Forecasting

Forecasting is the practice of predicting future events, conditions or outcomes by analyzing historical data, identifying patterns and applying analytical methods ^[1]. Organizations use forecasting to anticipate customer demand, project sales, estimate resource requirements and plan for various contingencies. The discipline combines quantitative techniques grounded in statistics and mathematics with qualitative approaches drawing on expert judgment and market intelligence. Accurate forecasting supports better decision making across functions including production, inventory, finance, marketing and human resources.

Purpose and importance

Businesses operate under conditions of uncertainty about future events. Customer demand fluctuates unpredictably. Economic conditions shift. Competitors take unexpected actions. Technology evolves in unanticipated directions. Forecasting attempts to reduce this uncertainty by providing informed estimates of what lies ahead ^[2].

The value of forecasting lies not in achieving perfect accuracy, which is impossible, but in improving the quality of decisions made under uncertainty. Even imperfect forecasts enable better resource allocation, inventory management, capacity planning and financial planning than operating without any forward view. Organizations that forecast effectively gain competitive advantages over those that react only after events occur.

Forecasting influences decisions across multiple time horizons. Short-term forecasts of days or weeks guide production scheduling and workforce deployment. Medium-term forecasts spanning months inform procurement, inventory positioning and promotional planning. Long-term forecasts extending years support capital investment, facility expansion and strategic planning ^[3].

Quantitative methods

Quantitative forecasting relies on historical numerical data and mathematical models to project future values. These methods work best when substantial historical data exists and when future patterns are expected to resemble past patterns ^[4].

Time series analysis

Time series methods examine sequences of observations recorded over time, identifying patterns that can be projected forward. The simplest approach assumes future values will equal past averages. More sophisticated techniques decompose historical data into trend, seasonal and cyclical components, then recombine these elements to generate forecasts ^[5].

Moving average methods calculate averages of recent observations, updating as new data becomes available. Simple moving averages weight all observations equally while weighted moving averages assign greater importance to more recent data. Exponential smoothing applies continuously declining weights to older observations, emphasizing recent experience while incorporating historical information.

Trend projection

When data exhibits consistent upward or downward movement over time, trend projection extends this pattern into the future. Linear trend models fit straight lines to historical data using regression techniques. Nonlinear approaches accommodate growth patterns that accelerate, decelerate or follow other curved trajectories ^[6].

Trend projection assumes underlying forces driving historical patterns will continue operating similarly. This assumption may fail when fundamental changes occur in markets, technology, competition or customer preferences.

Seasonal adjustment

Many business activities exhibit systematic variation across seasons, months, days of the week or hours of the day. Retail sales surge during holiday periods. Utility demand peaks in summer or winter. Restaurant traffic varies predictably through the week ^[7].

Seasonal adjustment techniques identify and quantify these recurring patterns, enabling forecasters to project seasonal fluctuations and adjust baseline forecasts accordingly. Seasonal indices express typical variation from average levels, applied to base forecasts to generate period-specific predictions.

Causal models

Unlike time series methods that project patterns without explaining them, causal models identify relationships between the variable being forecast and explanatory factors believed to influence it ^[8]. Regression analysis estimates the strength and direction of relationships between dependent and independent variables.

Economic models forecast aggregate measures such as gross domestic product, inflation or unemployment based on theoretical relationships among economic variables. Econometric approaches combine economic theory with statistical estimation techniques to generate forecasts. Input-output models trace how changes in one sector ripple through interconnected economic systems.

Qualitative methods

Qualitative forecasting draws on human judgment, expertise and opinion rather than numerical analysis. These approaches prove valuable when historical data is limited, when forecasting genuinely novel situations, or when significant discontinuities render historical patterns unreliable guides ^[9].

Expert opinion

Subject matter experts possess knowledge, experience and intuition that can generate useful forecasts even without formal analytical methods. Individual executives may provide forecasts based on their understanding of markets, customers and competitive dynamics. Panels of experts can combine multiple perspectives ^[10].

Unstructured expert opinion carries risks of bias, overconfidence and inconsistency. Experts may anchor on recent salient events, underweight base rates, or confuse confidence with accuracy. Structured approaches that decompose problems and aggregate multiple opinions mitigate some of these limitations.

Delphi method

The Delphi method structures expert opinion through iterative anonymous questionnaires. Experts independently provide forecasts with supporting rationale. Results are compiled and shared without identifying individual respondents. Subsequent rounds allow experts to revise their views considering others' perspectives ^[11]. The process continues until convergence or until remaining disagreements are clearly understood.

