Demand Forecasting is a function to estimating the quantity of a product or service that customer will purchase. The purpose of demand forecast is to avoid overproduction and underproduction.
Types of Demand Forecasting
Short term forecast used to improve raw material procurement and production scheduling. It is for daily and weekly scheduling.
Medium Term Forecast
Medium term forecast used for budgeting such as production planning, purchasing, distribution and etc. It related tactical, monthly, decisions and done for a month.
Long Term Forecast
Long term forecast used for business strategy and strategic management decisions sussed such as capacity planning, facility location, and strategic planning. It is done more than a year.
Factor of Select Demand Forecasting Method
Time Span
The choice of an appropriate forecasting method is affected by the nature of the production resource that is to be forecasted. For example, cash, workers, inventories are short term in nature and can be forecasted with exponential smoothing or moving average models. For long term production resource needs capital funds and factory capacities can be estimated by regression, market research and etc.
Data Available
The data available and relevant for forecasting is an important factor. For example, a survey of customer maybe an appropriate method for developing demand estimates if the intentions of customer and attitudes are a relevant factor in forecast. On the other hand, a survey of customers may not be a practical way to develop a forecast if the requirement is to forecast sales of a new product.
Cost & Accuracy
More forecast accuracy can be obtained at a cost. Normally, high accuracy approach uses more data to support it. If the data are more difficult to obtain then the models are more costly to design, operate, and implement. For example, statistical models, complex econometric models, and historical models. Every organizations basic on own situation to make the cost and accuracy trade-off.
Nature of Product and Service
Different forecasting methods for different products. For example, whether the product is a manufactured good or service, whether product is high volume and high cost, where the product is in its life cycle all affect the choice of a forecasting method.
Benefit of using Inventory Demand Forecasting System
Reduce Cost & Increase Profit
Inventory Demand Forecasting System can reduce shipping cost, maintenance cost, and etc. Shipping fee of board game in large quantity is cheaper than ship one by one. If customer wants to buy the new board game and no stock in the cafe then manager need order new board game from European supplier. It will increase our cost because shipping cost is higher compare with shipping large quantity. Besides that, it can also increase our profit because we would not have over quantity in our stock.
Improve Customer Service
Inventory Demand Forecasting System can improve customer service because this system can forecast how many quantity of product needed. It can reduce the situation that no enough stock to supplier to customers. Every board game products are import from European county, it may take 2 to 4 weeks delivery. If customer wants to purchase new board game then he/she no need book the board game and wait 2 to 4 weeks to delivery. Some of the customer maybe will change his/her decision within this few weeks.
Increase Productivity
Inventory Demand Forecasting System will directly show how many quantity of board game should order /purchase because the result will basic on the historical data to forecast the demand quantity for each product. When manager order/purchase the new board game, he/she can basic on the demand quantity to make the decision which board game want to purchase and how many order quantity. It can also save a lot of time to do the research and compare the product when place a purchase order.
Methods of Inventory Demand Forecasting
There are two types of Demand Forecasting Methods:
Qualitative Methods (Survey Methods)
Qualitative methods are mainly basic on educated opinions, workforce experience, and surveys. Besides that, it cans also using simple mathematical tools to combine different forecasts. It is usually used for short-term forecasts. For example, new product/service is launched on the market, changing product packing, or future demand pattern is expected to be affected by political changeovers. The most widely used qualitative methods are Expert Opinion Method and Consumer Survey Method. (Gianpaolo Ghiani, Gilbert Laporte, and Roberto Musmanno, 2006)
Expert Opinion Method
Expert Opinion Method is essentially based on the opinion of experts, either internal or external to the firm. The most widely used method is Delphi Method. (Kerstin Cuhls, 2006)
Delphi Method
Delphi Method relies on a pool of experts that an expert could be an ordinary employee, decision maker, or industry expert. The expert do not interact face-to face because they will answer questionnaires to estimate of the demand. After that, a vacillator basic on the questionnaires to provides an anonymous summary of the experts' forecast with the reasons they provided for their judgments. An iterative process is conducted until a consensus is reached by all. (Kerstin Cuhls, 2006)
There are advantages of Delphi Method:
Delphi Method does not require face-to-face meetings because it is conducted in writing. The responses of demand forecasting can be made at the convenience of the experts. Besides that, experts from different locations or backgrounds can work together on the same problems. It is relatively free of individual dominance, social pressure, and personality influence.
