Forecasting

Definition

A forecast is an estimation of some future
event that centres around a variable.

In SCM the major variables are:

Demand
Supply
Price

Forecasting
Demand Forecasts

Demand can be either at an overall or firm
level. An overall demand level is normally
representative of the total market demand
while the firm level demand is more specific
to a certain environment, e.g. global demand
would be an overall demand level while
national demand would be a firm level
demand.


Forecasting
Supply Forecasts

For certain types of materials the number
of suppliers are limited and if the firm
doesn't present supply forecasts it runs
the risk of having no supply; cartels and
monopolies are representative of supply
restrictions.

The supply of labour is another
consideration especially if specific skills
or knowledge are required.
Forecasting
Price Forecasts

Commodities are also subjective to price
variation and firms should seek to tie up
pricing contracts for materials and
commodities that are sensitive to speculative
or political influence, e.g. oil, gas, gold.
Currency fluctuations can also have a
significant impact upon material prices and
can impact significantly on the bottom line.
Hedging is the tool mostly utilised to alleviate
variances in price.
Forecasting
Reasons for Forecasting

Assess long-term capacity needs;
Develop budgets, hiring plans, etc.;
Plan production or order materials;
Get agreement within firm and across
supply chain partners;

Forecasting
Features of Forecasts

Forecasts are almost universally
incorrect;
As the forecast period draws near the
accuracy of the information tends to
improve;
Forecasts for collectives are more
accurate than specific items;
A forecast is no substitute for actual
information or data.


Forecasting
Qualitative Forecasting Techniques

Executive opinions
Sales force composite
Consumer surveys
Delphi method
Life cycle analogy

Forecasting
Quantitative Forecasting Techniques

Last period;
Moving average;
Weighted moving average;
Exponential smoothing;
Linear regression;
Seasonal adjustments;
Correlation coefficient;
Correlation of determination.

Forecasting
Qualitative Methods
Used when situation is vague
and little data exists, e.g.
new products, new
technology.
Involves intuition, experience
*****************************
E.g., forecasting sales to a new
market
Quantitative Methods
Used when situation is stable
and historical data exists, e.g.
existing products, current
technology.

Heavy use of mathematical
techniques
*******************************
E.g., forecasting sales of a
mature product
Forecasting
Period Demand
1 12
2 15
3 11
4 9
5 10
6 8
7 14
8 12
Time Series

A series of observations that
are arranged in
chronological order;

Can be used to estimate
future observations;

Randomness makes it
unreliable;

Can identify seasonality and
underlying trends.
Forecasting
Randomness
0
2
4
6
8
10
12
14
16
1 2 3 4 5 6 7 8
Peri od
D
e
m
a
n
d
Series1
Forecasting
Randomness & Trend
0
2
4
6
8
10
12
14
16
1 2 3 4 5 6 7 8
Peri od
D
e
m
a
n
d
Series1
Linear (Series1)
Forecasting
Last Period Forecast

A forecasting method that determines the next period
demand by using the actual demand from the previous
period.







Month Demand Forecast
Jan 12
Feb 8 12
Mar 10 8
Apr 13 10
Forecasting
Moving Average

A time series model that derives the
forecast for the current period by taking
the average of the preceding periods.

The number of periods to be used are
dependent upon the randomness of the
data the greater the range of the values
then the more periods should be used to
smooth out the results and avoid peaks
and troughs in production.

Forecasting
Period Demand
1 12
2 15
3 11
4 9
5 10
6 8
7 14
8 12
3-period moving average
forecast for Period 8:

F(8) = F7 + F6 + F5/3

= (14 + 8 + 10) / 3
= 10.67
n
D
F
n
i
i t
t

=
+
+
=
1
1
1
Forecasting
Weighted Moving Average

A variant on the moving average model
whereby a weight is applied to
observations in an attempt to add more
certainty to the forecast. The weights
applied would tend to be greater for the
most recent demand period as this
observation should be the most accurate.
The sum of the weights must add up to 1.



