You are on page 1of 23

Chapter 5

Forecasting

Key Terms:

Forecast

Qualitative Method

Quantitative Method

Delphi Method

Jury of Executive Opinion

Sales Force Composite

Consumer Market Survey

Time Series

Time Series Method

Trend

Seasonality

Cycles or Cyclical Components

Random Variations/Irregular Components

Smoothing Method

Smoothing Constant

Weighted Moving Average

Exponential Smoothing

Trend Projection

Mean Absolute Deviation (MAD)

Mean Squared Error (MSE)

Mean Absolute Percent Error (MAPE)

Trend Line Forecast


Learning Objectives

At the end of the chapter, the students are expected to:

1. Learn the meaning of forecasting

2. Know the significance of forecasting in business community

3. Identify the different types of forecasting methods

4. Differentiate the qualitative forecasting method to quantitative method

5. Determine the different types/components of forecasting

6. Describe the characteristics of judgemental forecasting

7. Forecasting simple average, weighted moving and exponential smoothing

8. Calculate errors in forecasting

9. Graph different forecasting methods

10. Relate the background of other subjects such as Management, Marketing


and Statistics to forecasting

11. Solve forecasting problems using computer

Introduction

In business of any field, anticipating the needs of your clients and


serving these needs, definitely spells success. A good management team
uses past and present data, analyzes the trend of the current situation, still
relies on “gut feeling” on what is likely to happen in the future. Decisions are
made every day, and coming up with the best and strategic plan to action in
the future is an essential tool. Consequently, managers try to minimize the
“uncertainty” of the future in forecasting. A forecast simply means a prediction
of what is likely to happen in the future. Converting the prediction into
numbers and concrete data is the main purpose of forecasting.

In this chapter, several types of forecasting techniques are to be


discussed, it is but worthy to mention that, managers accept the fact that
regardless of the technique used, there is no perfect forecast. There is
however eight basic steps to follow in forecasting. These are:

1. Establish the objective of the forecast we are aiming for.


2. Choose the items/quantities to be forecasted.
3. Determine the period of time for the forecast.
4. Decide on the forecasting technique.
5. Gather the data needed.
6. Utilize the forecasting method.
7. Make the forecast.
8. Apply the forecasting result.
These steps summarize the general procedure in forecasting.

What is Forecasting?

It is the process of estimation in unknown situations. Prediction is


similar, but in a more general term.

Business forecasting is an estimate or prediction of future development


in business such as sales, expenditures, revenues, income and profits. It is
considered as one of the most important aspect of corporate planning.

Demand forecasting is a forecast that projects the company’s sales.

Classification of Forecasting

There are two classifications of forecasting methods: Qualitative and


Quantitative.

FORECASTING METHODS

Forecasting
Methods

Quantitative Qualitative
Methods Methods

Delphi
Time Series Casual
Method
Methods Methods

Jury of
Regression Executives
Moving Analysis Opinion
Average

Exponential Sales Force


Multiple Composite
Smoothing
Regressions

Consumer
Trend
Market
Projections
Survey
Qualitative Forecast

This type of forecasting method is based on judgements or opinions


and is subjective in nature. It does not rely on mathematical computations.

Types of Qualitative Methods

Delphi Method

One of the most commonly used qualitative forecasting is the Delphi


Method. Here, a group of experts separated from each other makes a
forecast. In its usual application, they are asked to answer questionnaires.
These groups are called the "respondents." The answers and other data are
gathered, collected, summarized by staff personnel and are turned over to
another group, the decision makers, or the forecasters. It is important to note
that Delphi method principle is that, responses from the first questionnaires
are used and considered in preparing the second set of questionnaires.
Responses are gathered, and the process is repeated until a consensus of
the group is reached.

Jury of Executive Opinion

Jury of Executive Opinion is a forecasting method that is recommended


when situation is not likely to repeat itself. This is a qualitative forecast that is
based on e judgment of a single expert or a consensus of the group of
experts.

Sales Force Composite

In this method, a salesperson predicts his sales in his area based on


past performance and trend. This forecast is reviewed to be sure it is realistic
and attainable.

Consumer Market Survey

This method gathers information, data from customers and prospective


customers regarding their future needs and intended purchases, thereby
improving forecasting details and product design.

