Importance Of Predicting Demand For Organizations
In a world of extreme competition, expense reduction being the mantra for most organizations, primarily in the retail and CPG industries, they try to focus on cost cutting and maintaining optimum levels of inventory to gain the competitive edge. To accomplish this, forecasting demand is of utmost importance. It is also not enough to have a macro level sales forecast for the entire organization.
Efficient and accurate demand forecast enables organizations, to anticipate demand and consequently allocate the optimal amount of resources to minimize stagnant inventory. This will result in negligible wastage of resources as well as reduction of costs such as storage cost, transportation cost etc. Another side effect of accurate demand prediction is the prevention of shrinkage so that firms don’t have to give huge discounts to clear stock.
This excerpt will touch upon the steps in demand forecasting and briefly, talk about the different demand forecasting methods. The article ends with some challenges of demand forecasting.
Steps To Follow In Demand Forecasting/Prediction
- PURPOSE: The most important aspect in demand forecasting, is to specify the very purpose of the forecast. The technique, interpretation, and results, will obviously be subject to the purpose itself. Failing which the relevance of the exercise will forfeit significance. Is the demand forecasting for a firm or the industry? Is it for a specific hierarchy of products or an entire range of products a firm can offer? Is it to maximize market share or revenue?
- TIME: What is the time-period for which the forecast will happen? Is it for the short run or long? Depending on the time period, the factors affecting demand can be assumed to be fluctuating or constant. This determines the kind of Analytical Dataset that needs to be created for the forecast.
- DATA: Theoretical statistics would say, that the technique of demand forecast, should precede the collection of data – but here we are part of the real and imperfect world. Hence, data is what it is, everything must work around it. The method, the approach, the accuracy and the interpretation. The quality and quantity of data is a slave to none – hence it is the advantage as well as the limitation at the same time.
- METHOD: Choosing the method of forecasting is the next piece of the puzzle. There are multiple methods to demand forecast – both statistical and otherwise which we shall talk about later on the blog. This is where the business relevance of the forecast and the above aspects of time and purpose come into play. Also, the most crucial aspect in choosing the method sometimes turns out to be the data. The most appropriate technique given the time and purpose is often deemed irrelevant by the limitations of the data. That’s when we analysts ‘improvise’.
- INTERPRETATION AND PRESENTATION: This in today’s world is the most important aspect of the conclusion of a project. Even the simplest of projects, packaged in an attractive shroud, can earn more accolades, than the brilliance of the technique, relevance of purpose or even effort of resources. More often than not, the audience to a solution, are oblivious of the former four steps of forecasting. Explaining the complications, and the adopted solutions incomprehensible language, rather than the use of jargons, is most often the order of the day in front of clients. Of course, the attractiveness of the presentation is substantially underrated in today’s analytics industry.
Methods Of Demand Forecasts
Under the survey method, the consumers are contacted directly and are asked about their intentions for a product and their future purchase plans. This method is often used when the forecasting of a demand is to be done for a short period of time. The survey method includes:
1. Consumer Survey Method: Techniques of demand forecasting that involve direct interview of the potential consumers.
2. Complete Enumeration Method: This is a comprehensive and time-consuming method where the forecaster needs to be abreast with future purchase plans of all potential customers. The probable demand for a product can be calculated by taking an arithmetic sum of all quantities obtained from the customer opinion.
Dp = D1+D2+D3+D4+……+Dn
Where, D1, D2, D3 denote the demand indicated by children 1, 2,3 and so on. One of the major limitations of this method is that it is only applicable where the consumers are concentrated in a certain geographic vicinity. If the population is dispersed, then it can turn out to be very expensive. Besides, the other limitation is that the consumers might not know their actual demand in future. Due to this, they may give a hypothetical answer that may be biased to their own expectations regarding the market conditions.
3. Sample Survey: The sample survey method is often used when the target population under study is sufficiently large. Only a sample of potential consumers is selected for the interview. Here, the method of the survey can be a direct interview or e-mailed questionnaires to the selected sample. The probable demand, indicating the response of the consumers can be estimated by using the following formula:
Where Dp = probable demand forecast; H = Census number of households from the relevant market; Hs = number of households surveyed or sample households; HR = Number of households reporting demand for a product; AD = Average Expected consumption by the reporting households (total quantity consumed by the reporting households/ Number of households.
This method is overly simplistic, less expensive and even less time-consuming as compared to the comprehensive survey methods. This method is mostly used to estimate a short-run demand of business firms, households, government agencies who plan their future purchases. However, the major drawback of the process is that not a lot of reliability can be placed on the forecast.
