Everything you need to know about the methods and techniques of demand forecasting. Demand forecasting is done by companies and they employ different methods ranging from qualitative to quantitative methods.
The choice of method generally depends upon the size and complexity of the firm. Smaller firms usually have more informal form of human resource planning and hence many times rely on more qualitative methods.
Whereas, larger firms usually having multiple departments, levels and higher mobility of workforce both within and outside the firm, generally use more quantitative methods.
A: The following are the methods used in demand forecasting:-
1. Delphi Method 2. Product Life Cycle Model 3. Time Series Analysis 4. Regression Analysis 5. Exponential Smoothing 6. Extrapolation 7. Independent Technological Comparisons 8. Barometric Forecasting 9. Econometric Models 10. Opinion Polling.
B: A great variety of methods are available for forecasting the demand of a consumer:-
They are – 1. Jury Method/Executive Opinion Method 2. Survey of Expert Opinion Method 3. The Delphi Method 4. Demand force Composite Method 5. User Expectation Method/End-Use Method/Survey of Buyers Intentions Method 6. Market Share Method. 7. Analytical and Statistical Methods 8. Market Survey Method.
C: There are two main methods of demand forecasting:
They are – 1. Qualitative Methods 2. Quantitative Methods.
D: Following are the two methods adopted for demand forecasting:-
1. Survey Methods 2. Statistical Methods.
Although, there are several methods of forecasting the demand for a product or a service, yet no forecasting method is useful for all the products.
A company should use more than one forecasting method for correct decision making. The choice of the forecasting method also differs from product to product to the objective of the company to data availability besides others. Although forecasting may act as a light to guide the path, but it may not be accurate all the time in this dynamic business environment.
Learn about the Methods and Techniques of Demand Forecasting
Methods of Demand Forecasting – Top 10 Methods: Delphi Method, Product Life Cycle Model, Time Series Analysis, Regression Analysis and Opinion Polling
Forecaster use past data in the following way:
(a) The forecaster analyses the past data in order to identify a pattern that can be used to describe it.
(b) The pattern is extrapolated or extended into the future in order to prepare a forecast.
The following are the method used in demand forecasting:
This method involves using a panel of experts to produce predictions concerning a specific question, such as when a new development would occur. The use of the Delphi method assumes that the panel members are recognized experts, and it also assumes that the combined knowledge of the panel members will produce predictions atleast as good as those that would be produced by one member.
This relies on the judgment of a number of skilled judges such as experienced marketing managers, and marketing consultants. This is a method of scenario building, in which a group of experts are asked individually to provide their views about the demand of the product.
The Delphi technique adopts the following methodology:
Step 1 – Estimate of sales are obtained from each judge independently.
Step 2 – These independent estimates are circulated in an aggregated form for information and reflection to all other judges.
Step 3 – This leads to new estimates being made by all judges.
Step 4 – The opinions of each expert are collated. ‘Extreme’ views are discarded, and a draft ‘consensus’ view is formulated.
Step 5 – The draft ‘consensus’ is circulated to the experts for their further comments, and depending on how they respond, the ‘consensus’ might be amended.
Step 6 – The process will continue until a forecast for the future has been prepared which has the acceptance of all or most of the panel of experts.
Step 7 – When there is some uncertainty among the experts, probability weightage might be given to possible future ‘scenarios’ or ‘events’.
Step 8 – The entire process is repeated until consensus is reached on an acceptable sales forecast.
Delphi method is based on the assumption that the judgement and skill of experienced managers is the most valuable forecasting tool. However, the technique remains subjective and inexact. It can also be time consuming-especially if initial estimates differ significantly from one another.
Consider the case of a company introducing a new product and wishing to forecast sales of the product for the next three years. When predicting sales of a new product, it might be appropriate to consider the ‘product life cycle’. The product life cycle model splits the life of the product into four stages- Introduction, growth, maturity and decline. When forecasting sales of the product during the growth stage, the company might use the experts opinion of its sales and marketing personnel to subjectively construct an ‘S’ curve.
The S curve then be used to forecast sales during this stage. The company might use its experience with other products and knowledge concerning the new product, in constructing the S curve. It will need to determine how long it will take for the rapid increase in sales to begin, how long this rapid growth will continue, and when sales of the product will begin to stabilize. Estimating such curve is an example of subjective curve fitting and it requires a great deal of expertise and judgment.
Time series analysis means analyzing the historical patterns of sales that have occurred in the past as a means of predicting future sales.
