Judgmental forecasting
Definition
Judgmental forecasting
Judgmental forecasting predicts outcomes from structured expert opinion, not from statistical models. You reach for it when the data will not carry you: a new market, a first product, or a shock that makes every past number useless to the decision.
It is one half of forecasting as a discipline. The other half runs on regression and time series work, and judgment takes over where the model refuses to run.
Hyndman and Athanasopoulos put it plainly in Forecasting: Principles and Practice.
Judgmental forecasting is often “the only option, such as when there is a complete lack of historical data, or when a new product is being launched.”
That makes it the working tool for pharmaceutical launches, geopolitical planning, and complex operations where a spreadsheet alone would mislead the decision.
Key takeaways
- Judgmental forecasting relies on expert opinion when hard data is missing, dated, or unusable.
- Methods include Delphi rounds, forecasting by analogy, scenario building, Cooke’s method, and surveys.
- Anonymity, checklists, and documented assumptions protect the estimate from bias and drift.
- It suits new product launches, long-horizon strategy, and shocks that break historical patterns.
- Pair it with statistical forecasting whenever any usable numeric history exists.
How it works
Judgmental forecasting collects expert input under controlled conditions, corrects for known biases, and turns opinion into a documented estimate. The steps come from decision analysis, not pure statistics, so the audit trail matters as much as the number.
Analysts start by framing the question, then pick a method that matches the horizon and the data gap. A short-cycle demand call may need only an analogy comparison against a similar past product.
A ten-year strategy call usually needs scenario planning or a full Delphi method round. Panels typically run between five and twenty experts, and most converge after two or three rounds.
Panels stay anonymous so no single voice dominates. Facilitators aggregate the responses through medians, weighted averages, or Cooke’s calibration scoring — the choice depends on how well panellists scored on earlier seeded questions.
The same textbook is blunt about the risk. Judgmental forecasts “are subjective, and therefore do not come free of bias or limitations,” and anchoring pulls later estimates back toward the reference point the panel saw first.
Two of its principles do most of the protective work. Accuracy improves under a systematic approach involving checklists, and forecasters and users should be clearly segregated — otherwise targets bleed into forecasts and political agendas cloud the judgment.
The main methods sit in one table.
| Method | Best for | Inputs | Output |
|---|---|---|---|
| Forecasting by analogy | A new product close to a past one | Performance of the comparable case | Range prediction |
| Delphi method | Long-horizon, high-uncertainty calls | Anonymous expert rounds | Converged consensus |
| Scenario building | Strategy under deep uncertainty | Plausible future “worlds” | 3–5 narrative scenarios |
| Cooke’s method | Risk calls needing calibration | Answers to seeded questions | Weighted forecast |
| Structured surveys | Customer demand, sentiment, intent | Polling of buyers or staff | Probability estimate |
| Executive opinion | Launches with no sales history | A panel of senior managers | Single agreed figure |
| Sales force composite | Bottom-up demand from the field | Territory estimates rolled up | Aggregated demand curve |
| Composite | Problems needing more than one lens | Two or more methods combined | Blended estimate |
Scoring the result afterwards separates method from guessing. Teams that log each estimate against the outcome build a forecast accuracy rate for the panel, which is the input Cooke’s method needs next time.
In contact centres the same discipline shows up inside call center forecasting, where judgment covers the recalls, outages, and campaign spikes no arrival curve predicts.
Examples
Judgmental forecasting earns its keep when history is silent or misleading. Three documented cases show experts committing real money ahead of the data, and the last of them shows what happens when the judgment misses.
Royal Dutch Shell scenario planning (1967–1973). Pierre Wack pioneered the technique at Royal Dutch Shell, starting in 1967 and pushing oil-price scenarios through the business during 1972 and early 1973.
When the OPEC embargo ran from 17 October 1973 to March 1974, Shell had already rehearsed its response. The firm climbed from the seventh-largest oil major to the second by the late 1970s.
The Delphi method was developed at RAND Corporation in the same era, for the same class of long-horizon question.
Pfizer and BioNTech COVID-19 vaccine (2020). Both firms built manufacturing capacity months before Phase 3 data existed. No base rate existed to regress against, so the call rested on judgment about severity, spread, and public health need.
Albert Bourla, Pfizer’s chief executive, wrote that the company “has been investing at risk since the early days of the pandemic to perfect our manufacturing processes and rapidly build up capacity.” Emergency use authorisation followed on 11 December 2020.
Apple iPhone 12 mini (2021). A new form factor has no history to regress, so launch volumes rested on judgment. On 9 March 2021 Nikkei Asia reported that demand for the mini ran far lower than Apple had expected.
That is the honest case for the method — a documented estimate that misses still tells you which assumption broke, which a hidden model rarely does.
Related terms
Judgmental forecasting sits inside a family of estimation and planning terms. The cluster runs from the raw data layer through the forecast itself and out to the staffing decisions it drives. The boundary is whether the estimate starts from data or from people.
- Forecasting: the parent discipline covering both quantitative and qualitative methods.
- Predictive Analytics: statistical and machine learning methods used where a strong data trail exists.
- Prescriptive Analytics: the layer that recommends an action once the forecast is settled.
- Business Intelligence: the reporting layer that surfaces the signals feeding a forecast.
- Workforce Planning: headcount forecasting where judgment covers attrition and growth surprises.
- Workforce Management (WFM): the scheduling discipline that turns a demand forecast into published shifts.
- Capacity Planning: matching supply against a forecast of expected demand.
FAQ
Common questions on judgmental forecasting cover method choice, bias control, whether the output is solid enough for a board, and how the approach sits alongside statistical forecasting and the AI tools that now handle the data-rich half of the problem.
How is judgmental forecasting different from statistical forecasting?
Statistical forecasting projects the numeric past forward. Judgmental forecasting brings in expert opinion when that past is thin, broken, or irrelevant. Most mature teams run both and blend the outputs.
When should you use judgmental forecasting?
Reach for it during product launches, geopolitical shocks, regulatory shifts, and any decision carrying under three years of usable history. It also works as a sanity check on a purely statistical model, which is how Shell used it before 1973.
What are the main biases to control?
Anchoring, groupthink, overconfidence, and recency bias distort every panel. Hyndman and Athanasopoulos single out anchoring, where later estimates drift back toward the first reference point the panel saw. Anonymous rounds and checklists keep the estimate auditable.
Is judgmental forecasting reliable enough for board decisions?
Yes, when method selection is documented, assumptions are written down, and the estimate is checked against whatever data does exist. Boards usually want a scenario band rather than a single point — plus a record of who sat on the panel.
How does it fit alongside AI and predictive analytics?
AI handles the pattern-heavy problems with a long clean data trail, while judgmental forecasting covers the shocks, new categories, and strategy calls that still defeat it.
Explore more outsourcing terms and operator guidance at Outsource Accelerator.
Independent
