December 1996 € Volume 6 € Number 12

The Art and Science of Forecasting

How to achieve excellence in demand planning

By Hans Levenbach, Ph.D.

Some people call demand forecasting an art. Others consider it a science. Most call it impossible (although I do not adhere to this pessimistic point of view).

Demand forecasters are in the business of making statements about future demand for products and services in the face of uncertainty. A forecast is not just a number, outcome or task. It is part of an ongoing process, integrating excellence into sales, marketing, inventory, production and all other aspects of the supply chain. A sound forecasting process incorporates both art and science into a logical and coherent series of steps that if conducted in an organized, management-supported fashion, it can improve forecasting effectiveness, reliability and accuracy in your organization.


The art and science of business forecasting
The art of forecasting has to do with the relational aspects of forecasting: How well do you present the value of the forecast to its users? Weather forecasters are very good at this. But on a more practical front, many business executives, financial investors and government officials hope to find answers to questions such as:

• What will the information superhighway look like?
• If the electric vehicle becomes practical, how will things change?

A review of many important innovations, from the steam engine to the laser, shows that we can seldom predict the full technological, economic and social impact of new products and services. Scientists at AT&T's Bell Laboratories, which has recently become part of Lucent Technologies, invented the laser more than 30 years ago. Management initially hesitated to apply for a patent on it. Even IBM saw no large potential market for the computer in 1949 and predicted the eventual world market for computers at about 10 to 15. Today, the installed base worldwide in computers is in the hundreds of millions. The list goes on and on and can be very amusing.

The science of forecasting deals with the strategic issue, which has to do with how you prove the value of your forecast and models to management. In demand forecasting, there is still a big gap between what economics and statistics can offer us and the judgmental aspects of business decision-making required to develop accurate forecasts. Looking to the future we need new ways to close the gap.


The role of a business forecaster
At one time, demand forecasting was equated with economic or macro forecasting. In recent years economic forecasting has suffered from a lack of credibility, media ridicule and shortcomings in accuracy goals. Quotes from business journals and newspaper articles frequently bore into the quality of the ability of economic forecasters at predicting recessions. To quote from a recent issue of Fortune magazine: "The biggest problem with economic forecasters is that they generally can't tell us what we most want to know."

Nowadays, the meaning of business forecasting has broadened considerably and includes "micro forecasting," or "forecasting of the firm." This deals with more detailed elements than demand in classical economics. Expressions of such demands are shipments or sales of goods to supply warehouses, distributors, food brokers, channels and end customers. The need for demand-driven forecasting of the right amount of the right product in the right place and the right time is one of the key underpinnings of what is now known as supply chain forecasting.

Not everyone, however, has fully embraced this concept. While demand-based forecasting is the hallmark of grocery and discount retailers, many retailers in drugs, apparel and durables still do not trust their suppliers to deliver products without intervention.

How do we deal with this lack of credibility and turn a problem into an opportunity for excellence in forecasting? I am going to make a modest attempt to offer what I believe are the top 10 ingredients you need to have in place to succeed with a demand forecasting program in today's business world. This list is neither exhaustive nor exclusive, but it does include the essential elements of a successful forecasting process.


Management initiatives

1. Commit management to change and chance

Management tends to avoid dealing with forecasting issues because of its aversion to dealing with change and uncertainty. Forecasting is all about change and chance. Like many human beings, managers don't want to bother with forecasts, because when forecasts are OK, you don't need to hear from forecasters. But, when they are off, forecasters are wrong. Nobody wants to be wrong.

2. Accept forecasting as an evolutionary process

Forecasting seems to become a high priority with management at times of crises and when unexpected changes occur in the business environment. Forecasting must be accepted as an evolutionary process responsive to change. It does not deal with status quo situations.

3. Enhance quantitative skills in management

Managers can absorb and use tools that work for them. Most forecasting techniques, no matter how sophisticated, are only as good as the underlying data. With more and better data, the value of quantitative methods will become more evident to managers. By giving management access to internal and external data and the tools they need, you will have won as a forecaster.

4. Institute ongoing training

Our idea of what a forecaster needs to know today may be obsolete in the very near future. Companies need to maintain a training program in forecasting methods and best practices. Just as the management process is subject to change, so is forecasting.

5. Achieve consensus on assumptions and rationale

While sitting in forecast review meetings with marketing and production managers, forecasters must strive towards a consensus on assumptions and rationale, rather than a confrontation. Forecasters have techniques and methodologies for laying out these future scenarios based on assumptions about the economy, demographics, industry and market forces. Debate assumptions, not the numbers!

Business objectives

6. Forecast as advice

Sell a forecast as advice and seek management approval. The forecaster should be useful and provide information that is relevant to the decision maker. One does this by communicating with all user groups, including purchasing, finance, production and marketing. By providing advice that is useful, forecasters will not work themselves out of a job. Finally, in securing management approval, forecasters should not wow them with technical know-how, sophisticated modeling output or even excessive wit.

7. Accuracy as an attainable goal

Use forecast accuracy as an attainable goal, not as a fiat. Managers deal with changing market realities and uncertainties. Confronted with competitive pressures, management must view the future subject to a broad range of dissimilar influences. It is not certainty that forecasters want, it is an understanding of alternatives and possibilities.

8. Accumulate intuitive experiences

Forecasters must learn how to better incorporate their intuition with the formal, mechanical forecasting tools available on the computer. Forecasters often know what is going on by gut feeling, but can rarely formalize this experience in quantifiable terms. Usually, we tend to interact with mathematical and statistical models through software to build our forecasting models. We then take quantitative outputs and judgmentally (usually external to the computer) incorporate adjustments to formulate our forecasting advice for the decision maker. Expert systems are becoming more commonplace. What this will do for forecasting is to more closely link management judgment and quantitative modeling as a unified process.

9. Empirically validate statistical methods

Modern, computer-based techniques (data warehousing) now let us effectively analyze large amounts of data, much of this through effective graphical means, in less time than required before and with increased insight. The computer has made it feasible to warehouse relevant data and perform complex analytical calculations in a very short time. Better forecasting performance with these methodologies may be constrained by the quality, consistency and availability of relevant data. While accuracy of forecasts prepared by individual forecasters and years of experience may be related, an individual's judgment may be more important to the success of forecasting than reliance on increasingly sophisticated modeling efforts. As the well-known statistician George P. Box once noted, "All models are wrong, but some are useful."

10. Don't bet on the numbers

To remind you of some other statements about forecasts as numbers:

• Business Week in 1958 stated that "With over 50 foreign cars already on sale here, the Japanese auto industry isn't likely to carve out a big slice of the United States market for itself."
• Thirty years ago, Time magazine publisher Henry Luce was quoted as saying: "By 1980 all power (electric, atomic, solar) is likely to be costless."
• Only 20 years ago, a well known movie mogul said that television won't be able to hold on to any market it captures after the first six months. People will soon get tired of staring at a plywood box every night.

In business applications, demand forecasters must focus on the process, user and management needs, as well as their own needs, to do the best job possible. Arguing about numbers will be counterproductive!

I have presented 10 vital steps for achieving excellence in demand forecasting for business applications. Each step on its own will not blaze a trail to the top. But each is critical to the success of any forecasting endeavor designed to lower costs and improve customer satisfaction in the supply chain. Make forecasting crucial to your organization's goals and you will end up with the right quantity of the right product on the right truck.