3. Forecasting Process and Managerial Judgment

Introduction and Context

Demand forecasting is not merely the production of a numerical estimate. It is an organizational process through which a firm forms a reasoned view of future demand and converts that view into decisions about purchasing, production, inventory, capacity, transport, staffing, and customer service. In logistics, forecast quality matters because supply-chain decisions are made before actual demand is observed. A poor forecast can create excess inventory, stock-outs, emergency freight, idle capacity, or unnecessary working capital.

Forecasting is often presented as a choice between statistical models and managerial intuition. This is a misleading distinction. Statistical methods provide consistency, transparency, and a useful baseline, while managerial judgment contributes knowledge about customers, competitors, promotions, regulation, new products, and operational events that may not be represented in historical data. The central managerial challenge is to combine these sources without allowing authority, optimism, political pressure, or untested assumptions to distort the forecast.

A disciplined forecasting process therefore begins before a method is selected. The organization must clarify what is being forecast, for whom, at what level of detail, over what horizon, under which assumptions, and for what decision. Only then can it decide whether to use time-series analysis, causal models, market research, expert judgment, or a combination of approaches.

Learning Objectives

After studying this chapter, students should be able to:

1. Design a forecasting process that aligns objectives, granularity, horizons, assumptions, and decision rights.

2. Distinguish statistical forecasts from judgmental adjustments and explain how they should interact.

3. Apply Delphi, market research, sales-force composite, analogy, and structured judgment methods.

4. Identify cognitive, organizational, and strategic sources of forecasting bias.

5. Establish consensus processes, documentation standards, and governance controls.

6. Evaluate forecasting approaches in practical logistics and supply-chain situations.

Forecasting as a Managed Decision Process

A forecast has meaning only in relation to a decision. A distribution centre may need a daily forecast by product family to schedule labour, whereas a manufacturing network may require a monthly forecast by region for capacity planning. A board may need an annual demand outlook for investment decisions. These are not simply different versions of the same forecast; they involve different horizons, aggregation levels, uncertainties, and consequences.

The first task is to define the forecasting objective. Typical objectives include estimating replenishment requirements, planning production, allocating constrained inventory, budgeting transportation capacity, setting safety stocks, evaluating a market entry, or assessing the financial consequences of a commercial plan. A forecast designed for sales targets should not automatically be used as an unbiased operational forecast. Sales targets often express ambition, whereas an operational forecast should represent the most probable demand under stated conditions.

A useful forecasting specification answers five questions:

• What variable is being forecast: orders, shipments, consumption, revenue, or customer demand?
• At what level will it be forecast: item, category, location, customer, channel, or network?
• For which time horizon and review frequency?
• Under what assumptions about price, promotion, supply availability, competitors, and the economy?
• Which decision will the forecast support, and how will forecast error affect that decision?

The distinction between demand and sales is especially important. Observed sales may be lower than underlying demand because inventory was unavailable, deliveries were delayed, credit was restricted, or a warehouse had insufficient capacity. If constrained sales are treated as true demand, the forecasting system may systematically underestimate future requirements. Conversely, unusual forward buying may make recorded orders temporarily higher than sustainable consumption.

Granularity and Aggregation

Granularity refers to the level of detail at which forecasts are prepared. A highly disaggregated forecast may identify demand for every stock-keeping unit, customer, and location. Such detail appears precise but may produce unstable estimates, especially for intermittent or low-volume items. Aggregated forecasts are generally more stable, but they may conceal important differences between products, markets, or channels.

The appropriate level is determined by decision needs, data quality, demand behaviour, and the cost of error. Forecasting at the level at which replenishment decisions are made is often useful, but it is not always optimal. Forecasts can be generated at several levels and reconciled so that item-level figures add to category, regional, and corporate totals. This prevents different departments from working with incompatible versions of expected demand.

Forecast granularity should also reflect product life-cycle stage. New products have little or no history and may require analogies, market research, or expert assessment. Mature products may benefit from statistical models. Declining products require assumptions about substitution, phase-out, and residual demand. A single forecasting method rarely performs equally well across all these conditions.

