LECTURE 3:
Demand Forecasting
in
Supply Chains
Measures of Forecast Error:
MSE, MAD, & MAPE
1
Introduction
Eight steps to forecasting:
1. Determine the use of the forecast.
2. Select the items or quantities to be forecasted.
3. Determine the time horizon of the forecast.
4. Select the forecasting model or models. ©
5. Gather the data needed to make the forecast. 20
06
by
6. Validate the forecasting model. Pr
en
7. Make the forecast. tic
e
8. Implement the results. Ha
ll,
Inc
.
Up
pe
r
Sa
dd
These steps provide a systematic way of initiating, le
Riv
er,
designing, and implementing a forecasting system. NJ
07
45
8
RECALL:Types of Forecasts
Forecasting Techniques
No single method is superior
Qualitative Models: Time-Series Methods: Causal Methods:
attempt to include include historical data over a include a variety of
subjective factors time interval factors
©
Delphi Moving Regression 20
Methods Average Analysis
06
by
Pr
en
Jury of Executive Exponential Multiple tic
e
Opinion Smoothing Regression Ha
ll,
Inc
Trend Projections .
Sales Force Up
pe
Composite Sa
r
Decomposition dd
Consumer le
Riv
Market Survey er,
NJ
07
45
8
Measures of Forecast Accuracy
We compare forecasted values with actual values
to see how well one model works or to compare
models
Forecast error = Actual value – Forecast value
One measure of accuracy is the mean absolute deviation (MAD)
MAD
forecast error
n
Measures of Forecast Accuracy
Using a naïve forecasting model
ACTUAL ABSOLUTE VALUE OF
SALES OF CD ERRORS (DEVIATION),
YEAR PLAYERS FORECAST SALES (ACTUAL – FORECAST)
1 110 — —
2 100 110 |100 – 110| = 10
3 120 100 |120 – 110| = 10
4 140 120 |140 – 120| = 20
5 170 140 |170 – 140| = 30
6 150 170 |150 – 170| = 20
7 160 150 |160 – 150| = 10
8 190 160 |190 – 160| = 30
9 200 190 |200 – 190| = 10
10 190 200 |190 – 200| = 10
11 — 190 —
Sum of |errors| = 160
MAD = 160/9 = 17.8
Table 5.2
Measures of Forecast Accuracy
Using a naïve forecasting model
ACTUAL ABSOLUTE VALUE OF
SALES OF CD ERRORS (DEVIATION),
YEAR PLAYERS FORECAST SALES (ACTUAL – FORECAST)
1 110 — —
2 100 110 |100 – 110| = 10
3 120 100 |120 – 110| = 20
forecas
error
4
MAD
5
160
17
.
8
140
170
120
140
|140 – 120| = 20
|170 – 140| = 30
6 150 n 170 9 |150 – 170| = 20
7 160 150 |160 – 150| = 10
8 190 160 |190 – 160| = 30
9 200 190 |200 – 190| = 10
10 190 200 |190 – 200| = 10
11 — 190 —
Sum of |errors| = 160
MAD = 160/9 = 17.8
Table 5.2
Measures of Forecast Accuracy
There are other popular measures of forecast
accuracy
The mean squared error
MSE
( error ) 2
n
The mean absolute percent error
error
actual
MAPE 100%
n
And bias is the average error
Moving Averages
Moving averages can be used when demand
is relatively steady over time
The next forecast is the average of the most
recent n data values from the time series
This methods tends to smooth out short-term
irregularities in the data series
Sum of demands in previous n periods
Moving average forecast
n
Moving Averages
Mathematically
Yt Yt 1 ... Yt n1
Ft 1
n
where
Ft1 for time period t + 1
= forecast
= actualYtvalue in time period t
n = number of periods to average
Wallace Garden Supply Example
Wallace Garden Supply wants to
forecast demand for its Storage Shed
They have collected data for the past
year
They are using a three-month moving
average to forecast demand (n = 3)
Wallace Garden Supply Example
MONTH ACTUAL SHED SALES THREE-MONTH MOVING AVERAGE
January 10
February 12
March 13
April 16 (10 + 12 + 13)/3 = 11.67
May 19 (12 + 13 + 16)/3 = 13.67
June 23 (13 + 16 + 19)/3 = 16.00
July 26 (16 + 19 + 23)/3 = 19.33
August 30 (19 + 23 + 26)/3 = 22.67
September 28 (23 + 26 + 30)/3 = 26.33
October 18 (26 + 30 + 28)/3 = 28.00
November 16 (30 + 28 + 18)/3 = 25.33
December 14 (28 + 18 + 16)/3 = 20.67
January — (18 + 16 + 14)/3 = 16.00
Table 5.3
Exponential Smoothing
Exponential smoothing is easy to use and
requires little record keeping of data
It is a type of moving average
ecast = Last period’s forecast
+ (Last period’s actual demand
– Last period’s forecast)
Where is a weight (or smoothing constant) with a value
between 0 and 1 inclusive
Exponential Smoothing
Mathematically
Ft 1 Ft (Yt Ft )
here
Ft+1 = new forecast (for time period t + 1)
Ft = previous forecast (for time period t)
= smoothing constant (0 ≤ ≤ 1)
Yt = previous period’s actual demand
The idea is simple – the new estimate is the
old estimate plus some fraction of the error in
the last period
Exponential Smoothing Example
In January, February’s demand for a certain
car model was predicted to be 142
Actual February demand was 153 autos
Using a smoothing constant of = 0.20,
what is the forecast for March?
