Module 2

Lecture 5 Forecasting Demand

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 1

Outline
 Time-Series Forecasting

 Techniques for trend

 Techniques for Seasonality

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 2

Learning Objectives
When you complete this lecture you should be able
to :
 Comprehend the Techniques for Trend,
 Comprehend the Techniques for Seasonality,
 Comprehend the Simple Linear Regression,
 Explain three measures of forecast accuracy

3
Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT
Types of Forecasting

Quantitative Qualitative
Approaches Approaches

Time-series Associative Jury of Delphi Sales force Consumer

models model executive method composite Market
opinion Survey
Naïve Techniques for Techniques Simple Linear Judgmental Method
Forecasts Averaging for Trend Regression

Techniques for
Moving Seasonality
Average

Weighted Techniques for

Moving Cycles
Average

Exponential
Smoothing

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 4

III. Techniques for trend
 When a simple data plot (demand) shows the
existence and nature of a trend; then trend
equation can be used to develop equation for the
straight line (forecasts).
 A linear trend equation has the form:
Ft = a + b t
 Where:
 Ft = forecast for period t
 a = Value of Ft at t = 0
 b = slope of the line
 t = specified number of time periods
Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 5
III. Techniques for trend

demand

Slope of the line

Value of a when t=0 How much the value of
Starting point for demand change every
the line time.

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 6

Calculating a and b

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 7

Example
 Cell phone sales for a California-based firm over
the last 10 weeks are shown in the table below.
Week Unit sales
a. Plot the data, and visually = t (Value of the time series) = Y
check to see if a linear 1 700
trend line would be 2 724
appropriate. 3 720
4 728
b. Determine the equation of 5 740
the trend line. 6 742
c. Predict sales for weeks 11 7 758
and 12. 8 750
9 770
10 775 8
Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT
Solution - A
A- A plot suggests that a
linear trend line would
be appropriate

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 9

Solution - B
Week Unit t2 ty Determine the equation of
sales
the trend line: Ft = a + b t
(t) (y)
1 700 1 700
2 724 4 1448
3 720 9 2160 Then:
4 728 16
b = 10(41358) – 55(7407) = 7.51
2912
Then: 10(385) – 55(55)
5 740 25 3700 b = 10(41358) – 55(7407) = 7.51
6 742 36 4452 10(385) – 55(55)
a = 7407 – 7.51(55) = 699.4
7 758 49 5306 10
8 750 64 6000 a = 7407 – 7.51(55) = 699.4
The trend line is: Ft = a + bt
9 770 81 6930 10
Ft = 699.4 + 7.51t
10 775 100 7750 The trend line is: Ft = a + b t
Σt = 55 Σy=7407 Σt2 = 385 Σty = 41358
Production and Operation Management
Ft = 699.4 + 7.51 t 10
Solution - C

 Substituting the values of t into this equation, the

forecast for the next two periods (at t = 11 and t =
12) are:
 Ft = a + b t
 F11 = 699.4 + 7.51 (11) = 782.01 = 782
 F12 = 699.4 + 7.51 (12) = 789.52 = 789

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 11

Example
 The owner of “Darkest Tans Week No. of
Unlimited” in a local mall is visits
forecasting this week demand t
( ) y
( )
for the one new tanning booth, 1 100
based on the following historical
2 140
data.
3 110
a. Determine the equation of the 4 150
trend line.
5 120
b. Predict sales for weeks 7 and 8. 6 160

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 12

Solution
 Ft = a + bt

Week Unit t2 ty
(t) sales
(y)
Then:
1 100 1 100 b = 6(2870) – 21(780)
6(91) – 21(21)
=8
2 140 4 280
3 110 9 330
a = 780 – 8 (21) = 102
4 150 16 600
6
5 120 25 600 The trend line is: Ft = a + b t
6 160 36 960 Ft = 102 + 8 t
Σt = 21 Σy=780 Σt2 = 91 Σty = 2870
13
Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT
Solution
 Substituting the values of t into this equation, the
forecast for the next two periods (at t = 7 and t =
8) are:
 Ft = a + bt

 Ft = 102 + 8 t
 F7 = 102 + 8 (7) = 158
 F8 = 102 + 8 (8) = 166

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 14

Types of Forecasting

Quantitative Qualitative
Approaches Approaches

Time-series Associative Jury of Delphi Sales force Consumer

models model executive method composite Market
opinion Survey
Naïve Techniques for Techniques Simple Linear Judgmental Method
Forecasts Averaging for Trend Regression

Moving
Average

Weighted Techniques for

Moving Seasonality
Average

Exponential
Smoothing

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 15

Seasonality
 The amount that the actual values deviate from the
average value of a series.

