Analysis of Forecasting Demand of Aerosol

2.3.1 Forecasting
Forecasting is the process of making statements about events whose actual outcomes
(typically) have not yet been observed. A commonplace example might be estimation for
some variable of interest at some specified future date. Prediction is a similar, but more
general term. Forecasting is an important tool for the future of demand condition. This is
a prediction of future events used for planning purposes. Forecasting is needed to aid in
determining what resources are needed, scheduling existing resources and acquiring
additional resources.

2.3.2 Types of Forecasting

Mainly, the forecasting can be classified as:

Qualitative methods

Quantitative methods

Qualitative Methods: Qualitative forecasting techniques generally employ the judgment

of experts in the appropriate field to generate forecasts. A key advantage of these
procedures is that they can be applied in situations where historical data are simply not
available. Moreover, even when historical data are available, significant changes in
environmental conditions affecting the relevant time series may make the use of past data
irrelevant and questionable in forecasting future values of the time series.
Quantitative Methods: include causal methods and time-series analysis. Causal methods
are historical data on independent variables, such as promotional campaigns, economic
conditions and competitors actions to predict demand or supply. Time series analysis is a
statistical approach that relies heavily on historical demand data to project the future size
of demand or supply and recognizes trends and seasonal patterns.
To make our analysis we have followed the following techniques of forecasting.
1) Simple Moving Average
2) Single exponential Smoothing
3) Simple Linear Regression

Page | 1

Analysis of Forecasting Demand of Aerosol

1) Simple Moving Average
The simple moving average forecast uses a number of the recent actual data values in
generating a forecast. The moving average forecast can be computed using the following
equation:
n

A t i

Ft = t=1

A tn ++ A t2 + A t 1
n

Where,

Ft

= Forecast for the time period t

A t 1

= Actual value in period t 1

MA t

=n period moving average

N= number of periods (data points) in the moving average

2) Single Exponential Smoothing

Single Exponential Smoothing largely overcomes the limitations of moving average
models. Each new forecast is based on the previous forecast plus a percentage of the
difference between that forecast and the actual value of the series at that point.
Ft =F t1 + ( A t1F t 1 )
Where,

Ft
Ft 1

= Forecast for period t

= Forecast for previous period

= Smoothing constant
A t 1

= Actual demand for previous period

Page | 2

Analysis of Forecasting Demand of Aerosol

3) Simple Linear Regression
This method involves linear relationship between two variables. The object in linear
regression is to obtain an equation of a straight line that minimizes the sum of squared
vertical deviations of data points from the line. The equation is:
y=a+bx

Where, a = constant and b = slope of the line

And,
a= y b x

b=

n xy x y
n x 2( x2 )

2.3.2 Accuracy Test

After forecasting, applying different techniques, we have found out the forecasting
accuracy and the appropriate techniques for forecasting the sales of the product. To
calculate the forecasting accuracy, we have used the following techniques:
Mean absolute deviation (MAD)
Mean squared error (MSE)

MAD is the average absolute error, MSE is the average of squared error and MAPE is the
average percent error. The formulas used to compute MAD, MSE are given below:

ActualForecast
n
MAD=

Page | 3

Analysis of Forecasting Demand of Aerosol

2
( ActualForecast )

MSE=

n1

MSE is similar to the variance of a random sample; however, it is more sensitive to a few
large errors than MAD. Consequently, MAD, the average of the absolute discrepancies
between the actual and fitted values in a given time series is often preferred. If a model
fits the past time-series data perfectly, the MAD value would be zero. As the fit worsens,
the value of MAD increases. In other words, a small value of MAD is desirable. In
addition, when forecast errors are normally distributed, an estimate of the standard
deviation of the forecast error is given by 1.25 times MAD. We also considered MAPE
test. The advantage of this measure of accuracy is that MAPE is not dependent on the
magnitude of the values of demand.