Forecast Horizons

The forecast horizon should correspond to the lead time and reversibility of the decision. Short-term forecasts support order release, labour scheduling, transport planning, and near-term inventory control. Medium-term forecasts support supplier commitments, production planning, promotions, and capacity balancing. Long-term forecasts inform network design, capital investment, market entry, and strategic sourcing.

Uncertainty generally increases as the horizon extends, but the type of uncertainty also changes. Near-term variation may arise from order timing and promotions; long-term uncertainty may arise from technology, regulation, competitive structure, or changing customer preferences. Long-horizon forecasts should therefore be expressed through scenarios and ranges rather than false precision.

A rolling forecast updates the outlook regularly as new information becomes available. Rolling processes are useful, but frequent revision should not become uncontrolled volatility. Organizations should distinguish genuine changes in expected demand from normal random variation and should document why major revisions were made.

A Disciplined Forecasting Workflow

A robust forecasting process can be viewed as a sequence of linked activities.

1. Define the purpose and decision owner

The process begins by identifying the decision that the forecast will inform and the person accountable for using it. A forecast without an owner may become a reporting exercise rather than a decision tool. The decision owner also clarifies the cost of over-forecasting compared with under-forecasting. In a spare-parts operation, under-forecasting may cause severe service failure; in a perishable-food operation, over-forecasting may cause waste.

2. Establish the baseline

Historical data are cleaned, classified, and examined for stock-outs, one-time orders, returns, promotions, calendar effects, and changes in distribution. A statistical or operational baseline is then prepared. The baseline should be treated as a transparent starting point, not as an unquestionable truth.

3. Gather assumptions and external information

Relevant information may include confirmed promotions, price changes, new customer contracts, competitor actions, macroeconomic indicators, weather conditions, regulatory changes, product substitutions, and supply constraints. Each assumption should have an owner, a source, an effective period, and an assessment of confidence.

4. Obtain independent judgment

Judgment should be collected from people with relevant knowledge, such as sales managers, category specialists, procurement professionals, customers, engineers, or industry experts. Participants should explain the basis of their estimates rather than merely submit numbers.

5. Reconcile and challenge the forecast

The baseline and judgmental inputs are compared. Differences are investigated rather than averaged automatically. A change should be accepted only when the supporting information is relevant, sufficiently reliable, and not already reflected in the baseline.

6. Build consensus and approve the plan

Consensus does not mean that every participant receives the number they prefer. It means that the organization agrees on the most defensible forecast, its assumptions, its uncertainty, and the actions required. Disagreement should be recorded when it represents genuine uncertainty.

7. Communicate and execute

The approved forecast must be translated into inventory, capacity, purchasing, transport, and staffing decisions. Communication should include the forecast value, confidence range, key assumptions, risks, and trigger conditions for review.

8. Monitor performance and learn

Forecast accuracy is evaluated after actual demand becomes available. Errors should be analysed by product, location, horizon, method, and cause. The purpose is not to punish individuals for unavoidable uncertainty but to improve processes and identify systematic bias.

Managerial and Judgmental Forecasting Methods

Delphi Method

The Delphi method obtains forecasts from a panel of experts through several confidential rounds. In the first round, participants provide estimates and reasoning independently. A facilitator summarizes the responses without identifying individuals. Participants then reconsider their views in light of the group information and may revise their estimates. The process continues until the estimates converge sufficiently or the remaining disagreement is explained.

Delphi is valuable when historical data are limited, when the subject concerns a new technology or market, or when face-to-face discussion could be dominated by senior personalities. Anonymity reduces conformity and status effects. Its weaknesses include the time required, dependence on expert selection, and the possibility of artificial consensus. Delphi should therefore preserve minority views where they reflect plausible alternative futures.

Market Research and Customer Inquiry

Market research gathers information directly from customers or market participants through interviews, surveys, focus groups, experiments, concept tests, and analysis of stated purchase intentions. It is particularly useful for new products, service changes, and markets where historical demand is unavailable.