Anonymity prevents dominant personalities from swaying group judgment. Iteration allows learning and perspective taking. Written rationale forces explicit consideration of reasoning rather than mere opinion assertion.

Market research

Surveys of customers, potential customers or industry participants can reveal intentions, preferences and expectations that inform forecasts. Purchase intent surveys ask consumers about planned buying behavior. Concept tests gauge reactions to potential new products ^[12]. Trade channel partners may provide insights about emerging trends.

Market research faces challenges in translating stated intentions into actual behavior. Respondents may not accurately predict their own future actions. Survey samples may not represent broader populations. Questionnaire design can inadvertently bias responses.

Scenario planning

Rather than forecasting single expected outcomes, scenario planning develops multiple plausible futures representing different ways uncertain factors might unfold ^[13]. Organizations consider how they would respond under each scenario, building flexibility and resilience rather than betting on single predictions.

Scenarios typically span a range from optimistic to pessimistic, with intermediate cases reflecting most likely developments. The exercise value lies in expanding thinking beyond narrow expectations and preparing contingency responses.

Combining methods

Research consistently finds that combining forecasts from multiple methods outperforms reliance on any single approach. Simple averaging of different forecasts often proves remarkably effective despite its lack of sophistication ^[14]. Weighted combinations assign greater influence to methods with better track records.

Quantitative and qualitative methods can complement each other. Statistical analysis provides disciplined baseline projections while judgment adjusts for factors the models cannot capture. Integration of approaches leverages their respective strengths while compensating for limitations.

Forecast accuracy

No forecast achieves perfect accuracy, but forecasting value depends on how close predictions come to actual outcomes. Organizations should track forecast performance over time, measuring errors and identifying systematic biases that can be corrected ^[15].

Common accuracy metrics include mean absolute deviation, mean absolute percentage error and root mean squared error. Each emphasizes different aspects of forecast performance. Tracking multiple metrics provides comprehensive assessment.

Accuracy typically deteriorates as forecast horizons extend further into the future. Near-term predictions benefit from more relevant data and shorter periods for disruption. Long-range forecasts face compounding uncertainty as more intervening events can alter outcomes.

Forecasting — recommended articles

Market research — Decision making — Financial planning — Strategic planning — Risk management — Investment

References

• Armstrong J.S. (2001), Principles of Forecasting: A Handbook for Researchers and Practitioners, Springer.
• Hanke J.E., Wichern D.W. (2014), Business Forecasting, 9th Edition, Pearson.
• Heizer J., Render B., Munson C. (2017), Operations Management, 12th Edition, Pearson.
• Hyndman R.J., Athanasopoulos G. (2021), Forecasting: Principles and Practice, 3rd Edition, OTexts.
• Makridakis S., Wheelwright S.C., Hyndman R.J. (1998), Forecasting: Methods and Applications, 3rd Edition, Wiley.
• Ord K., Fildes R., Kourentzes N. (2017), Principles of Business Forecasting, 2nd Edition, Wessex Press.
• Stevenson W.J. (2018), Operations Management, 13th Edition, McGraw-Hill.

Footnotes

1. Makridakis S., Wheelwright S.C., Hyndman R.J. (1998), pp. 1-15
2. Stevenson W.J. (2018), pp. 78-92
3. Heizer J., Render B., Munson C. (2017), pp. 134-148
4. Hanke J.E., Wichern D.W. (2014), pp. 45-62
5. Hyndman R.J., Athanasopoulos G. (2021), Chapter 3
6. Ord K., Fildes R., Kourentzes N. (2017), pp. 156-178
7. Makridakis S., Wheelwright S.C., Hyndman R.J. (1998), pp. 234-256
8. Hanke J.E., Wichern D.W. (2014), pp. 312-340
9. Armstrong J.S. (2001), pp. 89-112
10. Stevenson W.J. (2018), pp. 102-118
11. Armstrong J.S. (2001), pp. 125-145
12. Heizer J., Render B., Munson C. (2017), pp. 162-178
13. Ord K., Fildes R., Kourentzes N. (2017), pp. 234-256
14. Armstrong J.S. (2001), pp. 267-285
15. Hyndman R.J., Athanasopoulos G. (2021), Chapter 5

Author: Sławomir Wawak