Delphi Method also allows experts provide a broad range of opinion that base analysis "two heads are better than one". All participants shared their information, and experience. An iteration process enables participants to re-evaluate, review, and revise all their previous judgement in light of comments made by their peers. Besides that, Delphi Method is inexpensive because it can save corporations money in travel expenses and it can use by email. (Roberts Evalution, 2010)
There are disadvantages of Delphi Method:
Delphi Method is time consuming to manage and coordinate because information comes from a selected group of experts and may not be representative.
The decision-making process can be more easily influenced by the coordinator, and it is less transparent than face to face meetings. Therefore, it can lead to less trust in the outcome and process by experts. (Roberts Evalution, 2010)
Organisation of a Delphi survey
Consumer Opinion Survey Method
Consumer Opinion Method is take opinion of the users or buyers of the product because they belief this is the best and most obvious way to gauge the demand for a commodity. This method is for short-term projections. The users or buyers can give their opinions about the particular product. The questionnaire must be carefully prepared bearing in mind the qualities of a good questionnaire with simple and interesting to evoke consumer response. (Geetika, Piyali Ghosh, Purba Roy Choudhury, 2005)
A consumer survey can be conducted into two ways:
Complete Enumeration Survey Method
This method is based on a complete survey of all the consumers. In questionnaire, consumers are asked about the quantity of the commodity they would like to buy in the forecast period. All the data is collected and added up to arrive at the total expected demand for that product.
DF = (ID1 + ID2 + ID3 + ... + IDn)
Where DF = Demand Forecast for all consumers
ID = Intended demand of consumer.
The advantages of Complete Enumeration Survey Method are quite accurate of demand forecast as it surveys all the consumers of a product. It is also simple to use and not affected by personal biases.
The disadvantages of Complete Enumeration Survey Method are it is costly and time consuming to manage. Besides that, it is difficult and practically impossible to survey all the consumers. The size of the data increases the chances of faulty recording and wrong interpretation. (Geetika, Piyali Ghosh, Purba Roy Choudhury, 2005)
Sample Survey Method
Sample Survey Method is select few consumers to represent the entire population of the consumers and their views on the probable demand are collected. The demand of the sample so ascertained is magnified to generate the total demand of all the consumers for that commodity in the forecast period.
DF = (ID1 + ID2 + ID3 + ... + IDn)N/n
Where,
DF = Demand Forecast for all consumers
ID = Intended demand of consumer.
N = population of consumers
n = sample picked up
The advantages of Sample Survey Method are it is simple and does not cost mush to manage. Besides that, this method can work more quickly because only few consumers are to be approached. The risk of the erroneous data is reduced.
The disadvantages of Sample Survey Method are the results based on the view of few consumers. So, the sample may not be true representation of the entire population. (Yogesh Maheshwari, 2005)
Quantitative Methods (Statistical Methods)
Quantitative methods are mainly basic on sufficient historical demand or relationships between variable to generate simulation models or mathematical. It is usually used for medium-term or long-term forecasts. The most widely used quantitative methods are Mechanical Extrapolation, Barometric Techniques, and Regression Method. (Gianpaolo Ghiani, Gilbert Laporte, and Roberto Musmanno, 2006)
Mechanical Extrapolation (Trend Projection Method)
This technique based on analysis of past sales patterns/historical data to predict the demand for a commodity in the future. These methods need for costly market research because the historical data is a time series data such as company files interns of different time periods. (Yogesh Maheshwari, 2005)
There are two main techniques of mechanical extrapolation:
Least Squares Method
Least Squares Method used statistical formula to find the trend line also called "best fit" of the available data. It can be used for forecasting demand basic on the corresponding values of product/service on the graph and extrapolate the line for future demand. (AR Aryasri, 2007)
There is the formula of least squares method (trend equation):
Sales = x+y(T)
Where,
x & y have been calculated from past data sales.
T = year number for which the forecast is made.
To determine the values of x & y, there are two normal equations need to be solved:
∑S = Nx + y∑T
∑ST = x∑T + y∑T2
Where,
S = Sales
T = Year Number
N = Number of Years/Months for which data is available
Example case: Basic on the annual sales data of company A to estimate sales for the year by using Least Squares Method.