Forecasting

=




Weights: Per. 5 = 0.2, Per. 6 = 0.3, Per. 7 = 0.5

Forecast for Period 8

F(8) = (0.5 * 10) + (0.3 * 8) + (0.2 * 14)
= 11.4

=
+ +
n
i
i t i t
D W
1
1 1
1 + t
F
Forecasting
Period Actual Demand
1 12
2 15
3 11 13.5
4 9 13 12.4
5 10 9 10.8
6 8 9.5 9.9
7 14 9 8.8
8 12 11 11.4
9 11.3 11.8
Two-Period
Moving Average
Forecast
Three-Period
Weighted Moving
Average Forecast

Forecasting
0
2
4
6
8
10
12
14
16
1 2 3 4 5 6 7 8 9
Period
V
o
l
u
m
e
Demand
2-PeriodAvg
3-PeriodWt. Avg.
Note how the forecasts smooth out demand variations
Forecasting
A major restriction in the use of moving
averages is that they don't
always produce results based upon the
whole data hence the results are
compromised and subject to prejudice.

X1 + X2 + X3 + X4
F(5) = ---------------------------
4

X2 + X3 + X4 + X5
F(6) = -----------------------
(data X1 eliminated)
4

Forecasting
Exponential Smoothing

Sophisticated weight averaging model that needs only three
numbers:

F
t+1 =
o D
t
+ (1 o) F
t

F
t
= Forecast for the current period t
D
t
= Actual demand for the current period t
= Weight between 0 and 1
Forecasting
Effects of Smoothing

The value assigned to the smoothing constant is
dependent upon the randomness of the data; if the data is
subject to large fluctuations then we need to smooth it by
giving more weight to the previous periods hence will be
low and vice vearsa. The idea is to reduce peaks and
troughs in demand in an attempt to achieve some form of
consistency.


Forecasting
Period Month Demand =0.5
1 Jan 37
2 Feb 40 37
3 Mar 41 38.5
4 Apr 37 39.75
5 May 45 38.37
6 Jun 50 41.68
7 Jul 43 45.84
8 Aug 47 44.42
9 Sep 56 45.71
10 Oct 52 50.85
11 Nov 55 51.42
12 Dec 54 53.21
13 Jan 53.61
F
2
= D
1
+ (1 - )F
1
= (0.50)(37) + (0.50)(37)
= 37
F
3
= D
2
+ (1 - )F
2
= (0.50)(40) + (0.50)(37)
= 38.5
F
13
= D
12
+ (1 - )F
12
= (0.50)(54) + (0.50)(53.21)
= 53.61
Impact of The Value of
Forecasting
Forecasting
Effects of a Trend

When there is a trend in the data then these time series have
a problem as they will always lag behind the trend as they
are based upon historical demand and take time to adjust to
variants.

To address this issue the forecast can either have a trend
constant and further smoothing applied to it or linear
regression is applied.
Forecasting



Adjusted Exponential Smoothing

A method that uses measurable, historical data observations
to make forecasts by calculating the weighted average of the
current period's actual value and forecast with a trend
adjustment added in.


Forecasting
Formula

AF
t +1
= F
t +1
+ T
t +1
where
T = a smoothed trend factor

T
t +1
= |(F
t +1
- F
t
) + (1 - |) T
t
where
T
t
= the last period trend factor
| = a smoothing constant for trend
Forecasting
Period Month Demand =0.5
1 Jan 37 0
2 Feb 40 37
3 Mar 41 38.5
4 Apr 37 39.75
5 May 45 38.37
6 Jun 50 41.68
7 Jul 43 45.84
8 Aug 47 44.42
9 Sep 56 45.71
10 Oct 52 50.85
11 Nov 55 51.42
12 Dec 54 53.21
13 Jan 53.61
Adjusted Exponential Smoothing (=0.30)
T
3
= |(F
3
- F
2
) + (1 - |) T
2
= (0.30)(38.5 - 37.0) + (0.70)(0)
= 0.45
AF
3
= F
3
+ T
3
= 38.5 + 0.45
= 38.95
T
4
= |(F4