It helps to identify and explain the following:
(a) Any regular or systematic variation in the series of data which is due to seasonality-the ‘seasonals’.
(b) Cyclical patterns
(c) Trends in the data
(d) Growth rates of these trends
This method can be useful when no major environmental changes are expected and it does highlight seasonal variations in sales and consumer demand. However, time series analysis is limited when organizations face volatile environments.
Time series analysis helps in understanding the past behaviour of a variable in determining the rate of growth and the extent and direction of periodic fluctuations. The study of past behaviour of a variable enables us to predict future tendencies. The actual performance can be compared with expected performance and the cause of variation analyzed. Different time series are compared and important conclusions drawn from time series data.
Regression analysis attempts to establish the nature of relationship between variables to study the functional relationship between the variables and thereby provide a mechanism for prediction or forecasting. Regression analysis is a statistical technique with the help of which we are in a position to predict or forecast unknown values of one variable from the known values of another variable.
For example by setting the amount for advertisement campaign we cm predict sales through regression analysis technique. Here, advertisement expenditure budgeted is called independent variable and the sales to be forecasted is called dependent variable.
The independent variable is denoted by V and the dependent variable is denoted by ‘y’. The analysis is called ‘simple linear regression analysis’, which is concerned with bivariate distributions, that is distributions of two variables. If two variables are functionally related then a knowledge of one will make possible an estimate of the other. If we know that advertising and sales are correlated, then for a given advertising expenditure, we can find out the probable increase in sales and vice versa.
Simple regression analysis provides estimates of values of the dependent variable from values of the independent variable. The device used to accomplish this estimation procedure in the regression line. The regression line describes the average relationship existing between V and ‘y’ variables i.e. it displays mean values of Y for given values of ‘y’. The equation of this line, known as the regression equation, provides estimates of the dependent variable when values of the independent variable are inserted into the equation.
The equation of this line is described by:
y = a + bx
y = Dependent variable i.e. sales
x = Independent variable i.e. advertisement expenses per unit of sales
a = Value of ‘y’ when ‘x’ is equal to zero
b = The amount of change that come in ‘y’ for a unit change in ‘x’
Method # 5. Exponential Smoothing:
Exponential smoothing is a forecasting technique that attempts to ‘track’ changes in time series by using newly observed time series values to ‘update’ the estimates of the parameters describing the time series.
The model used is:
Forecast for period T + 1 = Forecast for period T + α (Actual for period T – Forecast for period T)
The terms in the brackets are regarded as the error term, as it is the difference between the actual result and its forecast. The forecast for the next period is the previous forecast plus a proportion of the error term. The proportion of alpha lies between 0 and 1 and is known as the smoothing constant. Each new forecast depends upon the previous result and its forecast, but this old forecast depends upon the previous result and so on.
Extrapolation is the simplest yet often a useful method of forecasting. Extrapolation relies on the relative consistency in the pattern of past movements in some time series. Extrapolation is used frequently for sales forecasts and for other estimates when ‘^^’forecasting methods may not be justified.
Under extrapolation, the assumption that next year’s volume of sales will be equal to this year’s figure or that next year’s growth of sales will be equal to this year’s. A slightly more sophisticated version is to identify any trends over the recent past and then to extrapolate those trends forward into the future.
This method is often used to produce technological change. The method involves predicting changes in one area by monitoring changes that take place in another area. That is, the forecaster tries to determine a pattern of change in one area, called a primary trend, which he believes will result in new developments being made in some other area.
A forecast of developments in the second area can then be made by monitoring developments in the first area. This type of forecasting poses two basic problems. Firstly, the forecaster must identify a primary trend that will reliably predict events in the area of interest. Secondly, the forecaster must use his expertise to determine the precise relationship between the primary trend and the events to be forecast.
Time series analysis use information about the past in order to make forecast of the future. Barometric forecasting uses indicators of current activity in order to provide forecasts of the future. Perhaps the most common barometric technique is the use of leading indicators. A leading indicator is a variable which is known, or believed, to be correlated with the future behaviour of the variable for which a forecast is required.
Given the importance to businessmen of being able to predict general movements in the level of economic activity, there are a large number of leading indicators which are used in the attempt to identify changes in total spending, income and employment. These include new orders for machine tools, which often rise in advance of an increase in economic activity, the length of the working week, and the performance of the Stock Exchange.
