FORECASTING

 Forecasting
 The art and science of

predicting future events

 Forecast
 is a statement about the future

 takes into account 2 kinds of

information
 Current factors or

conditions
 Past experience in similar

situations
 Uses of Forecasts
 Plan the System
- long-range plans for new products and services

 Plan the Use of the System

- short-term and intermediate planning
 Features Common to All Forecasts
 Forecasting techniques generally assume that
same underlying causal system that existed in the
past will continue to exist in the future.
 Forecasts are rarely perfect.
 Forecasts for groups of items tend to be more
accurate than forecasts for individual items.
 Forecast accuracy decreases as the time period
covered by the forecast
Elements of a Good Forecast
 It should be timely.
 It should be accurate and the degree of
accuracy should be stated.
 It should be reliable; it should work
consistently.
 It should be in writing.
 The forecasting technique should be
simple to understand and use.
 7 Steps in the Forecasting System
1) Determine the use of the forecast.
2) Select the items to be forecasted.
3) Determine the time horizon of the forecast.
4) Select the forecasting model(s).
5) Gather the data needed to make the forecast.
6) Make the forecast.
7) Validate and implement the results (Monitor the
forecast and check if the forecast is performing in
a satisfactory manner).
Forecasting Approaches
 Qualitative Methods
- consist mainly of subjective inputs
 Quantitative Techniques
- involve either the extension of historical
data or the development of associative
models
- avoid personal biases
 Overview of Qualitative Methods
 Useful when management needs a forecast quickly when
historical data is not available especially during
introduction of new products
 Forecasts are made on the following basis:
1) Jury of Executive Opinion – has the advantage of
bringing together the knowledge and talents of a group of high
level experts or managers
2) Sales Force Composite – is a good source of information
because of direct customer contact
3) Delphi method – uses group process that allows experts to
make forecasts from anonymous responses from staff &
respondents
4) Consumer Market Survey – solicits consumer opinions; is
expensive and time-consuming
 Jury of Executive Opinion

 Involves small group of

high-level experts and
managers
 Group estimates demand
by working together
 Combines managerial
experience with statistical
models
 Relatively quick
 ‘Group-think’
disadvantage
 Sales Force Composite
 Each salesperson projects his or her sales

 Combined at district and national levels

 Sales reps know customers’ wants

 Tends to be overly optimistic

Delphi Method

 Iterative group process,

continues until consensus is
reached
 3 types of participants
 Decision makers
 Staff
 Respondents
 Consumer Market Survey

Ask customers about

purchasing plans

What consumers say, and what

they actually do are often
different

Sometimes difficult to answer

 Overview of Quantitative Methods
1. Naïve Approach
2. Moving Average Time-Series
3. Exponential Smoothing Models

4. Trend projection
5. Linear regression  Associative Model

Time-Series Models predict on the assumption that

the future is a function of the past.
Associative (or causal) Models incorporate the variables or
factors that might influence the quantity being forecast.
 Forecasting Based on
Time Series Data

 Time Series is a time-ordered sequence of

observations taken at regular intervals over a
period of time.
 Time Series Forecasting
 uses a series of past data points to make a
forecast; assumes that future values of the
series can be estimated from past values
 requires identifying the underlying behavior of
the series (pattern) after plotting the data.
 Decomposition of a Time Series
1. Trend is the gradual upward or downward
movement of the data over time.
2. Seasonality is a data pattern that repeats
itself after a period of days, weeks, months,
or quarter.
3. Cycles are patterns in the data that occur
every several years.
4. Random Variations are “blips” in the data
caused by chance and unusual situations.
 Product Demand Charted over 4 years with a
Growth Trend and Seasonality Indicated
Trend component
Seasonal peaks

D
Actual demand line
E
M
A Average demand
over four years
N
Random variation
D
Year 1 Year 2 Year 3 Year 4

T I M E
 NAÏVE METHODS
 Naïve Forecast – the forecast for any period
equals the previous period’s actual value
 Naïve Approach can be used with:
 a stable series (variations around an average) : The
last data point becomes the forecast for the next
period.
 seasonal variations : the forecast for this “season” is
equal to the value of the series last “season”.
 trend : the forecast is equal to the last value of the
series plus or minus the difference between the last
two values of the series.
EXAMPLE: Using an appropriate naïve method, predict orders for
the following day for each of the 3 products of a commercial
bakery that has recorded sales (in dozens) as shown below:
Day Blueberry Cinnamon Cupcakes
Muffins Buns
1 30 18 45
2 34 17 26
3 32 19 27
4 34 19 23
5 35 22 22
6 30 23 48
7 34 23 29
8 36 25 20
9 29 24 14
10 31 26 18
11 35 27 47
12 31 28 26
13 37 29 27
14
15
34
33
31
33
 24
22
 TECHNIQUES FOR AVERAGING
 Historical data typically contain a certain amount of
random variation, or “noise”.
 Averaging techniques smooth variations in the time-
series because the individual highs and lows in the data
offset each other when they are combined into an
average.
 Forecasts based on an average tend to exhibit less
variability than the original data.
 Averaging techniques include:
1. Moving average
2. Weighted moving average
3. Exponential smoothing
Moving Average
Technique that averages a number of recent actual values,
updated as new values become available
n
 Ai
i=1
Ft = MAn = ----------- , where
n

i = an index that corresponds to periods

n = number of periods (data points) in the moving average
Ai = actual value in period I
MA = moving average
Ft = forecast for time period t
 For example, MA3 would refer to a three-period moving
average forecast, and MA5 would refer to a five-period
moving average forecast.

