WHY Demand Managers Need to Manage a Demand Forecasting Process That is Independent of Plans and Targets
The demand forecasting process is a complex process. Complex processes require checklists for effective execution. Here is a comprehensive Demand Forecasting Principles Checklist I found on the web oriented for forecasters and planners of global health products. I think it applies more broadly to any industry. You can clearly see how this applies in general:
Checklist of the Ten Principles
- Have I identified the principal customers/decision-makers of the forecast and do I clearly understand their needs?
- Have I understood and clearly communicated the purpose of the forecast and the decisions that will be affected by the forecast?
- Have I created a forecasting process that is independent of plans and targets?
- Have I understood the political considerations and taken measures to protect the process from political interference? Is my process transparent?
- Have I understood the broader environment in which the forecasting process is occurring?
- Have I created the forecast in the context of market and policy trends, portfolio of investments, and new product developments by suppliers? Have I clearly communicated this context?
- Have I created a dynamic forecasting process that incorporates and will reflect changes in the market and in public policy as they occur?
- Have I selected the methods that are most appropriate for the forecast problem and data available? Do I understand how to apply the various methods that are most suitable? Have I obtained decision-makers’ agreement on the methods?
- Does my methodology reflect the appropriate level of accuracy and detail that is needed for the forecast? Have I explicitly identified confidence intervals in the forecast?
- Have I made my forecast assumptions clear and explicitly defined them for those who will use the forecast?
- Do I understand the data and their limitations? Have I searched for data from multiple sources and gathered both qualitative and quantitative data? Am I using these different types of data appropriately?
Problem definition is probably the most critical phase of any forecasting project. It is necessary in this preparatory stage to define what is to be done, design a data framework and establish the criteria for successful completion of the project or forecast. The four PEER stages of the forecasting process are independent of the item to be forecast and the input parameters. It is essential to agree on the required outputs, time, and resources to be devoted to solving a problem, the resources that will be made available, the time when an answer is required, and, in view of these, the level of accuracy that may be achievable.
Exploratory data analysis and forecast model building should only begin after these kinds of agreements have been reached. If the prospects for reasonably accurate, credible and defensible forecasts are good, the demand forecaster proceeds to the next step of the process.
In a previous post I talked about the distinction between methods and models, as they are not the same! When we use quantitative models, heuristic methods and qualitative approaches for demand forecasting, we should recognize that credible and successful approaches need to be assimilated within an efficient and effective process. In my book, I have incorporated these principles into a four-step PEER process discipline.
Stage 1. Preparing Data
Avoid surprise. Involve management. Management tends to avoid dealing with quantitative forecasting issues because of an aversion to change (numbers) and chance (uncertainty).
In dealing with uncertainty in the external environment, in the instability of organizational structure, or in changing user needs, demand forecasters must remain in constant dialogue with their end-users.
Understand the big picture. Understand that forecasts can be wrong. Forecasting tends to become a high priority with management only during times of crises and unexpected changes in the business environment (beware of Black Swan events – A Black Swan event is an event in human history that was unprecedented and unexpected at the point in time it occurred. However, after evaluating the surrounding context, domain experts (and in some cases even laymen) can usually conclude: “it was bound to happen”. Dealing with changing market realities and uncertainties and confronted by competitive pressures, management should view the future as subject to a broad range of dissimilar influences. Demand Forecasting must be accepted as an evolutionary process responsive to and molded by change. It does not deal with static, status quo situations.
Stage 2. Executing Models
Practice modeling. Incorporate intuition. Models help management learn quickly from the past without relying exclusively on trial-and-error approaches. Forecasters must learn how to better incorporate their intuition into the formal, quantitative forecasting tools. Usually, we tend to interact with mathematical and statistical models through software to build our forecasting models. As expert systems become more commonplace, the need for intuitive insight may link judgment and quantitative modeling more closely as a unified process.
Run the numbers. Don’t bet on them! When asked to predict the thickness of an ordinary sheet of paper when folded 40 times, most of us will miss the answer by orders of magnitude. Most will offer an answer between 3 inches and 10 feet. The actual answer is of the order of the distance between the earth and moon! In business applications, arguing about expectations can be counterproductive. For example, 40 years ago Time Magazine Publisher Henry Luce was quoted as saying: “By 1980 all power (electric, atomic, solar) is likely to be costless.”
Remember George Box, an eminent statistician, who is quoted with this advice: “All models are wrong, some are useful.” Validate quantitative methods empirically. Models allow for effective tracking of forward looking performance against expectations. Some of the methods are elegant; others can be misleading. Empirical evidence suggests that for industrial applications, the theory is much overrated and simpler methods can do as well, if not better at times. With modern computing power, we can (and must) effectively analyze large amounts of data, much of this through efficient graphical means in shorter time frames than before. The computer has made it feasible to warehouse lots of structured and unstructured data (Big Data) and perform complex analytical calculations (Predictive Analytics) in a flash. The availability of relevant data, simple paradigms, predictive visualization and the experience of individual forecasters may be more balanced that ever before.
Stage 3. Evaluating Performance
Avoid the illusion of accuracy. Pursue an attainable goal. Uncertainty is a certain factor. It is not certainty in the forecast that demand forecasters need; it is an understanding of alternatives and possibilities in the presence of uncertainty. Forecasts should simply be recognized as having ambiguities and uncertainties. The key is to identify a model with the most appropriate forecast profile and prediction limits over the forecast horizon. Complement traditional measures of accuracy (based on forecast errors) with nonconventional resistant metrics that give protection against outliers and unusual values of known sources. Beware of the Myth of the MAPE!
Training. Establish a training program. Somehow, institutionalized expertise has a way of falling through the cracks in a rapidly changing environment because new demand forecasters have not maintained continuity with best practices of experienced forecasters long on the job. Likewise, what a forecaster needs to know today may become obsolete in the future. Companies need to maintain a training program for the professional development of demand forecasters. Just as the management process is subject to change, so is demand forecasting.
Stage 4. Reconciling Results
Cooperate. Achieve consensus on assumptions and rationale. Demand forecasts are based on assumptions about the economy, demographics, habit, and industry and market forces. Recognize differences among common assumptions, not just the numbers! Forecasters must strive towards a consensus on assumptions and rationale, rather than a confrontation over what the right number is.
Collaborate. Enhance forecasting knowledge in management, the sales force and customers. Forecast users and providers can absorb and use tools that work for them. They will apply forecasting techniques, no matter how sophisticated, in their daily tasks, if they provide them with tangible benefits. With more and better data, the value of enhanced forecasting knowledge and methods will soon become evident to forecast users and providers, giving them access to better forecasting information.
Communicate. Sell a forecast as advice. The forecast should be useful and provide information that is relevant to the decision-maker. Advice that is useful will get management approval, not wow the administration with technical know-how, sophisticated modeling output or excessive wit.
Hans Levenbach is Executive Director, CPDF Training and Certification Programs for the professional development of demand forecasters. He conducts onsite and hands-workshops on Demand Forecasting for multi-national supply chain companies worldwide. He is group manager of the LinkedIn groups (1) Demand Forecaster Training and Certification, Blended Learning, Predictive Visualization, and (2) New Product Forecasting and Innovation Planning, Cognitive Modeling, Predictive Visualization.