A number of such general leading indicators used are as follows:
(a) New orders for machine tools
(b) Average hours worked in manufacturing
(c) Index of new business formation
(d) New orders for durable goods
(e) Orders for plant and equipment
(f) New building starts
(g) Changes in manufacturing inventories
(h) Industrial material prices
(i) Stock Exchange indices
(j) Profit figures
(k) Price to unit labour cost ratios
(l) Increase in consumer debt
Method # 9. Econometric Models:
A variety of forecasting models might be used by an organization. One type of model which can be used for both short-term and medium-term forecasting is an econometric model. Econometrics is the study of economic variables and how they are interrelated, using a computer model. Econometric techniques have recently gained popularity in forecasting.
The term ‘econometrics’ refers to the application of mathematical economic theory and statistical procedures to economic data in order to verify economic theorems and to establish quantitative results in economics. Econometric models take the form of a set of simultaneous equations.
The values of the constants in such equations are supplied by a study of statistical time series and a large number of equations may be necessary to produce an adequate model. The work of computations is greatly facilitated by electronic data processing equipment like computer etc.
(a) The likely rate of cost inflation
(b) The likely level of interest rates
(c) The expected growth in the economy and consumer demand
(d) Expected movements in foreign exchange rates
Method # 10. Opinion Polling:
Opinion poll is the survey of opinion of experts i.e. knowledgeable persons in the field whose views carry lot of weight. For example, a survey of opinion of sales representatives, wholesalers, retailers etc. shall be of great help in formulating demand projections.
Methods of Demand Forecasting – Top 8 Demand Forecasting Methods Adopted by a Firm
A great variety of methods are available for forecasting the demand of a consumer.
The noteworthy ones among them are listed below:
Demand Forecasting Methods:
1. Jury Method/Executive Opinion Method
(a) Top Jury Method
(b) Percolated Jury Method
2. Survey of Expert Opinion Method
3. The Delphi Method
4. Demand Force Composite Method
5. User Expectation Method/End-Use Method/Survey of Buyers Intentions Method
6. Market Share Method.
7. Analytical and Statistical Methods
(a) Simple Projection Method
(b) Extrapolation Method
(c) Moving Averages Method
(d) Exponential Smoothing
(e) Time Series Analysis
(f) Regression Analysis
(g) Complex Econometric Models
8. Market Survey Method
We shall discuss these methods sequentially:
The jury method is a commonly applied method of demand forecasting. It is also known as executive opinion method. Judgment is the crucial factor in this method. This is true of both the ‘top jury method’ and the ‘percolated jury method’. The difference is that in the former the participants are limited to the top executives and in the latter, a large number of marketing/ demand executives participate.
In both, the participants exercise their judgment and give out their opinions. By a rough averaging of these opinions, the final forecast is arrived at. Evidently, for the forecasts arrived at by this method to be reliable, the executives participating in the forecasting exercise must have versatile experience in and sound knowledge of the business.
They must also be well informed about the overall economic environment and the conditions prevailing in the industry. They should also know the strengths and weaknesses of their firm.
The jury method is a judgment method. The method gives due weight to the experience and judgment of people who know the market and the firm. It is an easy and simple method; the demand forecast can be ironed out in a short time. Further, when a firm lacks the experience and expertise required for using sophisticated analytical methods of forecasting, the jury method would perhaps be the only method that would be handy.
Again, when adequate past statistics on demand and market are not available with the firm, the jury method can be used. The jury method has some demerits too. After all, the estimates or forecasts arrived at by this method are based on ‘opinions’ and not ‘facts’.
And, the method disperses the responsibility for forecasting on a number of people. Another limitation is that the forecasts worked out by this method are not readily amenable for breaking down into Product-wise, Month-wise and territory-wise forecast, separate exercises are often required for this purpose.
This is yet another judgment based method of demand forecasting but is somewhat different from the jury method. In the jury method, opinions of executives give rise to the forecast. In survey of expert opinion method, experts in the concerned field inside or outside the organization are approached for their estimates. This method may be more useful in developing total industry forecast than consumer level demand forecast.
The Delphi method too is a kind of survey of expert opinion method; it is used more for broad-based futuristic estimates than for demand forecasts. In this method, experts in the field are interrogated by a sequence of questionnaires. Any information that is available with anyone member of the panel is passed on to others as well, enabling all the members of the panel to have access to all the available information.
This technique’ eliminates the ‘bandwagon’ effect of majority opinion. The panel members are asked to react to a checklist of questions which are significant to the forecast that is attempted. Their opinions and reactions are analyzed and where there is a sharp difference on an issue, interchanges are peffilitted and the final forecasts are presented issue by issue.