Stated intention is not identical to actual purchase behaviour. Respondents may express interest without purchasing, may react differently when prices are known, or may be unable to evaluate an unfamiliar product. Research design must consider sample selection, question wording, non-response, response bias, and the difference between hypothetical and actual choices. Where possible, stated preferences should be supplemented by pilot launches, test markets, or revealed-behaviour data.

Sales-Force Composite

In a sales-force composite, individual sales representatives estimate demand for their territories, accounts, or products. These estimates are aggregated to create a company forecast. Sales personnel possess valuable information about customer projects, competitive bids, and local conditions, especially in business-to-business markets.

However, sales representatives may understate demand to obtain easier targets or overstate it to secure inventory and demonstrate opportunity. Their estimates may also reflect only existing customers and may overlook market-wide trends. Controls include clear definitions, documented evidence, calibration against historical performance, independent review, and separation of the operational forecast from incentive targets.

Analogy Method

The analogy method forecasts a new product or market by comparing it with a similar product, customer segment, technology, or historical launch. The analogy may provide an initial demand curve, adoption pattern, seasonality profile, or substitution estimate.

A valid analogy requires more than superficial similarity. Analysts should compare price, use case, distribution, customer base, competitive intensity, availability, and marketing support. Differences should be adjusted explicitly. For example, a premium product distributed only through specialist outlets should not be forecast by copying the launch pattern of a mass-market substitute without adjustment.

Structured Judgment

Structured judgment converts expert knowledge into a transparent process. Rather than asking, “What is your forecast?” the facilitator asks participants to identify drivers, estimate a base case, state assumptions, provide an interval, assess confidence, and explain possible upside and downside conditions.

Useful techniques include independent initial estimates, evidence registers, premortem analysis, scenario construction, probability ranges, and written rationales. Independent estimates should be collected before group discussion because early numbers can anchor subsequent judgments. The final adjustment to a statistical baseline should be recorded separately from the baseline itself.

Judgmental Adjustment to a Statistical Baseline

Many mature supply chains begin with a statistical forecast and allow controlled adjustments. This hybrid approach is often superior to either mechanical forecasting or unrestricted intuition. The key principle is that judgment should add information, not merely restate an existing belief.

An adjustment is more defensible when it is based on a known event not represented in the data, such as a confirmed promotion, a new contract, a product discontinuation, or a documented competitor action. Adjustments based only on general optimism, pressure from management, or a desire to achieve a target should be challenged. Organizations should measure whether adjustments improve or reduce accuracy over time.

Bias, Power, and the Quality of Judgment

Forecasting bias is a systematic tendency for estimates to be too high, too low, too stable, or too reactive. Bias can arise from individual cognition and from organizational incentives.

Anchoring occurs when an initial number, previous forecast, or target exerts excessive influence. Confirmation bias leads people to select evidence supporting their preferred view. Availability bias gives disproportionate weight to vivid recent events. Overconfidence produces narrow prediction ranges and insufficient recognition of uncertainty. Recency bias causes the latest observation to dominate longer-term evidence. Groupthink suppresses disagreement, while optimism bias encourages favourable assumptions.

Organizational incentives can be equally important. Sales teams may inflate estimates to secure stock or deflate them to make targets easier. Senior managers may impose desired numbers. Finance may favour a conservative plan, while marketing may promote an ambitious scenario. When forecasts are linked directly to performance evaluation, participants may rationally distort their inputs.

Bias controls should be designed into the process. These include:

• separating the demand forecast from the sales target;
• collecting independent estimates before group discussion;
• requiring evidence and written assumptions;
• using ranges and scenarios rather than single-point certainty;
• comparing adjusted forecasts with unadjusted baselines;
• reviewing errors for persistent directional patterns;
• allowing anonymous input where hierarchy is influential; and
• assigning explicit responsibility for challenging assumptions.

The objective is not to eliminate judgment. It is to make judgment visible, contestable, and learnable.