Quantitative Forecast (Projective Forecast)

This type of forecasting method is based on quantitative models, and is


objective in nature. It relies heavenly on mathematical computations.
Types of Quantitative Methods

Time Series

It is an attempt to predict the future by means of historical data and a


set of ordered observations on a quantitative characteristic of an event at
equally spaced time positions.

Components of a Time Series

The pattern or behavior of the data in a time series has several


components. A time series typically has four separate components: Trend (T),
Seasonality (S), Cycles or Cyclical Components, and Random Variations (R)/
Irregular Components.

1. Trend (T) – the gradual shifting (upward or downward movement) of


the time series. The shifting or trend is frequently the result of long-
term factors such as changes in the community, demographic
characteristics of the population, technology, and consumer’s
preferences.

Figure 6.1 shows a straight line that may be a good approximation of the trend
in DVD sales. When a time series consists of random fluctuations move
around a long-term trend line, a linear equation may be used to estimate the
trend. This is shown in Figure 6.2
2. Seasonality (S) – is a pattern of the demand fluctuation above or
below the trend line that occurs every year. This is the component of
the time series that represents the variability in the data due to
seasonal influences.
3. Cycles or Cyclical Components – any recurring sequence of points
above and below the trend line lasting more than one year. These are
usually tied into the business cycle.

Fig. 6.3

Trend and Cyclical Components of a Time Series


4. Random Variations (R) – “blips” in the data caused by chance and
unusual situations and do not follow discernible pattern.

In Statistics, there are two general forms of time series models.

1. Multiplication Model – assumes that demand is the product of the four


components.

Demand = T x S x C x R

2. Additive Model – adds the components together to provide an estimate.


Demand = T + S + C + R
 If in case we do not know the value C and R, forecast becomes:

F = T + S (Additive Model)

And F = T x S (Multiplication Model)

Three more forecasting methods are to be discussed in this section.


These methods are moving averages, weighted moving averages, and
exponential smoothing. These methods “smooth out” random fluctuations are
caused by irregular components of the time series. They are best used for
stable time series where no movement in trend is expected.
Naïve Forecast

Naïve forecast is the simplest technique. It simply uses the actual


demand for the past period as the forecasted demand for the next period. This
makes the theory that the past will repeat. It also assumes that any time
series components (trends, seasonality or cycles) are either reflected in the
previous period’s demand or do not exist. An example of naïve forecasting is
illustrated in Table 6.1.

TABLE 6.1

Naïve Forecasting
Period Demand Forecast
1 35
2 40 35
3 55 40
4 65 55
5 60 65

*Notice the demand on Period 1 was 3; the naïve forecast for the
upcoming period is 35.

Moving Averages

These use the average of the most recent data values in the time
series forecast for the next period.

Formula:

∑ (most recent n data values )


Moving Average = n (6.1)

Where n is the number of period in the moving average.


Example 1.
Compute for a four-month moving average using the data given in
Table 6.2.
TABLE 6.2

Four-Month Average Forecast

Period Actual Demand Forecast

January 21

February 25

March 29

April 21
(21+25+29+ 21)
May 25 =24
4
(25+29+ 21+ 25)
June 20 =25
4
(29+21+25+ 20)
July 18 =24
4
(21+25+20+ 18)
August 21 =21
4
(25+20+ 18+21)
September 20 =21
4
(20+18+ 21+ 20)
October 19 =20
4
(18+21+20+19)
November 18 =20
4
(21+20+19+18)
December 15 =20
4

Weighted Moving Averages


This is a smoothing method that uses a weighted average of the recent
n data as the forecast.
Formula:
Weighted Moving Average =
Σ(weight for period n)(demand∈ period n)
(6.2)
Σweights
Example 1:

Compute for a three month weighted moving average (Table 6.3).

TABLE 6.3

Three-Month Weighted Moving Average Forecast


Month Demand Three-Month Weighted
Moving Average
January 21
February 25
March 29
April 21 [ ( 29 ×3 ) + ( 25 ×2 ) +21]
=26
6
May 25 [ ( 21× 3 ) + ( 29 ×2 ) +25]
=24
6
June 20 [ ( 25 ×3 ) + ( 21 ×2 ) +29]
=24
6
July 18 [ ( 20 ×3 ) + ( 25 ×2 ) +21]
=22
6
August 21 [ ( 18× 3 ) + ( 20 ×2 ) +25]
=20
6
September 20 [ ( 21× 3 ) + ( 18 ×2 ) +20]
=20
6
October 19 [ ( 20 ×3 ) + ( 21 ×2 ) +18]
=20
6
November 18 [ ( 19× 3 ) + ( 20 ×2 ) +21]
=20
6
December 15 [ ( 18× 3 ) + ( 19 ×2 ) +20]
=19
6
Example 2:

Compute a four-period weighted moving average forecast from Table 6.4

Table 6.4

Four-Period Weighted Moving Average Forecast


Demand Supply Forecast
1 35
2 50
3 65
4 55
5 70 [ ( 55 x 4 ) + ( 65 x 3 ) + ( 50 x 2 ) +35 ]
=55
10
6 75 [ ( 70 x 4 ) + ( 55 x 3 ) + ( 65 x 2 ) +50]
=63
10

Exponential Smoothing

This is a forecasting method that is a combination of the last forecast


and the last observed value. It uses a weighted average of past time series
value as the forecast and is based on the idea that as data gets older it
becomes less relevant and should be given less weight.

Formula:

New forecast = Last period’s forecast + α (last period’s demand – last


period’s forecast)

Mathematically:

F t=F t−1 +α ( At −1−Ft −1)

Where:

F t = new forecast

F t−1 = previous forecast

α = smoothing constant that has a value between 0 and 1 (the Greek


letter α pronounced “alpha”)

At −1 = previous period’s actual demand

Example 1:
In January, a demand for 200 units of Toyota car “Vios” for February
has predicted by a car dealer. Actual February demand was 250 cars.
Forecast the March demand using the smoothing constant of α=0.30.

New Forecast (March demand) = 200 + 0.30(250 – 200)

= 200 + 0.30(50)

= 200 + 15

= 215 cars

Therefore, the demand forecast for the Toyota “Vios” in March is 215.

Example 2:

Use exponential smoothing to compute for a series of forecast with


smoothing constant of:

a. α = 0.20,
b. α = 0.50, and
c. plot the actual data and both sets on single graph.

Table 6.5

Period Demand
1 20
2 35
3 46
4 40
5 50
6 55
7 45
8

 Forecast Error = Actual Demand – Forecast


 The forecast error tells us how well the method are performed
against themselves using past data.

Solution:

a. α = 0.20

Table 6.6

Period Demand Forecast Error


1 20 -
2 35 20 15
3 46 23 23
4 40 27.60 13.60
5 50 30.08 19.92
6 55 34.06 20.94
7 45 38.25 6.75
8 39.60

F 3=20+0.20 ( 35−20 )
F 3=23

F 4=23+ 0.20 ( 46−23 )


F 4=27.60

F 5=27.60+0.20 ( 40−27.60 )
F 5=30.08

F 6=30.08+0.20 ( 50−30.08 )
F 6=34.06

F 7=34.06+0.20 ( 55−34.06 )
F 7=38.25

F 8=38.25+0.20 ( 45−38.25 )
F 8=39.60

b. α = 0.50

Table 6.7

Period Demand Forecast Error


1 20 -
2 35 20 15
3 46 27.50 18.50
4 40 36.75 3.25
5 50 38.38 11.62
6 55 44.19 10.81
7 45 49.59 -4.59
8 47.30

F 3=20+0.50 ( 35−20 )
F 3=27.50

F 4=27.50+ 0.50 ( 46−27.50 )


F 4=36.75

F 5=36.75+0.50 ( 40−36.75 )
F 5=38.38

F 6=38.38+0.50 ( 50−38.38 )
F 6=44.19

F 7=44.19+ 0.50 (55−44.19 )


F 7=49.59

F 8=49.59+ 0.50 ( 45−49.59 )


F 8=47.30

c. Graphs at α = 0.20 and α = 0.50


60

50

40
Demand

30
α = 0.20
Actual Demand
20

10

0
1 2 3 4 5 6 7 8
Period

60 Period

50

40
Demand

30 α = 0.50
Actual Demand
20

10

0
1 2 3 4 5 6 7 8

 Other methods for determining and measuring the accuracy of forecast


error are MAD, MSE, MAPE and BIAS.

Mean Absolute Deviation (MAD) is a technique for determining the


accuracy of a forecasting model by taking the average of the absolute value.

Example 1. ( Forecast errors )


MAD= (6.4 )
n
Compute the MAD from Table 6.6 and 6.7.

