We all know the importance of demand forecasting in the supply chain, but the question people mostly ask which of the many forecasting methods is best suited for us?
Get Excess and Obsolete Inventory Policy
In my experience, Made-to-stock business has to forecast because that’s how they decide what to build. If there is no forecast and you are buying from long lead time suppliers, regardless of what planning method you are using, you might have no stock!
True make-to-order or engineer-to-order companies have forward order books going forward in future, fully covering the lead time. However, there are few true make-to-order companies because the lead time would be too long.
So, Who Owns the Forecast?
This answer can be found by asking another question… What is it a forecast of? Sales, of course. Then it should be sales team who should own the forecast!
However, in my experience, less than 30% of sales managers are willing to provide the sales forecast. And less than 50% are willing to look into statistical forecast data if provided by the supply chain.
I have heard this statement many times, “take last year sales data and add 5-7% on top for additional growth”, and they are same folks who want 95% on-time delivery performance on request😊.
In my humble but strong opinion sales should own the forecast and sales should want to measure their performance against it, so they must own it! Simply put sales and marketing teams are close to customers and close to the market and they should know what is happening.
In order to help business owners enjoy productive business activities even in the face of risks, many forecasting methods have been developed and used. However, let us first consider how to use demand forecasting methods.
Get The Ultimate Guide to Demand Forecasting
How to use Demand Forecasting Methods
Forecasting in itself is a risky adventure and the forecaster has a role to play in selecting the best demand forecasting method. Each of these methods has specially defined use and the appropriate one for a particular economic situation must be chosen carefully.
To use the method, factors which include the situational context of the forecast, access to historical data, level of accuracy expected, understand the 4 pillars of demand planning and the evaluated value of the forecast in respect to the business is important. It is important to assess them using available parameters before choosing a method.
Similarly, if a product is in question, the life cycle should be considered. Is it a perishable or durable product? Does it remain on shelves or leave the market as soon as it distributed to the middlemen or become obsolete? Is the product in a state of full growth or development? The ability to establish accurate data on these considerations will help choose an appropriate forecasting method.
Meanwhile, for successful forecasting, it is important for business managers and forecasters/ demand planners to work together. This will help to establish the goal of the forecast, how it is to be interpreted and applied to the situation at hand, tactics and strategy for precautionary motives, among others.
What are the Types of Forecasting Methods?
There are 14 forecasting method which can be applied to business situations. These various types are further categorized into quantitative and qualitative methods. Qualitative forecasting method is a subjective judgment based on the opinion expressed by consumers and market experts. This method is adopted when there is no historical data. On the other hand, quantitative forecasting method is engaged where there is access to historical data and unlike the former; it is objective to the degree of the reality of the data. They are presented below.
Quantitative Forecasting Methods
Moving Average. This is a time series method which involves a calculation to examine data points by creating an average series of various subsets from complete data. The formula involves a series of number and fixed subset size. The forecaster takes the average of the formerly fixed subset and then modifies it by taking out the first number of the series and adding the value that follows in the subset series. This method is statistical and it is usually adopted to deal with fluctuations only continuing for a short-term, technical analysis of financial data, evaluation of GDP, among others.
Exponential Smoothing. This is a simple method adopted to measure some determinations using existing assumptions by the user, such as seasonality. Using an algorithm that uses past data, the future is predicted. When compared to some other smoothing methods, it produces an easy result without requiring any minimum number of observations.
Regression Analysis. This covers a group of methods for forecasting that is dependent on information gathered from other variables (dependent and independent). Largely, it depends on the ability to use data generating process. There is simple linear regression which involves comparing an independent variable with a dependent variable and there is multiple linear regressions in which two or more independent variables are compared with one dependent variable.
Adaptive Smoothing. This method allows a business firm to key into different variables in order to arrive at every likely result from a particular business action or resolution. Statistical data and variable analysis are involved as well. It is common in firms without obvious quantities.