Year
2005
2006
2007
2008
2009
Sales (in 000's)
45
56
58
46
75
Solution:
Step 1: Create a table to determine N, ∑S, ∑ST, ∑T, and ∑T2
Year
Sales (S)
T
T2
ST
2005
45
1
1
45
2006
56
2
4
112
2007
58
3
9
174
2008
46
4
16
184
2009
75
5
25
375
N = 5
∑S = 280
∑T = 15
∑T2 = 55
∑ST = 890
Step 2: Basic on example case to substituting the above value in the normal equations.
∑S = Nx + y∑T
∑ST = x∑T + y∑T2
280 = 5x + 15 y
890 = 15x + 55y
Step 3: After solving these equations, we get x = 41, y = 5, and T = 6 because year 2010 on the year number is 6. By substituting these value in the trend equation Sales = x+y(T).
Sales (2010) = 41 + 5(6)
Thus, the forecast sale for year 2010 is 71000.
Moving Average Method
Moving Average Method is average of historical data determined the forecast. The average of this method keep on moving depends on the number of years selected. We can use this method to eliminate the effect of seasonality trend of sales. The advantages of moving average method are easy to compute and after computed, old data can be dispensed. (N. Kumar and R. Mittal, 2001)
There is the concept of moving average method:
Example case: Basic on daily sales data to compute 3-day moving average.
Date & Month
Daily Sales (in 000's)
3-day Moving Average
1 Jan
40
2 Jan
44
3 Jan
48
4 Jan
45
44
5 Jan
53
45.7
Solution: Calculate 3-day moving average
Sales (4 Jan) = (40+44+48)/3 = 44
Sales (5 Jan) = (44+48+45)/3 = 45.7
Regression Method
Regression Method used to estimate the value of one variable from assumed values of other variable. It can describe the relationship between the variable being forecast and other variables to determine the 'best fit' expression. The advantage of regression method is based on causal relationship to produces reliable and accurate results. The disadvantages of regression method are uses complex calculations, costly, and time consuming. (N. Kumar and R. Mittal, 2001)
The relevant equation of regression method:
Dx = a + bPx + cI + dA - ePy
Where,
a, b, c, d, & e = constants.
Dx = demand for X
Px = price of X
I = Consumer's income
A = Advertisement outlay
Py = Price of its substitute product Y.
Example Case: Basic on price and quantity data of pens sold by a company to estimate the demand for pen when price RM7 per pen.
Year
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Price(RM)
2
1
2
4
5
4
3
8
6
5
Quantity
('000 units)
9
10
8
7
5
6
8
3
4
7
Solution:
Step 1: Taking price and quantity as variable X and Y and tabulating them for calculations we get.
X (Price)
Y (Quantity)
X2
Y2
XY
2
9
4
81
18
1
10
1
100
10
2
8
4
64
16
4
7
16
49
28
5
5
25
25
25
4
6
16
36
24
3
8
9
64
24
8
3
64
9
24
6
4
36
16
24
5
7
25
49
35
40
67
200
493
228
Step 2: Use least square method to fit the regression line of the form
Y = a + bX
The set of normal equations for this are
∑Y = na + b∑X
∑XT =a∑X + b∑X2
Substituting the values of the variable
67 = 10a + 40b
228 = 40a + 200b
Step 3: After solving these equations, we get a = 10.7, b = -1. Thus,
Y = 10.7 - X
Hence, the regression line is
Q = 10.7 - P
When, P = the demand is
Q = 10.7 - 7 = 3.7
That is demand
Q = 3700 units.