 Regularly repeating movements in series values that

can be tied to repeated events.

 It is applied to monthly, weekly, daily and other

regularly repeated patterns in data.

 Example: sales of coca during summer, sales of

Blanket during winter
Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 16
Models of Seasonality

• Seasonality is expressed as a quantity that gets

Additive added to or subtracted from the time-series
average, in order to integrate seasonality.
• Forecasting = trend + seasonality + random +
(Absolute value) irregular + cycle

• Seasonality is expressed as a percentage to be

multiplied by the average of the series (or trend),
Multiplicative it is always called seasonal relatives or seasonal
index; in order to integrate seasonality.
(Relative value) • Forecasting = trend * seasonality * random *
irregular * cycle

Could be ignored
Mostly used as its weight is random and irregular
small, and need are hard to forecast 17
huge data.
Seasonal Relatives (Index)
 A seasonal relative, or a seasonal index
or seasonal factor; is how much the demand for
that particular period tends to be above (or below)
the average demand. A value equal to:
 Equal 1: the demand for that period is exactly the same
as the average
 Less than 1: demand are less than the average, which
indicates Low Season
 More than 1: demand are bigger than the average,
which indicates High Season
 Sum of relatives should be equal number of
seasons
Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 18
Integrating seasonality
to forecast
 Integrating seasonality into forecast is useful when
demand has both trend (or average) and seasonal
components; it is conducted as following:
 Obtain trend estimates for the desired periods using
trend equation.
 Add seasonality to the trend by multiplying these
trend estimates by the corresponding seasonal
relative (assuming multiplicative model is appropriate).

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 19

Deseasonalize the data
 For a good forecast, deseasonalize data in order to get
a clearer picture of the nonseasonal (e.g., trend)
components of the data series to be more accurate
(free from fluctuations exist in seasonality), to fit the
trend equation instead of the original data.
 To deseasonalize the data:
 Divide the demand of each period by the corresponding seasonal
relatives.
 Then multiply the trend by the associated relative of the period.

 N.B:
In sale period: we consider it seasonality because we have historical data.
If we have offer at certain time for the first time, then it is not seasonality. But if
repeated later, this will be used in the historical data later.
Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 20
Example

 The manager of a parking lot has computed

daily relatives for the number of cars per day in
the lot.
 The computations are repeated about three
weeks (as shown).

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 21

Example (cont.)
Day Cars
Tues 67
Wed 75
Thur. 82
Fri 98
Sat 90
Sun 36
Mon 55
Tues 60
Wed 73
Thur. 85
Fri 99
Sat 86
Sun 40
Mon 52
Tues 64
Wed 76
Thur. 87
Fri 96
Sat 88
Sun 44
22
Mon 50
Centered moving average = number of seasons = 7days.
Therefore, MA must centered between 7.

Day Cars Centered MA7

Tues 67
Wed 75
Thur. 82
Fri 98
Sat 90
Sun 36
Mon 55
Tues 60
Wed 73
Thur. 85
Fri 99
Sat 86
Sun 40
Mon 52
Tues 64
Wed 76
Thur. 87
Fri 96
Sat 88
Sun 44
23
Mon 50
MA must centered between 7 (67 + 75 + 82 + 98 + 90 + 36 + 55) / 7 = 71.86

Day Cars Centered MA7

Tues 67
Wed 75
Thur. 82
Fri 98 71.86
Sat 90 70.86
Sun 36 70.57
Mon 55 71
Tues 60 71.14
Wed 73 70.57
Thur. 85 71..14
Fri 99 70.71
Sat 86 71.29
Sun 40 71.71
Mon 52 72
Tues 64 71.57
Wed 76 71.86
Thur. 87 72.43
Fri 96 72.14
Sat 88
Sun 44
24
Mon 50
MA must centered between 7 ( 75 + 82 + 98 + 90 + 36 + 55 + 60) / 7 = 70.88