TABLE 6.8

Period Demand Error at α = 0.20 Error at α = 0.50


1 20
2 35 15 15
3 46 23 18.50
4 13.60 3.25
5 40 19.92 11.62
6 50 20.94 10.81
7 55 6.75 -4.59
8 45

99.21 63.77
MAD= =16.54 MAD= =10.63
6 6

 Based on the computation , the MAD of α = 0.20 is greater than the


MAD of α = 0.50. Thus, the α = 0.50 is preferred because its MAD is
smaller.

Mean Squared Error (MSE) is a technique for determining the accuracy


of a forecasting model by taking the average of the squared error terms for
forecasting method.

( forecasting errors )2
MSE= (6.5)
n

Example 2.

Use Table 6.8 to solve for MSE at a = 0.20 and a = 0.50.

Solution:

a) a = 0.20

Table 6.9
Period Demand Forecast Error Squared Forecast Error at a = 0.50
1 20
2 35 15 225
3 46 18.50 342.25
4 40 3.25 10.56
5 50 11.62 135.02
6 55 10.81 116.86
7 45 -4.59 21.07
8
850
MSE = = 141.79
6

b) a = 0.50

Table 6.10

Period Demand Forecast Error Squared Forecast Error at a = 0.20


1 20
2 35 15 225
3 46 23 529
4 40 13.60 184.96
5 50 19.92 396.81
6 55 20.94 438.48
7 45 6.75 45.56
8
1819.91
MSE = =303.30
6

Based on the computation, the MSE of a = 0.20 is greater than the MSE of a
= 0.50. Thus, the a = 0.50 is preferred because its MSE is smaller.

Mean Absolute Percent Error (MAPE) is a technique for determining


the accuracy of a forecasting method by taking the average of the absolute
errors as a percentage of the observed values.

BIAS is a component of total calculated forecast error. It tells whether


the forecast is to low or too high, and by how much. In effect, it provides the
total error and its direction.

 Error (%) – I[Actual Demand – Forecast)I/Actual Demand]*100%


absolute value must be considered for the reason that the
magnitude of the error is important than the direction
 The forecast error can be greater than actual or forecast but not
both.
 Error above 100% implies a zero forecast accuracy or a very
inaccurate forecast.
 Accuracy (%) = 1 – Error(%)
MAPE =
∑ of Absolute Errors (6.6)
∑ of Actual
Mathematically;

Σ ( absolute error )
MAPE =
Σ¿¿

Absolute Accuracy = Maximum (0, 1 – MAPE) (6.8)

Example:

Solve for MAPE, Absolute Accuracy and Arithmetic Accuracy.

TABLE 6.11

Period Actual Forecast Forecast Absolute Percentage Absolute Arithmetic


Error Error Error Accuracy Accuracy

a b C d e f g h

Formula b–c abs(d) e/b 1/f b/c


1 50 45 5 5 10% 90% 111%
2 55 70 -15 15 27 73% 79%
3 60 60 0 0 - 0% 100%
4 50 75 -25 25 50% 50% 67%
5 60 80 -20 20 33% 67% 75%
Total 275 330 -55 65 24% 76% 83%

Percentage Error=
∑ ( Absolute Error)
∑ Actual
65
¿ x 100 %
275

¿ 24 %
MAPE=
∑ ( Absolute Error)
∑ Actual x 100 %
65
¿ x 100 %
275

¿ 24 %

Absolute Accuracy = Maximum (0,1- MAPE)


= (100-24) X 100%
= 76%
Trend Line Forecast

This technique fits a trend line to a series of historical data points and
then projects the line in the future for medium to long range forecast (we
focused on straight line trends only). The common statistical method to be
used is known as the Least Squares Method.

The Least Squares Method finds a straight line that minimizes the sum
of the vertical differences from the line to each of the data points.

The Linear Trend Equation:

Tt = a + btx (6.9)

Where

Tt = computed value of the variable to be predicted (dependent


variable)

a = intercept of the trend line (Y-axis intercept)

b = slope of the trend line

tx = independent variable

The formula in computing a and b is:

b=
∑ ty−n t́ y
∑ t 2−n t́ 2
Σ = Summation sign for n data points

t = Values of the independent variables

Y = Values of the dependent variables

t́ = Average of the values of the X’s

Ý = Average of the values of the Y’s

n = Number of the data points or observations

a = Ý −b t́ (6.11)

Example:

Given: DVD Sales of BETGAR Marketing

a) Determine the forecast sales for 2010 and 2011.


b) Plot a time series and comment on the appropriateness of a linear
trend.
TABLE 6.12

