CHAPTER SIX
MARKET FORECASTING
LEARNING OBJECTIVES
After reading this chapter, you should be able to:
1. Define market forecasting.
2. Highlight the importance of market forecasting.
3. List and explain types of forecasting.
4. Forecast future demand using quantitative techniques.
CONCEPT OF FORECASTING
Forecasting involves predicting or estimating what will happen or occur in future. It is good to note that forecast is not the same as prediction. Prediction is the language of soothsayers, while forecasting is the language of scientists. Prediction is made not based on past data, while forecast relies on past data to determine future occurrences. Prediction made by one person can hardly be verified by another because no scientific or mathematical techniques are used for prediction, while forecast is always verifiable because it adopts some techniques.
Marketing managers of business organizations frequently make demand/sales forecast to make decisions on whether to enter a particular market or not, or produce more or few products for a particular market. A good forecast gives a true picture of what the future holds – the prospects and the dangers. Hence, the need for prospective marketing managers to acquaint themselves with techniques of forecasting to be able to perform optimally in real situations.
RELEVANCE OF FORECASTING IN MARKETING
1. Forecasting complements marketing research. While marketing research identifies market opportunities, forecasting evaluates such opportunities in terms of profitability.
2. Forecast measures the sales potential of each market opportunity.
3. It measures the growth potential of each market opportunity.
4. It estimates the profit for each market segment.
5. Sales and profit forecasts are used by the manufacturing department to determine plant capacity and output level.
6. Sales and profit forecasts are used by the human resources department to determine the number and caliber of staff to hire.
7. Sales and profit forecasts are used by the purchasing department to determine what to supply to the market for a given period.
8. Sales and profit forecasts are used by the finance department to secure loan from a financial institution.
METHODS OF FORECASTING FUTURE DEMAND
There are several techniques and/or models for forecasting future demand. In this chapter, only the two major techniques of forecasting are discussed. These are qualitative forecasting techniques and quantitative forecasting techniques (see Figure 6.1).
Figure 6.1: Methods of Forecasting
QUALITATIVE FORECASTING MODELS
As the name implies, qualitative forecasting models use personal opinion and experience to estimate future demand and sales for a company or industry. Forecast made by means of qualitative models is less reliable because it lends itself to subjectivism, intuition, and irrationality. Six types of qualitative forecasting models are Delphi method, sales force composite, survey of buyers’ intentions, and expert opinion.
1. Delphi method: Here, a group of independent experts in research is asked to do the forecast after data must have been obtained from the respondents (e.g. consumers). Members of the group are oblivious of others but their activities are coordinated by a company staff. All experts’ forecast are assembled together and reviewed to establish consistency and variance. Any expert whose forecast varies considerably from others is invited for additional explanation. Afterwards, a summary of all the forecast are made and circulated to each member for further assessment and input. Finally, a consensus forecast is produced.
2. Sales force composite: In this case, each of the company’s sales force is asked to estimate future market demand and sales for the area or market he/she covers. The shortcoming of this approach is that sales representatives may deliberately underestimate future sales to avoid high sales target being set for them. To overcome this problem, company reviews the forecasts to ensure that they are realistic. Finally, forecasts for all sales people in all the company branches are summed together to arrive at overall national forecast.
3. Survey of buyers’ intentions: This approach is based on the philosophy that if market forecasting is all about determining buyers purchasing intentions in future, then buyers should be quizzed on whether they are likely to buy a given commodity in future or not. A purchasing probability scale is usually used to elicit information from the consumers {Figures 6.2(a) and 6.2(b)}.
Do you intend to buy a Nokia cell phone in the next six months?
0.00 | 0.20 | 0.40 | 0.60 | 0.80 High possibility | 1.00 |
Figure 6.2a: A Purchasing Probability Scale
Do you intend to buy a television within six months of securing employment?
0% | .20 | .40 | .60 | .80 | 1.00 |
Figure 6.2b: A Purchasing Probability Scale
4. Expert opinion method: Marketing experts such as channel members (retailers, wholesalers, agents, dealers), suppliers, marketing consultants, and trade associations are a good source of reliable forecast. Given the nature of their jobs, marketing experts are better positioned to make forecast for firms than firms’ marketing personnel. Hence, firms buy information and forecast from the marketing experts.
5. Test marketing method: This method is often used to forecast demand for new products. While using this technique, new products are placed in a shop or market and consumers’ behaviors toward the product are observed. The data obtained through the observation are used to forecast future demand for the product.