Consensus Forecasting and Governance

Consensus forecasting integrates information from sales, marketing, operations, procurement, finance, and logistics. A typical process begins with a statistical baseline, followed by functional reviews, an exception meeting, and formal approval. The meeting should focus on material changes and unresolved risks rather than reviewing every item equally.

Good governance defines roles. The forecasting team prepares data and analysis; functional experts provide information; the supply-chain or business planning group coordinates alternatives; and an accountable executive approves the operating plan. Decision rights should specify who may alter the baseline, what evidence is required, and when an escalation is necessary.

Consensus can fail when meetings become political negotiations. A number that satisfies every department may be less accurate than a number that exposes genuine disagreement. A sound process records the base case, alternative scenarios, confidence level, major assumptions, and unresolved dissent. This record supports later learning and protects the organization from retrospective reinterpretation.

An Illustrative Case: Forecasting a New Cold-Chain Product

Background

Northstar Health Logistics, a fictional regional distributor, plans to introduce a temperature-sensitive diagnostic product. The product has no direct sales history. It will be supplied to hospitals and specialist clinics, and replenishment lead time is six weeks. Excess stock is costly because product shelf life is limited.

Situation and Problem

The commercial team expects rapid adoption and proposes a high first-year forecast. Operations is concerned that the proposed number reflects sales ambition rather than probable consumption. Procurement must reserve refrigerated capacity, while finance wants a defensible estimate for working-capital planning. The organization must prepare a monthly forecast for the first year and a capacity view for the following two years.

Analysis

A simple historical time-series model is impossible because no direct history exists. The team therefore identifies three information sources. First, market research estimates the number of eligible institutions and their likely adoption stages. Second, the sales force provides account-level assessments based on customer discussions, but each estimate is accompanied by evidence and a confidence rating. Third, the team identifies two analogous product launches, adjusting for the new product’s higher price, narrower clinical use, and stronger regulatory support.

The team constructs a base case, an upside case, and a downside case. It also identifies assumptions that could materially change the result: the date of regulatory approval, the number of trained clinicians, the supplier’s service level, and the rate at which hospitals substitute the new product for an existing one. The commercial target is kept separate from the operational demand forecast.

Approach and Solution

A facilitated Delphi process is used with representatives from clinical marketing, sales, procurement, logistics, and finance. Participants first submit independent estimates. The facilitator then presents the range, the underlying assumptions, and the evidence supporting major differences. After two rounds, the group agrees on a base case but retains a higher adoption scenario because two major hospital networks have not yet made final decisions.

The approved forecast includes monthly demand, a confidence interval, and trigger rules. If signed contracts exceed a specified threshold, the upside scenario will be activated. If training completion falls below plan, the downside scenario will guide inventory commitments. Initial procurement is staged rather than based entirely on the high scenario. The sales target is reported separately and is not used to determine safety stock.

Outcome

The process gives Northstar a transparent plan despite limited data. It reduces the risk of premature refrigerated-capacity commitments and establishes a mechanism for updating the forecast as customer commitments become observable. At the first quarterly review, the organization compares actual adoption with the assumptions rather than simply judging whether the original point estimate was correct.

Learning Points

The case illustrates that new-product forecasting is an exercise in structured uncertainty management. Analogies and expert judgment are useful, but only when assumptions are explicit. Scenario triggers allow the supply chain to respond to evidence. Separating targets from forecasts reduces incentive distortion, while staged commitments limit the cost of being wrong.

Limitations, Professional Responsibilities, and Digital Support

No forecasting method removes uncertainty. Delphi may reflect the composition of its panel; market research may not predict behaviour; sales composites may be politically influenced; analogies may be invalid; and statistical baselines may rely on distorted historical data. Forecasts should therefore be treated as conditional statements, not promises.

Professional responsibility requires honest communication of uncertainty. Managers should not suppress adverse information, present a target as an unbiased forecast, or use artificially precise numbers to create confidence. Assumptions affecting inventory, customer service, product safety, and financial commitments should be auditable. In regulated industries, records may need to demonstrate how demand assumptions affected procurement and service decisions.