New forecast (for March demand) = 142 + 0.2(153 – 142)
= 144.2 or 144 autos
If actual demand in March was 136 autos, the
April forecast would be
New forecast (for April demand) = 144.2 + 0.2(136 – 144.2)
= 142.6 or 143 autos
USING THE PREFERRED METHOD TO
FORECAST QUARTER 4, 2010
ACTUAL EXP SMTH
QUARTER, YEAR DEMAND 4QMA @ a=0.3
QUARTER 1, 2008 8,000
QUARTER 2,2008 13,000 8,000
QUARTER 3,2008 23,000 9,500
QUARTER 4,2008 34,000 13,550
QUARTER 1, 2009 10,000 19,500 19,685
QUARTER 2,2009 18,000 20,000 16,780
QUARTER 3,2009 23,000 21,250 17,146
QUARTER 4,2009 23,000 21,250 18,902
QUARTER 1, 2010 38,000 18,500 20,131
QUARTER 2,2010 12,000 25,500 25,492
QUARTER 3,2010 12,000 24,000 21,444
QUARTER 4,2010 ?????
15
NB. The forecasting model with the
lower/lowest Aggregate forecast error is the
preferred forecasting model.
e.g. Using either the MAD, MSE or the MAPE to
determine whether to use the 4 Quarter Moving
Average or the Exponential Smoothing @ a =0.3
The Method with the lower MAD, MSE or the
MAPE is the preferred forecasting method.
16
Regression Analysis
Regression Analysis: A procedure commonly
used by economists to estimate consumer
demand with available data.
◦ Cross-Sectional Data: provide information on
variables for a given period of time.
◦ Time Series Data: give information about variables
over a number of periods of time.
Regression Analysis
In
estimating the demand for a particular
good or service, one must first determine all
the factors that might influence this demand.
◦ E.g. price & non-price determinants of demand.
- tastes & preferences, income, prices of related
goods, future expectations, number of buyers,
etc.
- N.b. Not all determinants are necessarily
appropriate – given the need for good quality
data.
Regression Analysis
Regression equation: linear, additive
◦ Y = a + b1X1 + b2X2 + b3X3 + b4X4
◦ Y: dependent variable, amount to be
determined
◦ a: constant value, y-intercept
◦ Xn: independent, explanatory variables, used
to explain the variation in the dependent
variable
◦ bn: regression coefficients (measure impact of
independent variables)
Regression equation for Pizza
Demand
Y = a + b1X1 + b2X2 + b3X3 + b4X4
Y Quantity of pizza demanded (average number of
slices per capita per month
a Constant value or Y intercept
X1 Average price of a slice of pizza
X2 Annual tuition (in thousand of dollars
X3 Average price of a 12- ounce can of soft drink
X4 Location of campus (1 if located in a
concentrated urban area, 0 if otherwise.
b1, b2 , b3,b4 Coefficients of the X variables
measuring the impact of the variables on the
demand for pizza.