As per the demand force composite method, the demand forecasting is done by the demand force. It is also a judgment based method. Each demand man develops the forecast for his respective territory; the territory- wise forecasts are consolidated at branch/area/region level; and the aggregate of all these forecasts is taken as the corporate forecast.
It can be easily seen that the sale force composite method is similar to the jury method. The difference is that the jury method depends on the judgment of a few executives and the demand force composite method seeks to aggregate the judgments of the entire demand force
The responsibility for forecasting and the responsibility for achieving the demand as per the forecasts are entrusted with the same set of people viz the demand men, Since the demand men themselves have made the forecast, they will accept the demand quotas based on this forecast as fair allocation and try their best to achieve the quotas.
Moreover, the forecasts developed by this method have greater stability and reliability because of the largeness of the sample. Again, the forecasts derived by this method could be easily and meaningfully broken down territory-wise, product-wise, customer type wise and month- wise Since the forecast is developed on this basis, to start with, keeping the micro level conditions as the base. And coordination by field demand management becomes more meaningful when forecast are made by the demand force and integrated by the field demand management.
The method has some demerits too. Demand men are certainly not experts in forecasting. They cannot use sophisticated techniques of forecasting. Nor, do they have all the data required for ‘fact-based forecasting’. They are often heavily influenced by the conditions prevailing in their territories and tend to be over-optimistic or over- pessimistic about future demand.
Since their demand quotas are to be based on their demand forecasts, they may tend to under estimate demand and play it safe. Again, while the demand men may know their territories well, they may not know equally well the broad changes taking place in the economy and the given industry. Such knowledge will be necessary to predict the future, especially when major changes are taking place in the macro environment.
As per the ‘user expectation’ or’ end use’ method, the various users of the product under forecasting are listed first; then their individual likely demand of the product is ascertained and from that data, the demand forecast for the product u: consolidated.’ This method is alternately known as ‘Survey of buyers’ intentions.’
The user survey will give an idea of the total possible consumption of the product, the buying plans of the users and the possible market share for the consumer doing the survey. The user survey can be made either on a sampling basis or on a census basis depending on the size of the user group to the covered. Census survey would naturally provide a more reliable forecast.
The user expectation method is particularly suited for forecasting the demand of industrial goods-industrial raw materials as well as intermediates and industrial durables. And there is a variety of reasons for this. For example, unlike consumer goods, the users of industrial goods are limited in number thus making it possible to exhaustively survey them.
Again, the demand made to individual buyers are substantial in the case of industrial goods. In the case of consumer goods, several thousands of buyers make up the total, each with a very small quantity of purchase. The industrial customers are generally clustered in the industrial belts unlike the general consumers who are scattered everywhere. Industrial selling usually takes place direct from the suppliers to the user industries without going through a long-winding distribution channel.
The buyer behaviour in this case is also more rational and predictable as the buyers are professional people. And finally, applying the method of user survey to industrial demand forecasting is easy and inexpensive. In any user survey, the decision-making groups in the user companies as well as persons with good knowledge of the product are contacted.
For example, the production people, the finance people, the materials and corporate planning executives of the user consumer are contacted to know how much their firm is likely to consume and how much of it they are likely to buy from the consumer doing the survey.
Demand forecast can be developed by yet another method-the market share method. The planned market share of the firm is the key factor in this method. The firms first work out the industry forecast, apply the market share factor and deduce the consumer’s demand forecast. The market share factor is developed based on past trend, consumer’s present competitive position, brand preference, etc.
Such conversion of industry forecast into consumer demand forecast requires considerable expertise. By a detailed marketing audit, the firm must correctly appraise its market standing, brand image, market share and strengths and weaknesses as compared with the competitors in the industry.
It must also correctly assess through reliable marketing intelligence, its competitor’s plans policies and activities. Only then, the market share factor and therefore tile demand forecast arrived at by this method will have a good degree of reliability. Retail audit would also be of considerable help in employing the market share method; it would help assess the industry position as well as the individual firm’s market shares.
As mentioned earlier, besides judgment methods, a wide variety of analytical and statistical methods is available for forecasting the demand of a business firm. The firm can choose the most appropriate one depending on its forecasting needs.
The methods include:
i. Simple Projection Method
ii. Extrapolation Method
iii. Moving Averages Method
iv. Exponential Smoothing
v. Time Series Analysis
vi. Regression Analysis
vii. Complex Econometric Models
In what follows, we shall describe the salient features of each of these methods.
