Statistical approaches to forecasting can provide a framework for creating rolling budgets to which analytical skills and judgment can be applied in supporting a sound budgeting process. Users of statistical forecasting models have come to realize that their models can only be relied upon to provide a first approximation — a set of consistent forecasts which then must be ‘massaged’ with intuition and good judgment to take into account those influences on economic and business activity for which history is a poor guide.
    
In rolling budgeting applications, it may become apparent from creating rolling simulations that the actuals have exceeded the estimates for several successive periods, or that the forecasts for a given period underpredict the held-out actual value. Experience suggests that a model’s projections should be modified (adjusted upward or downward) by a given amount or percentage to account for the current deviation (bias) and the forecaster’s expectation of whether that forecast profile pattern will prevail.

Subjective judgment in forecasting practices should be based on all available information, including changes in company policy, changes in economic conditions and government regulations, and contacts with customers. Such judgment is a real measure of the skill and experience of the budget forecaster. Consequently, data and forecasting processes are only as good as the person interpreting them. To paraphrase world-renowned statistician George E. P. Box, I suggest that All DATA are Wrong, Some are Useful. This judgment operates on many inputs to reach a final forecast. 

Informed judgment plays a critical role in the determination of the final forecast numbers and, later on, in the determination of when a forecast should be revised.

Informed judgment is, by far, the most crucial element when we are trying to predict the future. Informed judgment is what ties the forecasting process and the extrapolative techniques into a cohesive effort that is capable of producing realistic predictions of future events or conditions. Informed judgment is an essential ingredient of the selection of the forecasting approach; the selection of data sources; the selection of the data collection and data cleaning methodologies; the selection of preliminary data analysis and extrapolative techniques; the use of exception-handling and root-cause analysis techniques during the forecasting process; the identification of forward-looking market and company factors that are likely to affect the future of the item to be forecast; the determination of how those factors will affect the item in terms of the direction, magnitude (amount or rate), timing, and duration of the expected impact; and the selection of the forecast presentation methodology.

Informed judgment plays a significant role in taming the uncertainty associated with all forecasting applications.

Automatic processes, models, and statistical algorithms are now increasingly used in computing future demand from a set of key factors. However, no such approach should reduce substantially the reliance upon sound judgment. Judgment must be based on a comprehensive analysis of market activities and a thorough evaluation of basic assumptions and influencing factors.

The limitations of a purely statistical modeling approach should be kept clearly in mind. Statistics, like all tools, may be valuable for one job but of little use for another.  An exploratory analysis of data patterns is basic to demand forecasting and a number of different statistical procedures may be employed to make this analysis more meaningful. However, the human element is required to understand the differences between what was expected in the past and what actually occurred and to predict the likely outcome of future events.

Dr. Hans is the author of a new book on business forecasting for the professional development and application in global demand/supply chain planning organizations.