- F
3
) + (1 - |) T
3
= (0.30)(39.75 - 38.5) + (0.70)(0.45)
= 0.69
AF
4
= F
4
+ T
4
= 39.75 + 0.69
= 40.44
Forecasting
Period Month Demand =0.5
1 Jan 37 0 - - -
2 Feb 40 37 37 0.00 37.00
3 Mar 41 38.5 38.5 0.45 38.95
4 Apr 37 39.75 39.75 0.69 40.44
5 May 45 38.37 38.37 0.07 38.44
6 Jun 50 41.68 41.68 0.07 41.75
7 Jul 43 45.84 45.84 1.97 47.81
8 Aug 47 44.42 44.42 0.95 45.37
9 Sep 56 45.71 45.71 1.05 46.76
10 Oct 52 50.85 50.85 2.28 53.13
11 Nov 55 51.42 51.42 1.76 53.18
12 Dec 54 53.21 53.21 1.77 54.98
13 Jan 53.61 53.61 1.36 54.97
F
t +1
T
t +1
AF
t +1
Forecasting
When to Employ

Preferred Scenario:
When a trend is present

Good Scenario:
When theres a cyclical or seasonal pattern

Least-effective Scenario
Working with random variations

Forecasting
Applications

Manufacturing Firms: to forecast demand

Service Organisations: to forecast customer arrival
patterns

Financial Analysts: to forecast revenues and profits

Investors: to forecast economic indicators
Forecasting
Linear Regression

A technique in which a straight line is fitted to a set of data
points to measure the effect of a single independent
variable. The slope of the line is the measured impact of that
variable; for a time series the variable is time. Also referred
to as a least squares method.




Forecasting
y = a + bx
where
a = intercept on y axis
b = slope of the line
x = time period
y = forecast for demand
for period x
_
y = mean of y
_
x = mean of x
We need to solve the equation to determine
values for a and b.
x b y a
n
x
x
n
y x
y x
b
n
i
n
i
i
i
n
i
n
i
n
i
i i
i i
=

=
=
=
= =
1
1
2
2
1
1 1
) (
) )( (
Forecasting
Methodology

Arrange the data into X, Y pairs
Compute the sum of the X^2 by squaring each value of
and adding the squares.
Compute the sum of each X value multiplied by its
corresponding Y value.
Calculate the mean of X.
Calculate the mean of Y.
Calculate the slope (b) of the line
Calculate the Y-intercept (a)
Place the slope and Y-intercept values into the equation
for a line.
Forecasting
Period Demand X^2 XY
1 110 1 110
2 190 4 380
3 320 9 960
4 410 16 1640
5 490 25 2450
Solve the regression equation
Forecasting
Period (X) Demand (Y) X2 XY
1 110 1 110
2 190 4 380
3 320 9 960
4 410 16 1640
5 490 25 2450
15 1520 55 5540
10
5
15
98
5
1520
98
5
15
55
5
1520 15
5540
2
= =
=

=
a
b
0
100
200
300
400
500
600
1 2 3 4 5
Y
X
Demand
Regres
Forecasting
Resulting Regression Model






Forecast = 10 + 98*Period (X)



Forecasting
Seasonally Adjusted Forecasts

Seasonal adjustment is the process of estimating and
removing seasonal effects from a time series in order to
better reveal certain non-seasonal features. Seasonal
movements are often large enough that they mask other
characteristics of the data that are of interest to analysts of
current economic trends. Seasonal adjustment produces
data in which the values of neighbouring months are usually
easier to compare. Many data users prefer seasonally
adjusted data because they want to see those
characteristics that seasonal movements tend to mask,
especially changes in the direction of the series.
Forecasting
Methodology
Determine the regression equation to evaluate unadjusted
forecast demand
demand = a + b*period
Determine demand/forecast for each period.
If series covers more than one year then determine the
average for demand/forecast to get the value for the
seasonal index; if less than one year then use value from
step 2.
Multiply unadjusted forecast by seasonal index to
determine adjusted forecast.
Forecasting
Quarter Period Demand
Winter 09 1 80
Spring 2 240
Summer 3 300
Autumn 4 440
Winter 10 5 400
Spring 6 720
Summer 7 700
Autumn 8 880
Forecasting
Year
Quarter
Period Demand Forecast Error
09 Winter 1 80 90 -10
Spring 2 240 198.6 41.4
Summer 3 300 307.1 -7.1
Autumn 4 440 415.7 24.3
10 Winter 5 400 524.3 -124.3
Spring 6 720 632.9 87.1
Summer 7 700 741.4 -41.4
Autumn 8 880 850 30
Forecasting
0
100
200
300
400
500
600
700
800
900
1000
1 2 3 4 5 6 7 8
Demand
Forecast
Forecasting
Determining Seasonal Index for Winter

