overforecast or underforecast.

MAD: Mean absolute deviation, an average absolute error measure.

Intermittent demand: Demand containing many zero periods and occasional non-zero observations.

Aggregate planning: Medium-term planning for product families and common capacity.

S&OP: Cross-functional sales and operations planning process.

Integrated business planning: S&OP extended to strategy, finance, risk and portfolio decisions.

Scenario planning: Analysis of coherent alternative futures and responses.

Postponement: Delaying final configuration or commitment until demand is clearer.

Service level: A specified measure of the ability to satisfy demand, such as cycle service or fill rate.

Review Questions

Short Answer Questions

1. Distinguish demand from sales and explain why the distinction matters for forecasting.
2. What are level, trend and seasonality in a demand time series?
3. State the assumptions of simple exponential smoothing.
4. Why can MAPE be unsuitable for intermittent demand?
5. What is the purpose of an executive S&OP meeting?

Long Answer Questions

1. Compare qualitative, time-series and causal forecasting methods, including their strengths, limitations and appropriate applications.
2. Explain ME, MAD, MSE, RMSE and MAPE. Discuss why no single metric is sufficient for all planning decisions.
3. Describe how a firm should design an S&OP process from data gathering through execution and learning.
4. Explain the differences among level, chase and hybrid aggregate-planning strategies.
5. Critically evaluate the role of AI and machine learning in demand forecasting, including data governance and fairness risks.

Analytical / Application Questions

1. Actual demand for three periods is 200, 240 and 180 units, while forecasts are 220, 230 and 200. Calculate the ME and MAD using error = actual − forecast, and interpret the sign of ME.
2. A firm has monthly demand of 1,000, 1,300 and 1,600 units, beginning inventory of 200 units, and regular capacity of 1,200 units per month. Determine the ending inventory after each month if no overtime, subcontracting or backlog is allowed. Explain the feasibility implication.
3. Design base, upside and disruption scenarios for a product launch. For each scenario specify a demand driver, supply condition, trigger indicator and management response.

Multiple Choice Questions with Answers

Question 1

Which method is most appropriate as a benchmark for many demand series?

A. Naïve forecasting

B. Delphi forecasting

C. Market research only

D. Executive judgment only

Correct Answer: A. Naïve forecasting

Question 2

If forecast error is defined as actual demand minus forecast demand, a positive mean error indicates:

A. Persistent overforecasting

B. Persistent underforecasting

C. Zero demand variability

D. Perfect forecast accuracy

Correct Answer: B. Persistent underforecasting

Question 3

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

A. MAPE

B. MSE

C. MAD

D. Tracking signal

Correct Answer: C. MAD

Question 4

Simple exponential smoothing is primarily designed for data with:

A. No level but strong seasonality

B. Level without systematic trend or seasonality

C. Only causal relationships

D. Guaranteed intermittent demand

Correct Answer: B. Level without systematic trend or seasonality

Question 5

For monthly data with annual seasonality, a seasonal naïve forecast for next January would generally use:

A. The most recent week

B. The average of all months

C. The previous January’s demand

D. The previous December’s forecast error

Correct Answer: C. The previous January’s demand

Question 6

Why is MAPE problematic when actual demand is zero?

A. It becomes undefined because of division by zero

B. It always equals zero

C. It measures only bias

D. It cannot be calculated for positive demand

Correct Answer: A. It becomes undefined because of division by zero

Question 7

Which activity is characteristic of aggregate planning?

A. Selecting a barcode format

B. Balancing workforce, inventory, capacity and subcontracting over the medium term

C. Designing an individual customer invoice

D. Choosing a product logo

Correct Answer: B. Balancing workforce, inventory, capacity and subcontracting over the medium term

Question 8

In S&OP, the supply review primarily examines:

A. Only advertising slogans

B. Capacity, materials, suppliers, labor and inventory constraints

C. Only historical revenue recognition

D. Employee holiday preferences

Correct Answer: B. Capacity, materials, suppliers, labor and inventory constraints

Question 9

A forecast scenario should ideally include:

A. A single unexplained number

B. A coherent driver, supply condition, implication and response

C. Only the most optimistic outcome

D. A target presented as a forecast

Correct Answer: B. A coherent driver, supply condition, implication and response

Question 10

A major governance risk in AI-based forecasting is that the model may:

A. Eliminate all uncertainty

B. Reproduce historical allocation bias contained in sales data

C. Make data definitions unnecessary

D. Guarantee causal interpretation

Correct Answer: B. Reproduce historical allocation bias contained in sales data