2.1.6 Data Analysis Tool

In order to get the forecasting output of aerosol, we have used Microsoft Word and
Microsoft Excel.

2.2 Limitations
This study has been done based on the data of sales volume and production from January
2014 to October 2015. We tried to propose suitable model for forecasting with the data
series in hand. We have also tried to forecast sales volume based on several forecasting
techniques. Scope of the study also included the Testing of the forecasting accuracy.
Here, we took quarter sales volume & used it for the remaining quarters. Not enough
statistical data has allowed little space for in depth analysis & thus to some extent might
have hampered the accuracy.

Page | 4

Analysis of Forecasting Demand of Aerosol

Chapter - 03
1.0 ANALYSIS
Data from December 2014-October 2015

Actual Demand
Aerosol

250ml

350ml

475ml

800ml

December

23,618

15,711

80,520

58713

January
February
March
April
May
June
July
August
September
October

21333
24436
22517
26966
39480
24446
21092
40733
56176
39711

29734
19165
40529
16229
31825
13885
39763
23667
25080
15711

114352
129425
215992
55095
177265
104231
152420
142918
121818
110088

57210
58573
56115
79781
65722
75154
53479
74942
63768
61962

Actual Demand
250,000
200,000
150,000
100,000
50,000
0

250ml

350ml

475ml

800ml

Page | 5

Analysis of Forecasting Demand of Aerosol

3.1 Simple Moving Average

The simple moving average forecast uses a number of the recent actual data values in
generating a forecast. The following figures shows the 3 simple moving average on the
basis of sales data which is the actual demand for 2013-2015, and we figure the forecast
of these three years for different SKU 250ml, 350ml, 450ml, and 800ml.
Data from December 2014-October 2015

Simple Moving Average

Following

Aero
sol

250ml

350ml

475ml

800ml

the difference

Dece
mber

18681

17505

113,819

58713

of forecasting

Janua

20180

18128

138297

57210

actual demand

data

with

of every SKU.
and
data

with

moving
not consistent
other.

Actual

ry
Febru
ary
Marc
h
April
May
June
July
Augu
st
Septe
mber
Octob
er
Nove
mber
MAD
MSE

graphs shows

data

24329

22726

127248

58573

forecasting

23107

21537

120834

56115

simple

29366
28493
30636
24109
25500

29809
25308
29528
20646
28491

108099
153256
133504
149451
112197

79781
65722
75154
53479
74942

25575

25772

144639

63768

28673

29503

133190

61962

26672

23779

124941

49028

2801.8472

3674.1666

11162.666

7502.3472

222222
12127079.

666667
17782769.

666667
128008791

222222
82131571.

24305

05556

5.24305

40972

average
with

are
each

Page | 6

Analysis of Forecasting Demand of Aerosol

250 ml
40000
35000
30000
25000
20000
15000
10000
5000
0

250 ml 2015 Actual Demand

250 ml 2015 Simple Moving Average

350 ml
45000
40000
35000
30000
25000
20000
15000
10000
5000
0

350 ml 2015 Actual Demand

350 ml 2015 Simple Moving Average

Page | 7

Analysis of Forecasting Demand of Aerosol

475 ml
250000
200000
150000
100000
50000
0

475 ml 2015 Actual Demand

475 ml 2015 Simple Moving Average

800 ml
120000
100000
80000
60000
40000
20000
0

800 ml 2015 Actual Demand

800 ml 2015 Simple Moving Average

Page | 8

Analysis of Forecasting Demand of Aerosol

3.2 Single Exponential Smoothing

Here we took

=0.3

to do our analysis because when we divide maximum value with

minimum value it comes out more than 2.

Data from December 2014-October 2015

Single Exponential
Smoothing

Following
the difference
data

with

of every SKU.
450ml

and

demand

data

two
are not.