Digital planning systems can support version control, collaborative assumption registers, forecast hierarchies, exception management, and automated accuracy measurement. Machine-learning tools may identify nonlinear relationships or external demand signals, but they do not eliminate the need for human review. Automated systems can reproduce historical bias, mistake stock-outs for weak demand, or generate opaque adjustments. Human judgment remains necessary when the environment changes, data are sparse, or the decision involves strategic commitments.

Emerging practice is moving toward probabilistic forecasting, scenario-based planning, demand sensing, and continuous monitoring of external signals. These developments are valuable when they improve decisions rather than merely increase technical complexity. The appropriate test is whether the process produces better, more explainable, and more actionable decisions.

Best-Practice Guidelines

A mature forecasting process should:

1. Begin with the decision, not the data or software.

2. Define demand, granularity, horizon, frequency, and ownership clearly.

3. Use a transparent statistical or operational baseline where possible.

4. Separate forecasts, budgets, targets, and scenarios.

5. Require evidence for judgmental adjustments.

6. Collect independent views before consensus discussion.

7. Record assumptions, confidence, alternatives, and dissent.

8. Reconcile forecasts across product, location, and organizational levels.

9. Evaluate both accuracy and directional bias.

10. Review the process when market conditions, products, or incentives change.

Chapter Summary

Forecasting in supply-chain management is a structured decision process rather than a single numerical exercise. Its quality depends on clear objectives, appropriate granularity, suitable horizons, explicit assumptions, and defined ownership. Statistical forecasts provide a consistent baseline, while managerial methods such as Delphi, market research, sales-force composites, analogies, and structured judgment contribute information about conditions not captured in historical data.

Judgment is most valuable when it is independent, evidence-based, documented, and separated from commercial targets. Bias may arise from anchoring, optimism, groupthink, incentives, hierarchy, and confusion between observed sales and underlying demand. Consensus forecasting should integrate functions without allowing political compromise to replace analysis. Scenario planning, confidence ranges, trigger conditions, and post-forecast evaluation help organizations manage uncertainty. Digital tools can improve collaboration and monitoring, but they require data governance and human oversight.

Key Terms

Forecasting process: The organized sequence through which demand information is defined, estimated, reviewed, approved, and evaluated.

Demand forecast: An estimate of future customer requirements under stated conditions.

Forecast baseline: The initial forecast produced from historical data, operational rules, or a standard model.

Granularity: The level of detail at which a forecast is prepared.

Forecast horizon: The future period covered by a forecast.

Rolling forecast: A forecast updated periodically as new information becomes available.

Consensus forecast: A forecast agreed through cross-functional review and reconciliation.

Delphi method: A multi-round, usually anonymous expert-elicitation process.

Market research: Systematic collection of information from customers or market participants.

Sales-force composite: An aggregate forecast developed from territory or account estimates by sales personnel.

Analogy method: Forecasting by adapting the observed pattern of a comparable product or market.

Structured judgment: A documented approach for eliciting, testing, and combining expert assessment.

Judgmental adjustment: A deliberate change to a baseline forecast based on additional information.

Forecast bias: A systematic tendency for forecasts to deviate in a particular direction.

Anchoring: Excessive influence of an initial number or reference point on later judgment.

Scenario: A coherent representation of possible future conditions and their implications.

Assumption register: A record of forecast assumptions, sources, owners, timing, and confidence.

Demand sensing: Use of recent operational and external signals to identify short-term demand changes.

Forecast reconciliation: Adjustment of forecasts so that estimates at different organizational levels are coherent.

Forecast governance: Roles, rules, controls, and decision rights governing the forecasting process.