(Actual / Forecast) for Winter Quarters:
Winter 02: (80 / 90) = 0.89
Winter 03: (400 / 524.3) = 0.76
Average of these two = 0.83

If index is <1 then the forecast is exaggerated; if index >1
then forecast is underestimated.


Forecasting
Year




Quarter
Period
(x)
Deman
d
(y)
Forecas
t Error
Demand/
Forecast
Seasona
l Index
Seasonally
Adjusted
Forecast
09 Winter 1 80 90 -10 0.89 0.83 75
Spring 2 240 198.6 41.4 1.21 1.17 233
Summer 3 300 307.1 -7.1 0.98 0.96 295
Autumn 4 440 415.7 24.3 1.06 1.05 435
10 Winter 5 400 524.3
-
124.3 0.76 0.83 435
Spring 6 720 632.9 87.1 1.14 1.17 740
Summer 7 700 741.4 -41.4 0.94 0.96 712
Autumn 8 880 850 30 1.04 1.05 893
Forecasting
Period (x) Demand (y) x^2 xy
1 80 1 80
2 240 4 480
3 300 9 900
4 440 16 1760
5 400 25 2000
6 720 36 4320
7 700 49 4900
n = 8 8 880 64 7040
Total 36 3760 204 21480
Average 4.5 470
Forecasting
Solving the regression equation:

b = 21480 (36)(3760)/8
204 (36^2)/8

= 4560/42
= 108.57

a = 470 (108.57)(4.5)
= -18.57
Forecasting

To forecast for 2011 we use the regression equation:
a + bx = -18.57 + 108.57x (x = period)

Winter 11 = -18.57 + 108.57(9) = 958.56
Spring 11 = -18.57 + 108.57(10) = 1067.13
Summer 11 = -18.57 + 108.57(11) = 1175.7
Autumn 11 = -18.57 + 108.57(12) = 1284.27




Forecasting











Seasonally Adjusted Forecast
0
200
400
600
800
1000
1 2 3 4 5 6 7 8
Demand
forecast
Forecasting
Causal Forecasting

Causal forecasting methods are based on the relationship
between the variable to be forecasted and an independent
variable other than time. A causal forecast is solved using
regression.

Causal forecasting is used when:
There is a belief something caused demand to act a certain way
Demand or sales patterns vary drastically with planned or
unplanned events
Forecasting
Correlation Coefficient, r

The quantity r, called the linear correlation coefficient,
measures the strength and the direction of a linear
relationship between two variables.

The mathematical formula for computing r is:



________nxy (x)(y)__________
[(nx^2) (x)^2] [(ny^2 - (y)^2]
Forecasting


The value of r is such that -1 < r < +1. The + and signs are used
for positive linear correlations and negative linear correlations
respectively.

Positive correlation: If x and y have a strong positive linear
correlation, r is close to +1. An r value of exactly +1 indicates a
perfect positive fit. Positive values indicate a relationship between
x and y variables such that as values for x increases, values for y
also increase and vice vearsa.


Forecasting



Negative correlation: If x and y have a strong negative linear
correlation, r is close to -1. An r value of exactly -1 indicates a perfect
negative fit. Negative values indicate a relationship between x and y
such that as values for x increase, values for y decrease.