Among the projection methods, the simplest is the one that uses the ‘rule of the thumb’ by which current year’s forecast is arrived at by simply adding an assumed growth rate to the last year’s demand; some firms go by the industry growth rate and project the demand; some others take the growth rate adopted by the industry leader.
Another formula as shown below is also used by some firms:
Only if the year by year demand sire stable and show an increasing trend, this formula will provide a reasonably reliable estimate.
The simple projection method provides a rough and ready forecast. Sometimes, the forecast arrived at by this method can be wide off the murk. The main limitation is that this method assumes past demand as the only factor influencing future demand. The method does not provide for the sharp changes that may happen in the current year, due to a variety of factors.
However, when the forecasting job of the firm is relatively simple and when the firm is in the mature stage of its business without much of growth or decline and when the external changes are not violent either, the simple projection method will serve the purpose. It has the added advantage of being inexpensive.
Some firms rely on the extrapolation method. Extrapolation is also a projection/ trend method. It involves the plotting of the demand figures for the past several years and stretching of the line, or the curve as the case may be. The extrapolation will give the figures for the coming years.
Extrapolation basically assumes that the variables will follow its previously established pattern. Accordingly” this method will be effective where the pattern of past movement has been relatively steady and abrupt disruptions are unlikely in the future. In other words, the assumption is that the future will mirror the past.
This method helps eliminate the effects of seasonality and other irregular trends in demand while forecasting the future figures. The method gives a time series of moving averages; each point of the time series is the arithmetical or weighted average of a number of preceding consecutive points of the series.
If seasonal effects are present in the demand pattern of the product, a minimum of two years demand history is needed for applying this method.
Exponential Smoothing is yet another projection method used for demand forecasting. It is similar to moving averages and used fairly extensively) It too represents the weighted sums of all past numbers in a time series wit.~ the heaviest weight placed on the most recent data. This method is particularly useful when forecasts of a large number of items are made.
It is not necessary to keep a long history of past data. -The method can have a stable response to changes and responses can be adjusted as required. This method is also adaptable for trend corrections and smoothing of forecast errors. Exponential smoothing is one of the most accurate statistical techniques available for forecasting.
Another statistical method that is extensively used in demand forecasting is the Time series analysis also known as Trend cycle analysis. A Time series is a set of chronologically ordered points of raw data, for example, demand of a given product, by month, for several continuous years.
Time series analysis helps identify and explain:
(i) Systematic variation or ‘seasonal’ variation which arises due to seasonality in the series of data,
(ii) Cyclical patterns that repeat every two years or every three years and so on.
(iii) Trends in the data,
(iv) Growth rates of these trends.
The main assumption in Time series analysis is that the factors influencing demand will not change very much over a period of time and that the future will reflect the past. In this sense, this method is also basically a projection method. But, in Time series analysis, a statistical procedure is used to analyze historical demand data. Projections of future demand are made by studying the interaction of the basic and significant influences of demand. A thorough and systematic analysis of past data is carried out.
All basic factors underlying demand fluctuations are analyzed.
The four main types of demand variations:
(1) Long term growth trends (secular trends),
(2) Cyclical changes,
(3) Seasonal variations, and
(4) Irregular or random fluctuations-are isolated and measured using the statistical procedure.
The trend lines for each type of variation are studied and demand estimates are made. A mathematical model describing the past behavior of the series is selected, assumed values for each type of variation are inserted and the demand forecast is cranked out.
Time series analysis is especially suited for long range forecasts. This method will give more reliable forecasts for a 10-15 year period than for a year to year prediction. When demand patterns and therefore, the demand variation patterns, are well defined and relatively stable and factors leading to variations are easily established, this method could be successfully used for even the short-term forecasts.
A problem with the Time series method is that whereas it is easy to explain why a particular trend is going the way it does, it is not equally easy to predict when the turns will actually set in.
vi. Regression Analysis:
Regression analysis is another analytical technique used for demand forecasting. This technique tries to functionally relate demand to those variables that influence demand. They may be economic factors, competitive factors or price. The variable which is to be forecast is the dependant variable and the factors which cause change.’) In the dependant variable are explanatory or causal variables.
‘The association between the dependant variable (i.e., the demand forecast of the consumer) and the explanatory or causal variables is determined and measured. An equation is fitted to explain the fluctuations in demand in terms of casual variables.
A study of past demand trend may show different relationships between demand and the factors influencing the demand this relationship can be expressed as:
y = a + b1* x1 + b2* x2 +… + bn *xn
Called the ‘regression equation’ representing the relationship between demand and a host of causal factors, y represents demand; XI, X2, , Xn represent causal factors; and a, b1, b2 , bn are constants indicating the extent of contribution of each causal factor to total demand.