4. End-use Method: The end-use method is primarily used to predict demand for inputs. This method of has considerable theoretical value. In this method, a forecaster builds the schedule of probable future demand for inputs by aggregating industries and several other sectors. In this method, during the estimation of a demand, changes in technological, structural and other factors that influence the demand is taken into the consideration.
The end-use method helps in determining the future demand for an industrial product in details by type and size. Also, with the help of end-use method, a forecaster can pinpoint or trace at any time in the future as to where, why and how the actual consumption has deviated from the estimated demand.
5. Opinion Poll Methods: The Opinion Poll Methods are used to collect opinions of experts with sufficient knowledge about the market, such as sales executives, professional marketing experts, sales representatives and market consultants.
6. Expert-Opinion Method: Companies with enough resources in terms of human and financial capital, can capitalize on them while assessing the demand for a product in a region or locality. Since sales representatives are in touch with the customer directly, they are supposed to know the future purchase plans of their customers, their preference, their reaction to the introduction of new products, their reactions to the market changes and the demand for competitive alternatives.
Thus, sales representatives are likely to provide an approximate, if not accurate, estimation of demand for a product in their respective regions or areas. In the case of firms, which lack sales representatives, they can collect information regarding the demand for a product through professional market experts or consultants, who can predict the future demand based on their expertise and experience.
Although the expert opinion method is too simple and inexpensive, it suffers serious limitations. First, there is extreme subjectivity in the estimates provided, contingent on the skill and awareness of the professionals involved in the forecast. Consequently, there are chances of severe over or under-estimation of demand due to the subjective judgment of the assessor. Secondly, the evaluation of market demand is often based on inadequate information available to the sales representatives since they have a narrow perspective of the market, specifically competition.
7. Delphi Method: The Delphi method is an extrapolation of the expert opinion method where the various expert opinions are consolidated to estimate a final demand. Under this method, the experts are provided with the information related to estimates of forecasts of other experts along with the underlying assumptions. The experts can revise their estimates in the light of demand forecasts made by the other group of experts. The consensus of experts regarding the forecast results in a final forecast.
8. Consumer Clinics or Controlled Laboratory Method: Here the consumers are given some money to make purchases in stipulated store goods with different prices, packages, displays, etc. This experiment is expected to display the responsiveness towards the changes made in the prices, packaging and a display of the product. One of the major limitations of market experiment method is that it is too expensive and smaller firms cannot afford it. Also, this method is based on short-term controlled conditions which might not exist in the uncontrolled market. Therefore, the results may not be applicable in the long term uncontrolled conditions.
Statistical methods are often used when the forecasting of demand is to be done for a longer period. These methods utilize time-series (historical) and cross-sectional data to estimate the long-term demand for a product. Statistical methods are used more often and are considered superior than the other techniques of forecasting due to the following reasons:
- There is a minimum element of subjectivity in the statistical methods.
- The estimation method is scientific and depends on the relationship between the dependent and independent variables.
- The estimates are more reliable
- Also, the cost involved in the estimation of demand is the minimum
Most widely used statistical methods, can be classified into three major heads:
- Trend Projection Methods
- Barometric Methods
- Econometric Methods
1. Trend Projection Methods
- Graphical Method: It is the simplest statistical method in which the annual sales data are plotted on a graph, and a line is drawn through these plotted points. A free hand line is drawn in such a way that the distance between points and the line is the minimum. In this method, it is assumed that future sales will assume the same trend as followed by the past sales records. Although the graphical method is simple and inexpensive, it is not considered to be reliable. This is because the extension of the trend line may involve subjectivity and personal bias of the researcher.
- Fitting Trend Equation or Least Square Method: The least square method is a more numerical technique in which the trend-line is fitted in the time-series using the statistical data to determine the trend of demand. The form of trend equation that can be fitted to time-series data can be determined either by plotting the sales data or trying different forms of the equation that best fits the data. Once the data is plotted, it shows several trends. The most common types of trend equations are:
- Linear Trend: when the time-series data reveals a rising or a linear trend in sales, the following straight line equation is fitted:
S = a + bT
Where S = annual sales; T = time (years); a and b are constants.
Exponential Trend: The exponential trend is used when the data reveal that the total sales have increased over the past years either at an increasing rate or at a constant rate per unit time.
- Box-Jenkins Method: Box-Jenkins method is yet another forecasting method used for short-term projections. This method is often used with stationary time-series sales data. A stationary time-series data is the one which does not reveal a long term trend. In other words, Box-Jenkins method is used when the time-series data reveal monthly or seasonal variations that reappear at intervals.