Review Questions

Short Answer Questions

1. Why should a forecast be linked to a specific decision?

2. Distinguish between a sales target and an operational demand forecast.

3. What is the purpose of an assumption register?

4. State two advantages and two risks of the Delphi method.

5. What is meant by a judgmental adjustment to a statistical baseline?

Long Answer Questions

1. Explain how granularity and forecast horizon should be selected for logistics decisions.

2. Describe the stages of a disciplined forecasting workflow.

3. Compare Delphi, market research, sales-force composite, and analogy methods.

4. Discuss the cognitive and organizational sources of forecasting bias and the controls used to reduce them.

5. Explain how consensus forecasting can be governed without allowing political compromise to replace analysis.

Analytical / Application Questions

1. A distributor’s sales history shows low demand for a product, but the product was unavailable for several weeks. Design a process for estimating underlying demand and explain the information you would require.

2. A sales department increases its forecast by 30 percent without providing evidence, arguing that the higher number will help secure inventory. How should the planning team evaluate and govern this adjustment?

3. A company is launching a product with no historical data and limited market research. Develop a forecasting approach using at least three judgmental methods, including assumptions, scenarios, and review triggers.

22. Multiple Choice Questions with Answers

Question 1

Which action best establishes a sound starting point for a judgmental forecasting process?

A. Adopt the sales target as the forecast

B. Prepare a transparent baseline and identify relevant assumptions

C. Ask the most senior manager for a single estimate

D. Average all departmental estimates without investigation

Correct Answer: B. Prepare a transparent baseline and identify relevant assumptions

Question 2

Why should observed sales sometimes be distinguished from underlying demand?

A. Sales data are never useful for forecasting

B. Observed sales may be constrained by stock-outs or supply failures

C. Underlying demand is always higher than sales

D. Demand can only be measured through surveys

Correct Answer: B. Observed sales may be constrained by stock-outs or supply failures

Question 3

Which feature most clearly distinguishes the Delphi method?

A. It relies only on historical shipments

B. It uses anonymous, repeated rounds of expert assessment

C. It requires sales representatives to set targets

D. It copies the demand pattern of a similar product

Correct Answer: B. It uses anonymous, repeated rounds of expert assessment

Question 4

A sales representative reports a large increase in expected demand but provides no customer evidence. What is the most appropriate response?

A. Accept the increase because local sales knowledge is always superior

B. Reject all sales-force input permanently

C. Request evidence, document the adjustment, and compare it with the baseline

D. Replace the forecast with the finance budget

Correct Answer: C. Request evidence, document the adjustment, and compare it with the baseline

Question 5

Which situation is most suitable for the analogy method?

A. A mature product with extensive stable history

B. A new product with no direct historical demand

C. A forecast requiring no assumptions

D. A situation in which comparable products do not exist

Correct Answer: B. A new product with no direct historical demand

Question 6

What is the main purpose of separating a forecast from a sales target?

A. To prevent the use of any managerial judgment

B. To ensure that an ambitious goal is not mistaken for probable demand

C. To make forecasts more optimistic

D. To eliminate cross-functional planning

Correct Answer: B. To ensure that an ambitious goal is not mistaken for probable demand

Question 7

Which practice best reduces anchoring during structured judgment?

A. Present one executive estimate first

B. Collect independent estimates before group discussion

C. Require participants to agree immediately

D. Use only the previous forecast

Correct Answer: B. Collect independent estimates before group discussion

Question 8

In a new-product launch, what is the principal value of scenario trigger rules?

A. They guarantee that the base forecast will be accurate

B. They allow commitments to change when observable evidence changes

C. They eliminate the need for market research

D. They convert a target into a forecast

Correct Answer: B. They allow commitments to change when observable evidence changes

Question 9

Which statement about consensus forecasting is most accurate?

A. Consensus requires every department to receive its preferred number

B. Consensus should conceal disagreement to support implementation

C. Consensus integrates evidence while documenting assumptions and unresolved uncertainty

D. Consensus makes statistical baselines unnecessary

Correct Answer: C. Consensus integrates evidence while documenting assumptions and unresolved uncertainty

Question 10

What is a major governance risk when an automated forecasting system is used without review?

A. It may reproduce historical bias or interpret stock-outs as weak demand

B. It will always produce forecasts that are too high

C. It prevents the storage of historical information

D. It removes all need for forecast evaluation

Correct Answer: A. It may reproduce historical bias or interpret stock-outs as weak demand