After establishing the relationship based on past data and with the estimated values for the factors for future years, we can get the demand estimates for the future years.
Where demand are influenced by two or more causal variables acting together, multiple regression analysis is applied. Computers Eire used for regression analysis involving complex calculations.
The regression method, in general, will give more accurate forecasts than the trend method since regression takes into account causal factors. At the same time, in regress ion analysis involving a number of causal variables, the error of forecasting will multiply along with the error in determining and measuring the relationship or influence of each of these variables.
Econometric models constitute yet another analytical method of demand forecasting. Econometrics basically attempts to express economic theories in mathematical terms so that they can be verified by statistical methods and used to measure the impact of one economic variable upon another for predicting future events. The econometric forecasting models vividly portray the real world situations and the multiple variables involved in the demand situation.
The econometric model is different from the regression model of forecasting.
The econometric model relies on the following principle/ theory:
i. Demand of product depends on several variables
ii. While demand is the dependant variable, the causal factors are the independent variables.
iii. There is constant interaction between demand and each of the causal/ independent variables.
iv. There is also constant interaction among the independent variables themselves.
v. The independent variables consist of two sets; there are exogenous variables constituted by non-economic forces such as nature, or politics, and there are endogenous variables constituted by economic forces such as income, employment, price, etc.
vi. The interrelationships between demand and independent variables can be estimated by statistical analysis of past data.
The econometric model is constituted by a set of interdependent equations that’ describe and simulate the total 3ales situation. The forecast is derived through this set of equations. Stated in simple terms, in this method, consumption figures for the past few years are taken as tile basic data; the relationship for analyzing the time series of consumption figures is provided by the model; the best trend is selected by adopting appropriate statistical tests for the goodness of fit; and based on the analysis of the time series of consumption figures, the forecast for the future specified period is derived.
The econometric models are quite complex and expensive to develop. But they predict Ute turning points more accurately. The econometric models are used more in forecasting the demand of durable goods-indu5trial as well as consumer durables, where ‘replacement demand’ is a significant factor to be projected. Likewise, they are used more for forecasting industry- level demand than consumer-level demand.
The market survey method is yet another method available for demand / demand forecasting. The term market survey is used by some as synonymous with market research or market analysis. This is incorrect. Market survey is only a technique or method of market research or market analysis. Its purpose is to collect specific data concerning the market that cannot be had from the consumer’s internal records or from external published sources of data.
When a market survey is used for generating relevant market information and such information forms the basis of the demand forecasts, the forecasting method is referred to as the market survey method of demand forecasting.
Normally, when a consumer wants to introduce a new product or an improved product, it resorts to a market survey to assess the likely demand for the product. Likewise, any new consumer entering the market for the first time, resorts to the market survey method for forecasting its demand/ demand. This is quite natural. The firm does not have any data of past demand or past demand patterns to fall back upon. It has to gather the information from the market and take decisions.
Usually, the firm conducts a survey among a sample of consumers and gauges their attitudes, likely purchases and purchase habits. Sometimes, a survey is conducted among the channel members -wholesalers, and/ or retailers to elicit information on their attitudes, likely purchases, etc. The merit of the market survey method lies in the fact that the method facilitates gathering of original or primary data that is specific to the problem on hand.
The main demerit is that it is time consuming and expensive. Moreover, the reliability of the information generated is dependent on the statistical accuracy of the survey procedures. Market survey, as a technique of market research and as a technique of data collection from the field.
The forecaster must carefully choose a method of forecasting from among the wide variety of methods available. Basically, the method chosen must match the requirements of his product and the organization. Each organization and each product has certain peculiarities from the standpoint of demand forecasting.
The forecaster must analyse thoroughly the organizational and product peculiarities and choose those methods of forecasting that match the requirements of his products and organization. The method should also lead to timely forecasts. It must also generate the forecasts in such a manner that they are readily discernible to the people implementing the demand forecasts demand plans.
Similarly, the chosen forecasting method should also match the environment in which the firm is operating-the technological environment, the competitive environment, the governmental influences, etc. The period range of the forecast is also a consideration in the choice of the forecasting method. And finally, the cost consideration and the extent of availability of qualified personnel for the forecasting job are also relevant factors in the choice of the forecasting methods.
In short, it must be understood that since all the methods have their associated merits and demerits, there is nothing like an ideal forecasting method that could be applied to advantage in the situations. The forecaster must assess the suitability of the specific method to his specific situation before commissioning the forecasting exercise.