2. Barometric Methods of Forecasting
The Barometric Method of Forecasting was developed to forecast the trend in the overall economic activities. This method can nevertheless be used in forecasting demand prospects, not necessarily the actual demand quantity expectations. The method is based on the approach of developing an index of relevant economic indicators and forecasting the future trends by examining the fluctuations of these indicators. Future trends can be examined using a time-series of relevant indicators. These can be classified as:
- Leading Series: The leading series comprises indicators which fluctuate up or down ahead of some other series. The most common examples of leading indicators are – net business investment index, a new order for durable goods, change in the value of inventories, corporate profits after tax, etc.
- Coincidental Series: The coincidental series include indicators which move up and down simultaneously with general economic activities. The examples of coincidental series – the rate of unemployment, the number of employees in the non-agricultural sector, sales recorded by manufacturing, retail, and trading sectors, gross national product at constant prices.
- Lagging Series: A series consisting of those indicators, which after some time-lag follows the change. Some of the lagging series are- outstanding loan, labor cost per unit production, lending rate for short-term loans, etc.
The following are the criteria on which the indicators are chosen:
- Time Series- statistical adequacy; a higher score is given to the indicator provided with adequate statistic
- The economic significance of the indicator; such as greater the significance the greater is the score of the indicator
- Conformity with the movement in overall economic activities
- Immediate availability of the time series
- The consistency of the series to the turning points in overall economic activities
- Smoothness of the series
The drawback of indicator selection is that some indicators appear in more than one class of the indicators. The only advantage of the barometric method of forecasting is that it helps to overcome the problem of finding the value of an independent variable under regression analysis. The major limitations of this method are:
- Often the leading indicator of the variable to be forecasted is difficult to evaluate or is not easily available.
- The barometric technique can be used only for a short-term forecasting.
3. Econometric Methods
Time series modeling
A time series model can predict trends based only on the original dataset that is used to create the model. Any new data added to the model when making a prediction is automatically incorporated into the trend analysis.
- Regression techniques: Regression analysis can be used to assess the relationship between one or more independent or predictor variables and the dependent or response variable – which is the variable we want to predict. Therefore, it can not only provide a forecast but also explain the relationships between dependent variables and independent factors. Multiple regression analysis involves two or more predictor variables, and is a slightly more advanced forecasting method, but believed to be the most accurate when used correctly. It can be used also for time-series analysis by incorporating trend, seasonality components into the analysis.
- Random Forest: Decision trees can select important “Features” and ignore irrelevant ones. It also gives a relationship between the features and predictions thus easing interpretation of models. However, in the case of a Random forest, it does not produce an explicit model to attain a relationship between the features and prediction, instead of as an ensemble of trees inherits the property of ability to select important features. This is a black box modeling approach.
- Gradient Boosting and Bagging: As the name suggests Bagging and Boosting are techniques that are used for creating multiple samples of the data and learning from each sample to accomplish accurate predictions. It is a machine learning algorithm that produces regression and classification predictions in the form of an ensemble of models.
Challenges Of Demand Forecasts
By far the greatest challenges to the most accurate forecast are:
- Forecasting of new products, that do not have sufficient history for the analyst to gauge its behavior. It poses a challenge in particularly retail sectors such as electronics, fashion, books and gardening, where new product introductions and heavily refreshed seasonal assortments account for the bulk of sales. One of the most extreme is the book industry where, as a rule, more than 90% of items sold are new that year. So, if you do not get the forecasting of product introductions right, you can forecast only a small fraction of your business.
- Tackling variance in sales volume is pretty much inevitable in modern retail environments. This should be a standard feature of most retail forecasting processes. The key to tackling these change situations effectively is to bring together statistical forecasting with human insight. Sometimes a change can indeed be traced to wider market conditions and that’s where our ability to infer a causal link from events serves us well; it’s where humans still outperform computers.
- Centralizing stock control and estimating supplier lead times, is another real-world scenario, where forecasts, can fall apart. Both excess stock and obsolete inventory are easier to prevent than to eliminate. Centralized systems can help prevent costly over ordering and reliance on suppliers and vendors can be lessened through inventory redistribution functionality across multiple stock locations. This enables stock to be moved where it is needed, eliminating unnecessary ordering and reducing inventory.
- Analyzing the impact of mid-season promotions, discounts, as well as potential changes to business (wins, losses, leads etc.). Thus, most top-down revenue estimates do not account for new sales opportunities. Few diligent analysts manually approximate incremental revenue from key new products, while assuming all else equal. This approach is better than not including any new revenue. But it still largely does not account for any bottom-up information from the field about new wins, losses or potential leads that could be converted to opportunities. It also does not account for any changes in competitive landscape or price degradation.
- Sales Forecasting for Retail Chains by Ankur Jain, Manghat Nitish Menon, Saurabh Chandra, 2015