Dec
emb
er

to

other

475ml

800m
l

Febr
uary
Marc
h
April
May
June
July
Aug
ust
Sept
emb
er
Octo
ber
Nov
emb
er

MAD

MSE

graphs shows
of forecasting
actual demand
Here

23,618

18,125

166,919

90072

21,333

16,018

107,470

57265

24,436

31,530

150,945

63819

22,517
26,966
39,480
24,446
21,092

50,975
46,078
27,479
20,853
26,776

180,449
198,100
149,831
94,837
181,209

97654
123017
75897
31945
88608

40,733

23,234

154,016

60561

56,176

32,300

209,475

66367

39,711

30,134

186,485

72156

23,101

40,692
8304.58

174,636
26203.7

90072

333333

7777777

9149.5
112772

33

8
1071606

476.18

108471

969.555

75

879.75

56

Janua
ry

forecasting
consistent

Aer
oso 250m 350m
l
l
l

only

800ml
and
data

are

each other and

SKUs

data

18698.3
472222
22
559189
441.076
39

Page | 9

Analysis of Forecasting Demand of Aerosol

250ml
60000
50000
40000
30000
20000
10000
0

250 ml 2015 Actual Demand

250 ml 2015 Single exponential Smoothing

350 ml
60000
50000
40000
30000
20000
10000
0

350 ml 2015 Actual Demand

350 ml 2015 Single exponential Smoothing

Page | 10

Analysis of Forecasting Demand of Aerosol

475 ml
250000
200000
150000
100000
50000
0

475 ML 2015 Actual Demand

475 ML 2015 Single exponential Smoothing

800 ml
140000
120000
100000
80000
60000
40000
20000
0

800 ml 2015 Actual Demand

800 ml 2015 Single exponential Smoothing

Page | 11

Analysis of Forecasting Demand of Aerosol

3.3 Simple Linear Regression

Data from December 2014-October 2015

Single Linear Regression

Aeros
ol

250ml

350ml

475ml

800ml

January

23,378
25557

24134
24365

121276
122550

66051
65127

February

25737

24596

123825

64202

March

25916

24827

125099

63277

April

26096

25058

126373

62352

May

26275

25289

127648

61428

June

26454

25520

128922

60503

July

26634

25752

130196

59578

August
Septemb
er

26813

25982

131471

58653

26993

26213

132745

57729

October
Novemb
er

27172

26444

134019

56804

25199

23903

120002

MAD

538.1666
383524.38

693
635883.

3823
19351211.91

66976
2774.166
7

MSE

89

25

67

Decemb
er

10190340
.5

Following graphs shows the difference of forecasting data with actual demand of every
SKU.

Page | 12

Analysis of Forecasting Demand of Aerosol

250 ml
40,000
35,000
30,000
25,000
20,000
15,000
10,000
5,000
-

Actual Demand

Simple Linear Regression

350 ml
40,000
35,000
30,000
25,000
20,000
15,000
10,000
5,000
-

Actual Demand

Simple Linear Regression

Page | 13

Analysis of Forecasting Demand of Aerosol

475 ml
160,000
140,000
120,000
100,000
80,000
60,000
40,000
20,000
-

Actual Demand

Simple Linear Regression

800 ml
80,000
70,000
60,000
50,000
40,000
30,000
20,000
10,000
-

Actual Demand

Simple Linear Regression

Page | 14

Analysis of Forecasting Demand of Aerosol

3.4 Forecasting Differences

Data from December 2014-October 2015

Month

Forecasting of
ACI

250ml
Simple Moving
Average

December
January
February
March
April
May
June
July
August
Septembe
r
October
November

22615
22614
22614
22614
45228
22614
22614
45230
67845
45230

18681
20180
24329
23107
29366
28493
30636
24109
25500
25575

22615

28673
26672

Single
Simple
Exponential
Linear
Smoothing Regression
23,618
21,333
24,436
22,517
26,966
39,480
24,446
21,092
40,733