No correlation: If there is no linear correlation or a weak linear
correlation, r is close to 0. A value near zero means that there is a
random, nonlinear relationship between the two variables

Note that r is a dimensionless quantity; that is, it does not depend on
the units employed.

Forecasting
A perfect correlation of +1 occurs only when the data points all
lie
exactly on a straight line.

If r = +1, the slope of this line is positive. If r = -1, the slope of
this line is negative.

A correlation greater than 0.8 is generally described as strong,
whereas a correlation less than 0.5 is generally described as
weak. These values can vary based upon the "type" of data
being
examined. A study utilizing scientific data may require a
stronger
correlation than a study using social science data.
Forecasting
Car
A 9 28.4
B 15 29.3
C 24 37.6
D 30 36.2
E 38 36.5
F 46 35.3
G 53 36.2
H 60 44.1
I 64 44.8
J 76 47.2
Age
(months)
Minimum
Stopping at 40
kph
(metres)
Forecasting
x y x2 y2
9 28.4 81 806.56 255.6
15 29.3 225 858.49 439.5
24 37.6 576 1413.76 902.4
30 36.2 900 1310.44 1086
38 36.5 1444 1332.25 1387
46 35.3 2116 1246.09 1623.8
53 36.2 2809 1310.44 1918.6
60 44.1 3600 1944.81 2646
64 44.8 4096 2007.04 2867.2
76 47.2 5776 2227.84 3587.2
Totals 415 375.6 21623 14457.72 16713.3
xy
X-bar = 415/10 = 41.5 n = 10
y-bar = 375.6/10 = 37.6

r = (10)(16713.3) (415)(375.6) / ((10)(21623) (415^2)(10)(14457.72)
(375.6^2)
r = 11259 / (44005 x 3501.84)
r = 11259 / 124.14
r = 0.91
Forecasting



Coefficient of Determination

The coefficient of determination, r^2, is useful because it gives the
proportion of the variance (fluctuation) of one variable that is
predictable from the other variable.

It is a measure that allows us to determine how certain one can
be in making predictions from a certain model/graph.

The coefficient of determination is such that 0<r 2<1, and
denotes the strength of the linear association between x and y.
Forecasting
The coefficient of determination represents the percentage
of the data that is the closest to the line of best fit. For
example, if r = 0.922, then r 2 = 0.850, which means that
85% of the total variation in y can be explained by the linear
relationship between x and y (as described by the regression
equation). The other 15% of the total variation in y remains
unexplained.

The coefficient of determination is a measure of how well the
regression line represents the data. If the regression line
passes exactly through every point on the scatter plot, it
would be able to explain all of the variation. The further the
line is away from the points, the less it is able to explain.
Forecasting
Measuring Forecast Accuracy

There are generally two methods employed to test the
accuracy of forecasts and to determine if the model being
used is generating reliable data.

Mean forecast error (MFE);
Mean absolute deviation (MAD).



Mean Forecast Error

This measures the bias of the model being used;
0 = unbiased model;
<1 = model over-forecasts;
>1 = model under-forecasts.

FE = forecast error, n = no. of periods.

( )
n
E
n
i
i
MFE

=
=
1
Forecasting
Mean Absolute Deviation

Measures the average size of the errors regardless of whether
there is under or over forecasting.

0 = perfect model;
>0 = flaws in model

E = absolute deviation, n = no. of periods.
n
E
n
i
i
MAD

=
=
1
Forecasting
Forecasting
Period Demand Forecast Error Absolute Error
3 11 13.5 -2.5 2.5 MFE = -2/6 = -0.33
4 9 13 -4 4 MAD = 14/6 = 2.33
5 10 10 0 0
6 8 9.5 -1.5 1.5
7 14 9 5 5
8 12 11 1 1
N = 6 -2 14


Interpretation of MFE & MAD

Low MFE and MAD:
The forecast errors are small and unbiased

Low MFE, but high MAD:
On average, the forecast is reliable.

High MFE and MAD:
The forecasts are inaccurate and biased.
Forecasting