Graphical methods. This involves a statistical method but it is simple and useful for the sales forecast. With this method, periodic sales data for various years can be illustrated graphically with meeting points established by drawing free-hand lines. Based on the graph, the distance between points and line determines the minimum.
Econometric modeling. This is an improvement on regression analysis. It involves the calculation of independent regression under equation, variables and data. Meanwhile, it is economic theories that are adopted in the statistical method to determine the influence of one economic variable on another.
Life-cycle modeling. This method analyses and forecast the growth and development rates of a new product. The model brings together data that have been subjected to acceptance or rejection by different market groups such as creators, early and late adopters, early and late majority. It is the result that is used in forecasting sales for a new product.
Qualitative Forecasting Methods
All of the methods that will be discussed below are after the same goal- to forecast a useful market reception of a product. The forecasting will help to make a useful decision on the quantities to the produced and even distributed across the market one step at a time. Also, it will help to inform other decisions like budgeting, capital investment, and management of inventory.
Expert opinion. Here, the opinions of experts in the area where the forecast is to be made are weighed in order to make meaningful projections. It does not depend on statistical data hence it can be done where measurable data is lacking. This method is easy and quick and the team usually adjusts the result of the projection to their anticipation.
Market Research. It is possible for a business firm to carry out market research that will enable its sales forecast. This method may be executed by the staff members of the firm or another organisation, research firm, which it has been outsourced to. Whichever way, market research could involve strategies which include telephone, opinion poll or personal interviews and questionnaires.
Focus groups. This is a popular qualitative forecasting method. It involves engaging about five to ten people from a business firm’s target customers in an open-ended discussion. Usually, there is a moderator who sees to turn-taking among the participants and also asks questions related to their perception of the brand, products, slogans, design and related concepts. It is expected that participants will provide insightful responses which represent the opinion of a larger market it targets. Focus group discussions may involve incentives like a financial reward or any equivalent measure in terms of free good items.
Historical analogy. This is a forecasting method in which a sales history of a product having a parallel relationship with a present product is studied to predict future sales. It can be utilized to predict the market reception for a new product or group of products. This is done by using the historical data gathered over a period of time from a similar existing product either by the company or a formidable competitor.
Delphi method. This is a forecasting method in which market orientation and judgments of a small group of experts are combined using a function of iteration. The results of these iterated combinations help to develop the next parallel meeting points in order to discover an accurate forecast. Mind you, the opinions of the experts are gathered individually in order to avoid the influence of the opinion of a dominating personality if it were to be a group discussion method. Rather, an outsourced party handles opinions gathering; summarises and bring them before the same experts. New questions may be attached to it and the circle continues until a meeting point is arrived at. This method has proved effective and dependable for long-term forecasting.
Panel consensus. This method brings together members of a business firm across all levels to establish its forecast. It is an open process that allows all the participants to express themselves. Since it usually includes participants from the low level of the organisation’s hierarchy to superiors, there may be feelings of intimidation and suppression of opinions on the part of the former. For example, a sales manager who actually understands the market well may feel reluctant to contradict the opinions expresses by top managers like president and vice-president. In the end, the process of a panel consensus may not be truly open, fair and reliable.
Whilst the above traditional demand forecasting methods are tried and tested for decades, but they are now challenged and upgraded by modern demand forecasting methods using Machine Learning and Artificial Intelligence. You might find my blog “5 Reasons Why Machine Learning Forecasting Is Better Than Traditional Forecasting Techniques”, useful!
Forecasting is a tool that helps management to deal with the fluctuating market. With reliable historical data or opinions of the expert, a business firm will be able to predict future trends that can be manipulated to make the market work in their favour. The various demand forecasting methods available are categorized into quantitative and qualitative.
Meanwhile, it should be noted that there is no strict rule on the use of any forecasting method. If need be, it can be adjusted to the particular need of a business firm. Also, two or more forecasting methods can be adopted at a time by a business.
By Dr. Muddassir Ahmed