Data for 10 colleges
COLLEGE # Y X1 X2 X3 X4
1 10 100 14 100 1
2 12 100 16 95 1
3 13 90 8 110 1
4 14 95 7 90 1
5 9 110 11 100 0
6 8 125 5 100 0
7 4 125 12 125 1
8 3 150 10 150 0
9 15 80 18 100 1
10 12 80 12 90 1
21
SUMMARY OUTPUT
Regression Statistics
Multiple R 0.934702
R Square 0.873667
Adjusted R Square 0.772601
Standard Error 1.933765
Observations 10
ANOVA
df SS MS F Significance F
Regression 4 129.3028 32.32569 8.644509 0.018063
Residual 5 18.69724 3.739448
Total 9 148
Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Lower 95.0% Upper 95.0%
Intercept 32.00304 5.728278 5.586852 0.002534 17.27803 46.72805 17.27803 46.72805
X Variable 1 -0.17318 0.068768 -2.51834 0.053282 -0.34995 0.003592 -0.34995 0.003592
X Variable 2 -0.07831 0.186601 -0.41966 0.692156 -0.55798 0.401365 -0.55798 0.401365
X Variable 3 -0.02085 0.063827 -0.32672 0.757117 -0.18493 0.14322 -0.18493 0.14322
X Variable 4 -0.91025 2.070162 -0.4397 0.678511 -6.23177 4.411276 -6.23177 4.411276
22
Regression Analysis – An Example Case of
Pizza Demand at some University Campuses
COLLEGE
# Y X1 X2 X3 X4
1 10 100 14 100 1 R Square 0.874
2 12 100 16 95 1
3 13 90 8 110 1
Standard Error 1.934
4 14 95 7 90 1
Coefficients
5 9 110 11 100 0
Intercept 32.00304111
6 8 125 5 100 0
7 4 125 12 125 1
X Variable 1 -0.173180264
8 3 150 10 150 0 X Variable 2 -0.078308964
9 15 80 18 100 1 X Variable 3 -0.020853396
10 12 80 12 90 1 X Variable 4 -0.910245686
Regression Equation:
Y = 32.0 - 0.17 X1 - 0.078 X2 - 0.021 X3 - 0.910 X4
Regression Analysis
Regression Results
◦ Negative coefficient shows that as the
independent variable (Xn) changes, the
quantity demanded changes in the opposite
direction. (inverse relationship)
◦ Positive coefficient shows that as the
independent variable (Xn) changes, the
quantity demanded changes in the same
direction. (positive relationship)
◦ Magnitude of regression coefficients is
measured by elasticity of each variable.
Regression Analysis
Statistical evaluation of regression results
◦ t-test: test of statistical significance of each
estimated regression coefficient of
Y = a + b1X1 + b2X2 + b3X3 + b4X4
b̂
t
SE b̂
◦ b: estimated coefficient
◦ SEb: standard error of the estimated coefficient
◦ Rule of 2: if absolute value of t is greater than 2,
estimated coefficient is significant at the 5% level
◦ If coefficient passes t-test, the variable has a true
impact on demand
t-test: test of statistical significance of
each estimated regression coefficient of
Standard b̂
t
Coefficients Error t Stat SE b̂
Intercept 32.0030 5.7283 5.5869
X Variable 1 -0.1732 0.0688 -2.5183
X Variable 2 -0.0783 0.1866 -0.4197
X Variable 3 -0.0209 0.0638 -0.3267
X Variable 4 -0.9102 2.0702 -0.4397
• Rule of 2: if absolute value of t is greater than 2, estimated
coefficient is significant at the 5% level
• If coefficient passes t-test, the variable has a true
impact on demand
Regression Equation:
Y = 32.0 - 0.17 X1 - 0.078 X2 - 0.021 X3 - 0.910 X4
Regression Analysis
Statistical evaluation of regression
results
◦ Coefficient of determination (R2):
percentage of variation in the dependent
variable (Y) accounted for by variation in all
explanatory variables (Xn)
Value ranges from 0.0 to 1.0
Closer to 1.0, the greater the explanatory power
of the regression equation
R Square 0.874
Regression Analysis
Statistical
evaluation of
regression
results
F-test: measures statistical
significance of the entire
regression as a whole (not each
coefficient)
F
8.644509169
Regression Results
Steps for analyzing regression results
◦ Check signs and magnitudes
◦ Compute elasticity coefficients
◦ Determine statistical significance
◦ Implications for management
Regression Problems
Identification Problem: The estimation
of demand may produce biased
results due to simultaneous shifting of
supply and demand curves.
Advanced estimation techniques, such
as two-stage least squares and
indirect least squares, are used to
correct this problem.
Regression Problems
Multicollinearity: two or more
independent variables are highly
correlated, thus it is difficult to
separate the effect each has on the
dependent variable.
Passing the F-test as a whole, but
failing the t-test for each coefficient is
a sign that multicollinearity exists.
A standard remedy is to drop one of
the closely related independent
variables from the regression.
Problems
Autocorrelation: also known as serial
correlation, occurs when the dependent variable
relates to the independent variable according to
a certain pattern.
Possible causes:
◦ Effects on dependent variable exist that are not
accounted for by the independent variables.
◦ The relationship may be non-linear
The Durbin-Watson statistic is used to identify
the presence of autocorrelation.
To correct autocorrelation consider:
◦ Transforming the data into a different order of
magnitude
◦ Introducing leading or lagging data
Subjects of Forecasts
Gross Domestic Product (GDP)
Components of GDP
◦ E.g. consumption expenditure, producer durable
equipment expenditure, residential construction
Industry Forecasts
◦ Sales of products across an industry
Sales of a specific product
Prerequisites of a Good Forecast
A good forecast should:
◦ be consistent with other parts of the business
◦ be based on knowledge of the relevant past
◦ consider the economic and political environment as
well as changes
◦ be timely
============================