The forecaster can improve his forecast by choosing a combination of more than one method. Since no one method of demand forecasting is perfect or foolproof, it would be generally advantageous to try out a combination of different methods. Using more than one method would give a better insight into the situation.
Cross-checking between one method and the other would minimize the risks involved in the forecast and improve the reliability of the forecast. Combinations would also help the forecaster to probe deeply the reasons for wide variations between the forecasts arrived at by different methods.
Such a probing will eventually enable the forecaster to arrive at a more accurate and reliable forecast. Conversely, if the results arrived at by the different methods converge, the confidence in the forecasts is enhanced to that extent.
Depending totally on a single method or one category of methods has certain pitfalls. For example, if the firm depends total on the jury method, it exposes itself to one type of bias. For, the jury method relies heavily on the judgment of persons who are not experts in forecasting and who normally do not employ elaborate database in working out the forecasts. It would be advantageous to supplement this method with one of the statistical methods.
The different methods of forecasting are not mutually competitive, nor intractably exclusive. Quite often they supplement one another and can be easily used in conjunction. The forecaster cam choose one method from the statistical/ analytical group and one method from the non-statistical group and compare the forecasts; he can also use two different methods from the same group and compare the position.
For example, in a consumer product, both the jury method and the demand force composite method could be tried out and the variations between the two forecasts could be studied. Such an approach may lead to a more reliable forecast. For, even though both the jury method, and the demand force composite method belong to the category of judgement methods, they have complementary merits and demerits.
Whereas the jury method is a ‘break- down’ method, the demand force composite method is a “build-up” method. The jury method involves working out the overall consumer forecast in the first instance and then breaking it into parts demand man-wise, territory-wise, channel-wise and month-wise; the demand force composite method involves the reverse procedure. When both the methods are employed, the accuracy of the forecast is improved through comparison and. through an analysis of why significant variances occur in the two methods. And when an analytical or quantitative method is also used in addition to the above two methods, the comparison would become complete.
Methods of Demand Forecasting – Qualitative and Quantitative Methods
Demand forecasting is done by companies and they employ different methods ranging from qualitative to quantitative methods. The choice of method generally depends upon the size and complexity of the firm. Smaller firms usually have more informal form of human resource planning and hence many times rely on more qualitative methods. Whereas, larger firms usually having multiple departments, levels and higher mobility of workforce both within and outside the firm, generally use more quantitative methods.
Let us have look at the various qualitative and quantitative methods that can be employed to forecast the demand of workforce:
a. Judgmental Methods:
Employee and managerial judgement are used to forecast the demand of labour. The approach could differ from choosing to employ only managerial judgement (more of a top to down driven forecasting) to something like using a combination of employee and managerial judgement (more of a down-up driven approach).
b. Delphi Technique:
The Delphi technique employs the judgement of the experts. Generally a panel of experts is chosen. They are then polled for their forecasts. The average of such forecasts is then taken as the demand of workforce in that firm.
Qualitative methods have their own limitations and hence reliance is more on more hard-data driven forecasts.
Two commonly used quantitative techniques are:
a. Trend Analysis:
In trend analysis first a business factor relevant to human resource needs is chosen, for example sales, production, etc. After this a historical trend of the business factor in relation to number of employees is plotted. The ratio of employees to the business factor provides a labor productivity ratio (for example- sales per employee or units produced per employee etc.).
Then such ratios are compared for at least the past five years. Finally, human resources demand is calculated by dividing the business factor by the productivity ratio and projection of human resources demand is done to the target year.
b. Workload Analysis:
In workload analysis, after considering the workload, the planned man-hours are calculated. Then the productive hours per worker is estimated. The total of planned-man-hours is divided by the productive hours per worker to forecast the demand of workers.
Although there are several methods of forecasting the demand for a product or a service, yet no forecasting method is useful for all the products. A company should use more than one forecasting method for correct decision making. The choice of the forecasting method also differs from product to product to the objective of the company to data availability besides others. Although forecasting may act as a light to guide the path, but it may not be accurate all the time in this dynamic business environment.
For example, the forecasting done for two cars i.e. Maruti Suzuki Swift and Mahindra Logan, met with different fate in reality. Maruti Suzuki Swift was launched in 2005 with the sales target of 4000 units a month. But it became a runaway hit among the car buyers and sold more than 40000 cars in the first six months, which is almost 65% higher than the forecast.
The success story does not end here. It went on to sell as many as 12000 units a month on an average. To continue the success story, Maruti also introduced the sedan version of swift in the name of Swift Dzire. Now Dzire is also selling 10000 units a month on an average.