Day Cars Centered MA7

Tues 67
Wed 75
Thur. 82
Fri 98 71.86
Sat 90 70.86
Sun 36 70.57
Mon 55 71
Tues 60 71.14
Wed 73 70.57
Thur. 85 71..14
Fri 99 70.71
Sat 86 71.29
Sun 40 71.71
Mon 52 72
Tues 64 71.57
Wed 76 71.86
Thur. 87 72.43
Fri 96 72.14
Sat 88
Sun 44
25
Mon 50
Solution (cont.) To calculate the
Estimated Seasonal Relatives

Day Cars Centered MA7 Cars/MA7

Tues 67
98 / 71.86 = 1.36
Wed 75
Thur. 82
Fri 98 71.86 1.36
Sat 90 70.86
Sun 36 70.57
Mon 55 71
Tues 60 71.14
Wed 73 70.57
Thur. 85 71..14
Fri 99 70.71
Sat 86 71.29
Sun 40 71.71
Mon 52 72
Tues 64 71.57
Wed 76 71.86
Thur. 87 72.43
Fri 96 72.14
Sat 88
Sun 44
26
Mon 50
Solution (cont.) To calculate the
estimated seasonal relatives

Day Cars Centered MA7 Cars/MA7

Tues 67
98 / 71.86 = 1.36
Wed 75
Thur. 82
Fri 98 71.86 1.36
Sat 90 70.86 1.27
Sun 36 70.57 0.51
Mon 55 71 0.77
Tues 60 71.14 0.84
Wed 73 70.57 1.03
Thur. 85 71..14 1.19
Fri 99 70.71 1.4
Sat 86 71.29 1.21
Sun 40 71.71 0.56
Mon 52 72 0.72
Tues 64 71.57 0.89
Wed 76 71.86 1.06
Thur. 87 72.43 1.2
Fri 96 72.14 1.33
Sat 88
Sun 44
27
Mon 50
Add all Fridays together then divide by number of days, to
calculate the estimated seasonal relatives

Day Cars Centered MA7 Cars/MA7

Tues 67
Wed 75
Thur. 82
Fri 98 71.86 1.36
Sat 90 70.86 1.27
Sun 36 70.57 0.51
Mon 55 71 0.77
Tues 60 71.14 0.84
Wed 73 70.57 1.03
Thur. 85 71..14 1.19
Fri 99 70.71 1.4
Sat 86 71.29 1.21
Sun 40 71.71 0.56
Mon 52 72 0.72
Tues 64 71.57 0.89
Wed 76 71.86 1.06
Thur. 87 72.43 1.2
Fri 96 72.14 1.33
Sat 88
Sun 44
28
Mon 50
Solution (cont.)
The estimated seasonal relatives will be:
 Friday: (1.36 + 1.4 + 1.33)/3 = 1.36
Sum is 6.985
 Saturday: (1.27 + 1.21)/2 = 1.24
 Sunday: (0.51 + 0.56)/2 = 0.535
 Monday: (0.77 + 0.72)/2 = 0.745
Must equal the number
 Tuesday: (0.84 + 0.89)/2 = 0.865
of seasons = 7
 Wednesday: (1.03 + 1.06)/2 = 1.045
 Thursday: (1.19 + 1.2)/2 = 1.195

 If it is not, you have to multiply by a correction factor =

7/6.985
Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 29
Solution (cont.)
The new relatives will be:
 Friday: 1.36 * 7/6.985 = 1.363 → High Season
 Saturday: 1.24 * 7/6.985 = 1.243→ High Season
Sum is
 Sunday: 0.535 * 7/6.985 = 0.536 → Low Season
7
 Monday: 0.745 * 7/6.985 = 0.746 → Low Season
 Tuesday: 0.865 * 7/6.985 = 0.867→ Low Season
 Wednesday: 1.045 * 7/6.985 = 1.047→ High Season
 Thursday: 1.195 * 7/6.985 = 1.198→ High Season
 Equal 1: the demand for that period is exactly the same as the average
 Less than 1: demand are less than the average, which indicates Low Season
 More than 1: demand are bigger than the average, which indicates High Season