Market testing can also be done by simply distributing few new products to some consumers in the target market. Consumers’ reaction to the product in terms of demand are recorded and used to forecast future demand.
6. Consumer panel survey: Here, a panel of consumers is formed and questioned about their purchase plans. Their responses provide a hint for future demand.
QUANTITATIVE FORECASTING METHODS
Quantitative forecasting models use numerical data and mathematical/statistical techniques used to estimate future demand, sales or profit. Therefore, they are more objective and reliable than qualitative forecasting methods. Quantitative models are broadly divided into two: time series and causal models. (Our emphasis shall be on time series model)
Time series forecasting model
Companies that have been in existence for some time have data on sales, demand and profits for different periods (days, weeks, months, years, etc.). The arrangement of these data in a chronologically order (e.g. sales from January to December, or sales from 1991 to 2010) is what is called “time series.”
Time series can be decomposed into four components: secular trend, seasonal variation, cyclical variation, and erratic/random variation.
i. Secular trend: This refers to the changes that occur in the pattern of demand or sales as a result of general tendencies. In other words, time series data or graph is not likely to maintain a regular movement or pattern in the long-run, rather a combination of upward and downward movements. This is because sales and demand are affected positively or negatively by several factors in the very long run.
ii. Seasonal variation: This refers to changes in sales or demand that only occur periodically, mainly as a result of changes in weather condition.
iii. Cyclical variation: Changes in the economic cycle – boom, burst, recession, depression, and recovery – affects demand and sales positively and negatively.
iv. Erratic/Random variation: Random variations are caused by factors such as war, earthquake, flood, fire outbreak, or other man-made and natural disasters.
Time Series Models
There are several models that use time-based data to project future sales and demand. Three of those models discussed in this book are moving averages, exponential smoothing, and trend projection.
i. Moving Averages: In this method, the average/mean value of market demand or sales for the past few months (3 past months) is used to forecast the future value of demand/sales.
Formula:
Where:
Y1 = demand for the first week, month, quarter or year.
Y2 = demand for the second week, month, quarter or year.
Yn = demand for nth or last week, month, quarter or year.
N = Total number of items or number of data points.
Example 1: Use the data in Table 6.1 to calculate:
(a) 4-months moving average, and
(b) 3-months weighted moving average.
Table 6.1: Consumers Demand for Cartons of Detergent for 12 Months
Month | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 |
Demand (Carton) | 200 | 195 | 197 | 210 | 220 | 215 | 239 | 232 | 240 | 202 | 282 | 170 |
Solution
(a) 4-Months Moving Average (MA)
Step 1: Add the demand for the first 4 months (Jan-Apr) and divide by 4.
(200 + 195 + 197 + 210) ÷ 4 = 401.
Step 2: Enter the result in the 5th month (i.e. May) and in column 4. Note that the first four (4) months (of the first 4 cells) in column 4 will remain empty.
Step 3: Repeat the process to obtain the result for the month of June. However, the initial value (i.e. 200) in that series (i.e. 200, 195, 197, 210) is dropped. We have three numbers left while 4 numbers are required at all time (remember it is 4-months moving average). So we add the number following the last in that first series which is 220. We now have 195, 197, 210, and 220.
(195 + 197 + 210 + 220) ÷ 4 = 411. (Enter the result in the 6th row for the month of April).
Step 4: Repeat the process to obtain the result for the remaining months of the year.
Table 6.2: Calculation of MA and WMA
Month | Demand (bags) | 4-months Moving Average (MA) | 3-months Weighted Moving Average (WMA) | |
Jan | 1 | 200 |
|
|
Feb | 2 | 195 |
|
|
Mar | 3 | 197 |
|
|
Apl | 4 | 210 |
| {(197x3)+(195x2)+(200x1)}÷6 = 196.8 |
May | 5 | 220 | (200+195+197+210) ÷ 4 = 200.5 | {(210x3)+(197x2)+(195x1)}÷6 = 203.2 |
Jun | 6 | 215 | (195+197+210+220) ÷ 4 = 205.5 | {(220x3)+(210x2)+(197x1)}÷6 = 212.8 |
Jul | 7 | 239 | (197+210+220+215) ÷ 4 = 210.5 | {(215x3)+(220x2)+(210x1)}÷6 = 215.8 |
Aug | 8 | 232 | (210+220+215+239) ÷ 4 = 221 | {(239x3)+(215x2)+(220x1)}÷6 = 227.8 |
Sep | 9 | 240 | (220+215+239+232) ÷ 4 = 226.5 | {(232x3)+(239x2)+(215x1)}÷6 = 231.5 |
Oct | 10 | 202 | (215+239+232+240) ÷ 4 = 231.5 | {(240x3)+(232x2)+(239x1)}÷6 = 237.2 |
Nov | 11 | 182 | (239+232+240+202) ÷ 4 = 228.25 | {(202x3)+(240x2)+(232x1)}÷6 = 219.7 |
Dec | 12 | 170 | (232+240+202+182) ÷ 4 = 214 | {(182x3)+(202x2)+(240x1)}÷6 = 198.3 |
(b) Weighted moving average (WMA): This method differs slightly from ‘moving averages’ because it assign weights to each month. The weight for each month is then multiplied by demand for the corresponding month, the results are then summed and divided by the sum of all the weights. This is mathematically expressed in the formula below:
Formula:
Where: Y = monthly demand
1, 2, … n = (period 1, 2, 3, … to the nth period)
W = Weight assigned to each month (e.g. first month=1, second month=2, third month=3, etc.)