On the other hand, Mahindra and Ranault’s partnership came to an end with the debacle of their joint venture product Logan. With a sales forecast of 50000 units a year, Logan could not achieve even its 15% of its target.
Following are the forecasting methods, which are generally used. All these methods are divided into two categories i.e. Survey methods and Statistical Methods:
Survey methods help a marketer to forecast the demand based on the buyer intentions, expert opinions and the feedback from the market
i. Survey of Buyer Intentions:
In survey of buyer intentions, a sample size of current and / or potential customers is selected and is asked about the purchase intentions of a particular product at a given price within a specified period. Then the data is extrapolated to the whole population, to have the right forecast.
But the sample may not be the representative of the population. This may result in inaccurate demand forecast. It may also be possible that the intentions may not translate into the actual purchases by the consumers.
ii. Sales Force Composite:
Sales force composite is based on the collective data provided by the sales people all across the region or state or territory or country. In terms of reliability, it scores better than the other methods, as it is based on the real data collected by right people. Although there are still chances of being over estimation or under estimation.
iii. Reasoned Opinion:
Reasoned opinion is also known as Delphi method, which is a variant of poll and survey method. This method was developed by Rand Corporation, USA for technical changes in the late 1940s. Just like sales force composite, this method uses the opinions of the experts and a group of knowledgeable individuals anonymously to estimate the future demand. But, this method can also be subjective, as it is based on somebody’s estimation.
iv. Test Marketing:
Under test marketing, a company test markets its products in a limited geographical area, which may be representative of the whole population to measure the demand. Test marketing also facilitates feedback on the packaging, pricing, design, likability etc. besides gauging the demand. In India, Pune and Hyderabad has been a major testing ground for FMCG companies.
Now Lucknow is also added to the list, with Hindustan Unilever Ltd. (HUL) test marketing its mobile tea/coffee vendors. Now the test marketing has expanded to Hyderabad, Mumbai and Patna.
Demand forecasting based on statistical methods is used a lot by companies, due to its nature of being logical and unbiased. As it depends upon the historical data and its extrapolation, it is reliable and useful for a company. But, it lacks the dynamism associated to the business, where past may not give correct picture of the present. Secondly, it cannot be used for new products, as there will be no historical data.
Demand forecasting can either be done by time series analysis or regression analysis.
i. Time Series Analysis:
Time series is a statistical arrangement of data in a chronological order. Its’ analysis helps a company to analyse the trend in the future. The arrangement of data can be done in the form a graphical representation or it can be in the form of a table. Time series analysis helps to give us a trend, long term tendency of the data. It can also show us the seasonal variations due to festivals, holidays, weather, fashions etc.
Similarly, cyclical variations like boom, recession, depression and recovery can also help us in analysis. Sometimes unforeseen events like flood, war, epidemic, fire etc. leads us to residual variations, which can also help in demand forecasting.
Following methods can be used under time series analysis:
a. Graphical Method
b. Semi Averages Method
c. Moving Averages Method
d. Least Squares Method
ii. Regression Analysis:
Regression analysis is one of the most popular methods of demand forecasting. It analyses the average relation between two or more variables mathematically. Under simple regression, analysis is done for two variables only. While under multiple regression, it is done for multiple factors simultaneously.
This method is not only descriptive, but also prescriptive and objective in nature. It analyses time series and cross section data for more accuracy. But, when explanatory variables are not chosen realistically, it may be misleading. In the case of autocorrelation, the analysis results may be biased.
Demand forecasting for new products is more complex than the old and existing products. It requires a good understanding of the consumer per se, besides the points of differentiation from the competition. Generally it has been found that demand forecasting for new products in the year one has been off the target. It becomes more difficult, if the product or service is totally new to the market like Tata Nano, Ginger Hotels, Mahindra Scorpio, Tata Ace or say Mahindra Quanto.
The demand for new products can be estimated on the basis of old products and its data in the similar category or product segment. Another tried and tested method is the test marketing for the product to understand the interest and willingness to buy the product at a given price point.
For FMCG products, another method, which works for calculating the year one data, is the retail simulation. In retail simulation, the consumer is asked to choose the top 5 or 10 brands in a product category. Then the new product is added to the list and the feedback in the form of choice is counted. Thus survey of buyers’ intention is understood in the above way, besides the survey approach. When the product is just an improvement on existing product, then the demand can be estimated with an evolutionary approach and be projected as an outgrowth.