23,378
25557
25737
25916
26096
26275
26454
26634
26813

56,176
39,711
23,101

26993
27172
25199

Page | 15

Analysis of Forecasting Demand of Aerosol

Forecasting Differences Between Different Methods
80000
70000
60000
50000
40000
30000
20000
10000
0

Forecasting of ACI

Simple Moving Average

Single Exponential Smoothing

Simple Linear Regression

Page | 16

Analysis of Forecasting Demand of Aerosol

Data from December 2014-October 2015

Month

Forecasting of
ACI

350ml
Simple Moving
Average

December
January
February
March
April
May
June
July
August
Septembe
r
October
November

16150
32300
64608
48456
32300
16150
32300
16150
36000
32300

17505
18128
22726
21537
29809
25308
29528
20646
28491
25772

48450

29503
23779

Single
Simple
Exponential
Linear
Smoothing Regression
18,125
16,018
31,530
50,975
46,078
27,479
20,853
26,776
23,234

24134
24365
24596
24827
25058
25289
25520
25752
25982

32,300
30,134
40,692

26213
26444
23903

Page | 17

Analysis of Forecasting Demand of Aerosol

Forecasting Differences Between Different Methods
70000
60000
50000
40000
30000
20000
10000
0

Forecasting of ACI

Simple Moving Average

Single Exponential Smoothing

Simple Linear Regression

Page | 18

Analysis of Forecasting Demand of Aerosol

Data from December 2014-October 2015

Month

Forecasting of
ACI

475ml
Simple Moving
Average

December
January
February
March
April
May
June
July
August
Septembe
r
October
November

119020
166628
202317
190432
190432
59510
214200
154700
238000
214200

113,819
138297
127248
120834
108099
153256
133504
149451
112197
144639

202300

133190
124941

Single
Simple
Exponential
Linear
Smoothing Regression
166,919
107,470
150,945
180,449
198,100
149,831
94,837
181,209
154,016

121276
122550
123825
125099
126373
127648
128922
130196
131471

209,475
186,485
174,636

132745
134019
120002

Page | 19

Analysis of Forecasting Demand of Aerosol

Forecasting Differences Between Different Methods
250000

200000

150000

100000

50000

Forecasting of ACI

Simple Moving Average

Single Exponential Smoothing

Simple Linear Regression

Page | 20

Analysis of Forecasting Demand of Aerosol

Data from December 2014-October 2015

Month

Forecasting of
ACI

800ml
Simple Moving
Average

December
January
February
March
April
May
June
July
August
Septembe
r
October
November

63684
63594
113056
127188
98924
7060
105900
49420
70600
84720

58713
57210
58573
56115
79781
65722
75154
53479
74942
63768

42360

61962
49028

Single
Simple
Exponential
Linear
Smoothing Regression
90072
57265
63819
97654
123017
75897
31945
88608
60561

66051
65127
64202
63277
62352
61428
60503
59578
58653

66367
72156
90072

57729
56804
66976

Page | 21

Analysis of Forecasting Demand of Aerosol

Forecasting Differences Between Different Methods
140000
120000
100000
80000
60000
40000
20000
0

Forecasting of ACI

Simple Moving Average

Single Exponential Smoothing

Simple Linear Regression

Page | 22

Analysis of Forecasting Demand of Aerosol

3.5 Accuracy Test

SKU
Simple
Moving
Average
Single
Exponential
Smoothing
Simple
Linear

250ml

350ml

475ml

800ml

MAD
MSE

2802
12127079

3674
17782769

11163
1280087915

7502
82131571

MAD
MSE

9150

8305

26204

18698

112772476
538
383524

108471879
693
635885

1071606969
3823
19351211

559189441
2774
10190340

MAD
MSE

Regression

In the above chart we can see that simple linear regression is giving out better figures in
terms of comparison between it, simple moving average and single exponential
smoothing for forecasting. As we know the lower the figure the better the model. Thus for
further forecasting we can use the simple linear regression.

Page | 23