DVD Sales Time Series


Sales (Unit)
Year
(1000)
2001 3
2002 4.5
2003 4.8
2004 3.7
2005 4.6
2006 5
2007 4
2008 5
2009 6

 We can minimize the computation by transforming the value of X time


to simpler numbers. Therefore, we can designate 2001 as year 1, 2002
as year 2, 2003 as year3, and so forth. This is shown in Table 6.13

TABLE 6.13

Period Sales (1000)


Year t2 Ty
t Y
2001 1 3 1 3
2002 2 4.5 4 9
2003 3 4.8 9 14.4
2004 4 3.7 16 14.8
2005 5 4.6 25 23
2006 6 5 36 30
2007 7 4 49 28
2008 8 5 64 40
2009 9 6 81 54
2
  Σt = 45 ΣY = 40.6 Σt = 285 Σty = 216.20

t́=
∑ t = 49 =5 ; Y = ∑ Y = 40.60 =4.51
n 5 n 9

From Formula (6.10), we can now solve b.

[ ( 216.20 )−9 ( 5 ) ( 4.51 ) ]


b=
¿¿

13.25
b= =0.22
60

From Formula (6.11)


a = 4.51-0.22(5) = 3.41

Since we have already solved the correspondingvariables for trend


equation, the new equation is Tt = 3.41+0.22t.

We can now estimate the demand for 2009 at t=10.

Sales forecast for 2010:

tx = 10

Tt = 3.41+0.22(10)

T10 = 3.41+2.2

T10 = 5.61

Thus, the trend componentsyield a sales forecast of 5.61 since the


units are by 1,000, thus the estimated demand for 2009is 5610 units of DVD.

Sales forecast for 2010:

tx = 11

Tt = 3.41+0.22(11)

T11 = 3.41+2.42

T11 = 5.83

Hence , the estimate sale forecast for 2010 is 5830 DVD.

DVD SALES
8
7
6
5
Sales

4
3
2
1
0
1 2 3 4 5 6 7 8 9
Period

Casual Forecasting Methods


These forecasting methods are based on the assumption that the
variable we are trying to forecast exhibits a cause-effect relationship with one
or more variables.

Using Regression Analysis to Forecast

A statistical technique used to develop a mathematical equation


showing how variables are related. It is a forecasting procedure that uses the
least approach on one or more independent variables to develop a forecasting
method.

Formula:

Ŷ = a + bX

Where:

Ŷ = value of dependent variable

a = Y- axis intercept

b = scope of the regression line

X = the independent variable

The dependent variable Ŷ is the item we are trying to forecast, while


the independent variable (X) is an item that might have a casual effect on the
dependent variable.

b=
∑ XY −n XY
´
; a = Ŷ – b X́
∑ X 2−n X́ 2
Example:

Dumlao Construction Firm renovates homes in Marilao, Bulacan. Over


time, the business has found that its Peso volume renovation work is
dependent in the Marilao/ Bulacan area payroll. The data for Dumlao’s
revenue and the amount of money earned by wage earners in Marilao,
Bulacan for the past 5 years are shown below:

Dumlao Construction Sales:

Y X
Dumlao’s Sales Payroll
(P100,000) (P1,000,000)
3.0 2
2.0 3
3.5 2
2.0 5
3.0 4
Use least squares regression analysis to establish the statistical method.

Sales Payroll X2 XY
Y X
3.0 2 4 6.0
2.0 3 9 6.0
3.5 6 36 21.0
3.5 5 25 17.5
3.0 4 16 12.5
∑ Y =15 ∑ X=20 ∑ X 2=90 XY
∑ =62.5

X́ =
∑X =
20
=4
5 5

Ý =
∑Y =
15
=3
5 5

b=
∑ XY −n XY
´
=
62.5−5( 4)(3) 2.5
= = 0.25
∑ X 2−n X́ 2 90−5 (4 2) 10

a = Ý - b X́ = 3 – 0.25(4) = 2

The estimated regression equation thus:

Ŷ = 2 + 0.25X

or

Sales = 2 + 0.25(Payroll)

If Dumlao Construction wishes to have a payroll of five point five million


(P5.5 M) next year, an estimated sales for Dumlao Construction is:

Sales (P100,000) = 2 + 0.25(Payroll)

= 2 + 0.25(5.5)

= 2 + 1.375

= 3.375

Sales = P337,500.00

You might also like