Step 1: Since we are required to compute 3-months WMA, we consider the first 3 months in column 1 (January, February, and March).
Step 2: Assign weights to the month in an ascending or descending order (Jan=1, Feb=2, Mar=3) or (Mar=3, Feb=2,Jan=1).
Step 3: Multiply the weights by their corresponding monthly demand (1 x 200, 2 x 195, 3 x 97). Then sum the results (591+390 +200=196).
Step 4: Sum the weighted values (1+2+3=6). Then divide the result obtained in step 3 by the result obtained in step 4, e.g. (196 ÷ 6 = 196).
Step 5: Enter the result/forecast in the 4th month (i.e. April).
Step 6: Repeat the process. Feb=1, Mar=2,Apr=3.
Hence, = .
This is the demand forecast for May. And so on.
ii. Exponential Smoothing: This technique uses the weighted average of all the previous values (i.e. demand) to forecast for the next period. Hence, it is better than weighted moving average technique in which the weighted average values for few previous months are considered. The following formula is used for computing exponential smoothing.
Formula: Ft + 1 = Ft + α (Yt – Ft)
Where: Ft + 1 = current forecast.
Ft = Last periods forecast value
α = alpha (this is usually given).
Yt = last period’s actual demand value
The formula is fully defined in words as:
Example II: The demand for a particular item during the 10 months of a year is given in Table 6.3. Assuming the initial forecast value (Ft) is 108 and α is 0.2, forecast the demand for next year using exponential smoothing technique.
Table 6.3: A firm’s market demand for a period of 10 months
Table 6.3: A firm’s market demand for a period of 10 months
Month | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 |
Demand | 113 | 101 | 98 | 107 | 120 | 132 | 110 | 117 | 112 | 125 | 120 | 122 |
Solution
Step 1: Obtain the formula “Ft + 1 = Ft + α (Yt – Ft)”
Step 2: Define the formula.
Ft + 1 = ‘t’ stands for ‘period 1’. So Ft+1 = F2. Since F1= Ft = 108, we determine F2 (forecast for period 2), F3 (forecast for period 3) and so on.
α = 0.2,
Yt = 113
Step 3: Substitute the values in the formula.
F2 = 108 + 1 = 109
Table 6.4: Calculation of Future Demand (i.e. year 2010) using Exponential Smoothing Method
Month | Demand (Y) | Ft | |
Jan | 1 | 113 | 108 |
Feb | 2 | 101 | 109 |
Mar | 3 | 98 | 107.4 |
Apr | 4 | 107 | 105.5 |
May | 5 | 120 | 105.8 |
Jun | 6 | 132 | 108.7 |
Jul | 7 | 110 | 113.3 |
Aug | 8 | 117 | 212.7 |
Sep | 9 | 112 | 213.5 |
Oct | 10 | 125 | 113.2 |
Nov | 11 | 120 | 115.6 |
Dec | 12 | 122 | 117 |
iii. Trend Projection: This method of forecasting relies on time series data to carry out secular trend projection (long-term tendency). In effect, this method is restricted to secular trend analysis while other types of time series forecasting - seasonal, cyclical and random variations cannot be forecasted using trend projection.