Steps to Select Method of Inventory Demand Forecasting
Identification of objective
Each demand forecasting would be different objective. We should clear about the uses of forecast data and how it related to forward planning by the company. For example, estimation of quantity and composition of demand, inventory control and etc. (MBA Knowledge Base, 2010)
Select the product to be forecast
After identified objective, the next step is select which product to be forecast. We should clearly to identify nature of product which examine whether the product is consumer products or producer products, final or intermediate demand, new demand or replacement demand type and etc. (MBA Knowledge Base, 2010)
Determine the type of demand forecasting
The next step is determining the type of the forecast for the product. Since we had identified objective of demand forecasting and nature of product, we can depend to identify which type of demand forecasting need to select such as short-term, long-term or medium-term. (MBA Knowledge Base, 2010)
Select the demand forecasting method
Select the forecasting method is a very import step. We need to choose a particular demand forecasting method from among various methods. We need to expose to all methods because different methods maybe appropriate for forecasting method for different product depending upon their nature and objective of the demand forecast. It may be possible to use more than one method in some case. However, we need to be logical and appropriate to select method of demand forecasting. (MBA Knowledge Base, 2010)
Validate and implement results
This is the final step of demand forecasting. We need to basic on the data and apply it into the formula. After that, we need to testing accuracy of the result. (MBA Knowledge Base, 2010)
Implementation of Inventory Demand Forecasting System
Product Purchase Quantity
Meeples BoardGame Cafe has many board game products and each of the product purchase from European country. When manager place a purchase order with an overseas suppliers, it may take 2 to 4 weeks for supplier delivery the products. During these periods, it might happen some accidently things such as supplier delay in delivering order, products missing and etc. Therefore, it should have enough inventories provide to customer. This is a purpose of forecast inventory order quantity.
Product to be forecast is board game that quantity on hand equal reorder level or high demand board game. Manager will purchase new board game every month. So, this is a medium-term forecasting.
Basic on this situation, it suitable using Moving Average Method to calculate the purchase quantity of product basic on last 3 month sales data. This method is easier to use and calculate and no need to keep the old data. There is the example of Meeples BoardGame Cafe:
Month
Monthly Sales (units)
Jan
10
Feb
20
March
15
Solution:
Sales (April) = (10+20+15)/3
= 15 units.
New Product
Supplier will selling the new board games every month or few months. But, not all the product is suitable for the customers of Meeple BoardGame Cafe. Some of category of games is the most famous in Meeple BoardGame Cafe. For example, memory games, war games, and etc. The purpose of forecasting demand of new product is avoiding order the wrong product or no enough stock to sell to customers.
Product to be forecast is new product selling by supplier or product new launch in the market. This is the short-term forecast.
Basic on this situation, it suitable use Sample Survey Method to calculate demand forecast of new product because new product does not has historical data to calculate the demand forecast. So, we must the qualitative method. There is the sample of Meeples BoardGame Cafe by using this formula:
DF = (ID1 + ID2 + ID3 + ... + IDn)N/n
Where,
DF = Demand Forecast for all consumers
ID = Intended demand of consumer.
N = population of consumers
n = sample picked up
Intended demand of customer (units)
Customer A
1
Customer B
2
Customer C
0
Customer D
1
Solution:
DF = (1+2+0+1)10/4
= 10 units
Product Price Change
Sometimes some of the product will change the selling price. The selling price change will affect the sales of the product also. So, we need the forecast the demand of the product after selling price is change. It can avoid ordering the wrong quantity of product or because of the price change and affecting the sales of the product.
Product to be forecast is product that changes the selling price. This is the long-term forecasts. There is the example of Meeple BoardGame Cafe basic on the historical data by using the formula:
Dx = a + bPx + cI + dA - ePy
Where,
a, b, c, d, & e = constants.
Dx = demand for X
Px = price of X
I = Consumer's income
A = Advertisement outlay
Py = Price of its substitute product Y.
Year
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Price(RM)
2
1
2
4
5
4
3
8
6
5
Quantity
9
10
8
7
5
6
8
3
4
7
Solution:
Step 1: Taking price and quantity as variable X and Y and tabulating them for calculations we get.
X (Price)
Y (Quantity)
X2
Y2
XY
2
9
4
81
18
1
10
1
100
10
2
8
4
64
16
4
7
16
49
28
5
5
25
25
25
4
6
16
36
24
3
8
9
64
24
8
3
64
9
24
6
4
36
16
24
5
7
25
49
35
40
67
200
493
228
Step 2: Use least square method to fit the regression line of the form
Y = a + bX
The set of normal equations for this are
∑Y = na + b∑X
∑XT =a∑X + b∑X2
Substituting the values of the variable
67 = 10a + 40b
228 = 40a + 200b
Step 3: After solving these equations, we get a = 10.7, b = -1. Thus,
Y = 10.7 - X
Hence, the regression line is
Q = 10.7 - P
When, P = the demand is
Q = 10.7 - 7 = 3.7
That is demand
Q = 3700 units.
Conclusion