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 30

Deseasonalize the data 67/0.867= 77.28

Day Cars New Relatives Des. Data=Cars/New Relatives

Tues 67 0.867 77.28
Wed 75 1.047 71.63
Thur. 82 1.198 68.45
Fri 98 1.363 71.90
Sat 90 1.243 72.41
Sun 36 0.536 67.16
Mon 55 0.746 73.73
Tues 60 0.867 69.20
Wed 73 1.047 69.72
Thur. 85 1.198 70.95
Fri 99 1.363 72.63
Sat 86 1.243 69.19
Sun 40 0.536 74.63
Mon 52 0.746 69.71
Tues 64 0.867 73.82
Wed 76 1.047 72.59
Thur. 87 1.198 72.62
Fri 96 1.363 70.43
Sat 88 1.243 70.80
Sun 44 0.536 82.09
31
Mon 50 0.746 67.02
Period (t) Day Des. Data (Y) t2 t*y
1 Tues 77.28 1 77.28
2 Wed 71.63 4 143.26
To
3 Thur. 68.45 9 205.35
Forecast
4 Fri 71.90 16 287.6
5 Sat 72.41 25 362.05
6 Sun 67.16 36 402.96
7 Mon 73.73 49 516.11
8 Tues 69.20 64 553.6
9 Wed 69.72 81 627.48
10 Thur. 70.95 100 709.5
11 Fri 72.63 121 798.93
12 Sat 69.19 144 830.28
13 Sun 74.63 169 970.19
14 Mon 69.71 196 975.94
15 Tues 73.82 225 1107.3
16 Wed 72.59 256 1161.44
17 Thur. 72.62 289 1234.54
18 Fri 70.43 324 1267.74
19 Sat 70.80 361 1345.2
20 Sun 82.09 400 1641.8
21 Mon 67.02 441 1407.42

Σt2 = 3310 Σty = 16625.97 32

Calculating Trend Equation
Determine the equation of the trend line from the
deseasonalized data : Ft = a + bt

Then multiply the trend by the associated relative of the

period (Tuesday).

F22 = [a + b (22)] * 0.867

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 33

Period (t) Day Des. Data (Y) t2 t*y Seasonal Relative
1 Tues 77.28 1 77.28 0.867
2 Wed 71.63 4 143.26 1.047
3 Thur. 68.45 9 205.35 1.198
4 Fri 71.90 16 287.6 1.363
5 Sat 72.41 25 362.05 1.243
6 Sun 67.16 36 402.96 0.536
7 Mon 73.73 49 516.11 0.746
8 Tues 69.20 64 553.6 0.867
9 Wed 69.72 81 627.48 1.047
10 Thur. 70.95 100 709.5 1.198
11 Fri 72.63 121 798.93 1.363
12 Sat 69.19 144 830.28 1.243
13 Sun 74.63 169 970.19 0.536
14 Mon 69.71 196 975.94 0.746
15 Tues 73.82 225 1107.3 0.867
16 Wed 72.59 256 1161.44 1.047
17 Thur. 72.62 289 1234.54 1.198
18 Fri 70.43 324 1267.74 1.363
19 Sat 70.80 361 1345.2 1.243
20 Sun 82.09 400 1641.8 0.536
21 Mon 67.02 441 1407.42 0.746

Σt2 = 3310 Σty = 16625.97 34

Seasonal Relatives
The number of periods needed in a centered moving
average is equal to the number of “seasons” involved.
 With monthly data, a 12-period moving average is
needed.
 With daily data, a 7-period moving average is needed.
 With quarterly data, a 4-period moving average is
needed.
Note:
When the number of period is even, one additional step is
needed, because the middle of an even period falls
between two periods.
An additional step is required; by taking two-period moving
average of the even-numbered centered moving average.

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 35

Summary
 Time-Series Forecasting
 Techniques for trend
 Techniques for Seasonality

Production Operations Management – Dr. Nevien Farouk Khourshed - AASTMT 36