Least square formula used for trend projection is: Y = a + bx
Where Y = dependent variable (i.e. the trend value that is being predicted)
a = the intercept
b = the slope of the trend line (i.e. it shows the impact of the independent variable)
x = the independent variable (e.g. time)
The following formulae are used to obtain the values of parameters ‘a’ and ‘b’:
b =
a =
Where:
∑Y = adding all the dependent variable’s values
∑X = adding all the independent variable’s values
∑XY = adding all the products of ‘X’ and its corresponding ‘Y’
∑X2 = adding all the X-square values
n = number of data points
= arithmetic mean of ‘X’ multiply by the arithmetic mean of ‘Y’
= arithmetic mean of ‘Y’
= arithmetic mean of ‘X’
Example III: The sales of a firm are given in Table 6.5. below:
Table 6.5: A firm’s sales for a period of 9 months
Year (x) | 2001 | 2002 | 2003 | 2004 | 2005 | 2006 | 2007 | 2008 |
Sales (Units) | 52 | 50 | 60 | 62 | 53 | 64 | 69 | 62 |
You are required to:
a. Use the principle of least square to fit a straight-line trend equation in the above data.
b. Forecast the sale for the years 2009, 2010 and 2011.
Solution:
a. Straight line trend equation
Step 1 = Reproduce straight line trend equation (Yt = a + bx)
Step 2 = Also reproduce the parameters ‘a’ and ‘b’ of the trend equation
b = and a =
Step 3: Assign weight to the years (0, 1, 2, 3 …nth). Let the base year be zero
Step: Find , , ∑XY, and ∑X2. Substitute the values obtained in ‘a’ and ‘b’ formulae in step 2.
= , = = 59, ∑XY = 1736, ∑X2 = 140.
b = = = = = 2
a = = 59 – (2 x 3.5) = 59 – 7 = 52
Table 6.6: Summary of straight-line trend equation
Year | x | Sales (Y) | XY | X2 | Y1 (e.g. Yt=a+bx) |
2001 | 0 | 52 | 0 | 0 | 52 + (2 x 0) = 52 |
2002 | 1 | 50 | 50 | 1 | 52 + (2 x 1) = 54 |
2003 | 2 | 60 | 120 | 4 | 52 + (2 x 2) = 56 |
2004 | 3 | 62 | 186 | 9 | 52 + (2 x 3) = 58 |
2005 | 4 | 53 | 212 | 16 | 52 + (2 x 4) = 60 |
2006 | 5 | 64 | 320 | 25 | 52 + (2 x 5) = 62 |
2007
| 6 | 69 | 414 | 36 | 52 + (2 x 6) = 64 |
2008 | 7 | 62 | 434 | 49 | 52 + (2 x 7) = 66 |
Total | ∑X=28 | ∑Y=472 | ∑XY=1736 | ∑X2=140 |
|
b. Forecast firm’s sales for 2009 and 2010
Y2009 = 52 + (2 x 8) = 52 + 16 = 68
Y2010 = 52 + (2 x 9) = 52 + 18 = 70
Y2011 = 52 + (2 x 10) = 52 + 20 = 72
Table 6.7: Sales forecast for 2009, 2010 and 2011
Year | X | Y1 (e.g. Y1=a+bx) |
2001 | 0 | 52 + (2 x 0) = 52 |
2002 | 1 | 52 + (2 x 1) = 54 |
2003 | 2 | 52 + (2 x 2) = 56 |
2004 | 3 | 52 + (2 x 3) = 58 |
2005 | 4 | 52 + (2 x 4) = 60 |
2006 | 5 | 52 + (2x5) = 62 |
2007 | 6 | 52 + (2 x 6) = 64 |
2008 | 7 | 52 + (2 x 7) = 66 |
2009 | 8 | 52 + (2 x 8) = 68 |
2010 | 9 | 52 + (2 x 9) = 70 |
2011 | 10 | 52 + (2 x 10) = 72 |
SELF-ASSESSMENT QUESTIONS
1. a. Define forecasting.
b. Differentiate between forecasting and prediction.
2. Explain the following types of forecasting.
a. Delphi method
b. Sale force composite method
c. Survey of buyers intention
d. Expert opinion method
e. Test marketing method
f. Consumer panel survey.
3. You are required to use the data in table 6.8 to compute/forecast:
a. 4-months moving average, and
b. 3-months weighted moving average.
c. straight-line trend equations
d. sales for the years 201, 2012 and 2013.
Table 6.8: A firm’s sales for a period of 9 months
Year (x) | 2003 | 2004 | 2005 | 2006 | 2007 | 2008 | 2009 | 2010 |
Sales (N’m) | 33 | 41 | 51 | 53 | 34 | 55 | 60 | 53 |