Compute a three-period moving average forecast given the above

demand for shopping carts for the last five periods.
Period Age Demand
1 5 42
2 4 40
3 3 43 The 3 most
4 2 40 recent
5 1 41 demands
Solution:
43 + 41 + 40
What is F6 = MA4 ?
F6 = MA3 = ------------------ = 41.33
If Actual = 39 for Period 6,
3
what is F7 = MA3?
 Weighted Average
More recent values in a series are given more weight in computing a
forecast. [Note: The sum of the weights must be 1.00 and the
heaviest weights are assigned to the most recent values]
Example: Compute a weighted average forecast using a weight of
0.40 for the most recent period, 0.30 for the next most recent, 0.20
for the next, and 0.10 for the next.
Period Demand Solution:
1 42 F6 = .40(41) + .30(40) + .20(43) + .10(40)
2 40
= 41.0 ans.
3 43
4 40 If actual demand is 39 for Period 6, F7 = ?
5 41 F7 = .40(39) + .30(41) + .20(40) + .10(43)
= 40.2 ans.
NOTE: If four weights are given, only the four most recent data points are used.
 Exponential Smoothing
Weighted averaging method based on previous forecast plus a
percentage of the forecast error:
Next Forecast = Previous Forecast + (Actual – Previous Forecast)
where (Actual – Previous Forecast) = Forecast Error
 = percentage of the error

F t = F t-1 + (A t-1 - F t-1) , where

Ft = Forecast for period t

F t-1 = Forecast for the previous period
 = Smoothing constant
A t-1 = Actual demand or sales for the previous period

or F t = (1- )F t-1 +  A t-1

 Exponential Smoothing
 Smoothing Constant determines the quickness of forecast adjustment
to error.
 Low values of  are used when the underlying average tends to be
stable; higher values when average is susceptible to change.
 Example: Prepare a forecast using exponential smoothing with a
smoothing constant of 0.40 given the following data:
Period No. of Complaints Forecast Calculations
1 60 -
(The previous value of the series
2 65 60 is used as the starting forecast.)

3 55 62 60 + .40(65 – 60) = 62
4 58 59.2 62 + .40(55 – 62) = 59.2
5 64 58.72 59.2 + .40(58 – 59.2) = 58.72
6 ? 60.83 58.72 + .40(64 – 58.72) = 60.83
 TECHNIQUES FOR TREND
1. Trend Projection
2. Trend-Adjusted Exponential Smoothing

TRENDS:
a) Linear A Linear Trend Equation has the form:
b) Parabolic yt = a + bt , where
c) Exponential t = specified no. of time periods from t = 0
yt = Forecast for period t
a = value of yt at t = 0
b = slope of the line
 Linear Trend Equation
The coefficients of the line, a and b, can be computed from historical
data using these two equations:
n ty -  t y y
b = ---------------------
n t2 – ( t)2

y - b  t
y y
a = -------------------
t b = t
n
where: a
n = number of periods
y = value of the series 0 t
 Example: Develop a line trend equation for the
following data. Then use the equation to predict the
next two values of the series.
n ty -  t y
Period (t) t2 Demand (y) ty b = n t2 – ( t)2
1 1 44 44
2 4 52 104 9(2,545) – 45(488)
= 1.75
= 9(285) – (45)2
3 9 50 150
y - b  t
4 16 54 216
a = n
5 25 55 275
= 488 – 1.75(45) = 45.47
6 36 55 330
9
7 49 60 420
y = 45.47 + 1.75 t
8 64 56 448
y10 = 45.47 + 1.75 (10) = 62.97
9
_______ 81
_________ 62
_________ 558
____________
 t = 45  t2 =285  y = 488  ty = 2,545 y11 = 45.47 + 1.75 (11) = 64.72
 Trend-Adjusted Exponential Smoothing
 Variation of exponential smoothing used when a time series
exhibits trend.
 Simple exponential smoothing fails to respond to trends as
it will exhibit a severe lag from the actual data points as the
period t increases (as shown in the table below).
Month Actual Demand Forecast for Month t (Ft) w/  =0.40
1 100 F1 = 100 (given)
2 200 F2 = 100 + 0.40(100 – 100) = 100
3 300 F3 = 100 + 0.40(200 – 100) = 140
4 400 F4 = 140 + 0.40(300 – 140) = 204
5 500 F5 = 204 + 0.40(400 – 204) = 282
 Trend-Adjusted Exponential Smoothing
 The idea is to compute an exponentially smoothed average of the data and
then adjust for positive or negative lag in trend. Thus, the new formula is
Forecast including trend (FITt)= exponentially smoothed forecast (Ft)
+ exponentially smoothed trend (Tt)
 With this model, estimates for both the average and the trend are smoothed.
This procedure requires two smoothing constants,  for the average and  for
the trend. Then, the average and trend are computed for each period:

Ft =  (Actual demand last period) + (1 -  ) (Forecast last period +

Trend estimate last period)
or Ft =  (A t-1) + (1 -  ) (F t-1 + T t-1)

Tt =  (Forecast this period – Forecast last period)

+ (1 - ) (Trend estimate last period)
or Tt =  (Ft - F t-1) + (1 - ) T t-1
 Trend-Adjusted Exponential Smoothing
3 Steps in Computing a Trend-Adjusted Forecast:

1) Compute Ft , the exponentially smoothed forecast for

period t, using the equation
Ft =  (A t-1) + (1 -  ) (F t-1 + T t-1)

2) Compute the smoothed trend Tt using the equation,

Tt =  (Ft - F t-1) + (1 - ) T t-1

3) Calculate the forecast including trend, FITt = Ft + Tt

 Example: A large Portland manufacturer uses exponential
smoothing to forecast demand for a piece of pollution control
equipment. It appears that an increasing trend is present.

Month (t) Actual Month (t) Actual

Demand (At) Demand (At)

1 12 6 21
2 17 7 31
3 20 8 28
4 19 9 36
5 24 10 ?

Smoothing constants are assigned the values of  = 0.2 and  = 0.4.

Assume the initial forecast for month 1 (F1) was 11 units and the trend
over that period (T1) was 2 units.
 Trend-Adjusted Exponential Smoothing
Solution to Sample Problem

 Do the 3-step calculations for Month 2

Step 1 : Forecast for month 2

F2 =  (A 1) + (1 -  ) (F 1 + T 1)
F2 = 0.2(12) + (1 – 0.2)(11 + 2)
= 2.4 + 0.8(13) = 2.4 + 10.4 = 12.8
Step 2 : Compute the trend in period 2
T2 =  (F2 - F1) + (1 - ) T1
T2 = 0.4(12.8 – 11) + (1 – 0.4)2
= 0.4(1.8) + 0.6(2) = 0.72 + 1.2 = 1.92
Step 3 : Compute the forecast including trend (FIT2)
FIT2 = F2 + T2
= 12.8 + 1.92 = 14.72
 Trend-Adjusted Exponential Smoothing
Solution to Sample Problem (con’t)

 Do the same calculations for the third month

Step 1 : Forecast for month 3

F3 =  (A 2) + (1 -  ) (F 2 + T 2)
F3 = 0.2(17) + (1 – 0.2)(12.8 + 1.92)
= 3.4 + 0.8(14.72) = 3.4 + 11.78 = 15.18
Step 2 : Compute the trend in period 3
T3 =  (F3 – F2) + (1 - ) T2
T3 = 0.4(15.18 – 12.8) + (1 – 0.4)1.92
= 0.4(2.38) + 0.6(1.92) = 0.952 + 1.152 = 2.10
Step 3 : Compute the forecast including trend (FIT3)
FIT3 = F3 + T3
= 15.18 + 2.10 = 17.28
 Trend-Adjusted Exponential Smoothing
Solution to Sample Problem (con’t)
The following table completes the forecast for the 10-month period:
Actual Smoothed Smoothed Forecast
Month Demand Forecast, Ft Trend, Tt including Trend, FITt
1 12 11 2 -
2 17 12.80 1.92 14.72
3 20 15.18 2.10 17.28
4 19 17.82 2.32 20.14
5 24 19.91 2.23 22.14
6 21 22.51 2.38 24.89
7 31 24.11 2.07 26.18
8 28 27.14 2.45 29.59
9 36 29.28 2.32 31.60
10 - 32.48 2.68 35.16
 TECHNIQUES FOR SEASONALITY
 Seasonal Variations are regularly repeating upward or
downward movements in series values that can be tied up
to recurring events.
 Six common seasonality patterns
Period of “Season” Number of
Pattern Length “ Seasons ” in Pattern
Week Day 7
Month Week 4–4½
Month Day 28 – 31
Year Quarter 4
Year Month 12
Year Week 52
 TECHNIQUES FOR SEASONALITY
 Seasonality in a time series is expressed in terms of
a) Moving Average, if the series tends to vary around
an average value
b) Trend Value, if the series exhibits a trend

 2 Models of Seasonality
1) Additive Model – seasonality is expressed as a
quantity, w/c is added or subtracted from the series
average in order to incorporate seasonality.
2) Multiplicative Model – seasonality is expressed as a
percentage of the average (or trend), w/c is used to
multiply the value of a series to incorporate
seasonality.
 Multiplicative Seasonal Model
 is more widely used in business than the
additive model
 entails computation of
 a seasonal index (when time series has
seasonality and trend is not present), or
 a seasonal relative (when data has both
trend/average and seasonal components)
 Steps in Making a Forecast Using a
Seasonal Index (Data w/o Trend)

1) Find the average historical demand each season.

2) Compute the average demand over all seasons.
3) Compute a seasonal index for each season.
4) Estimate next period’s total volume.
5) Do the seasonal forecast by dividing the volume
by the number of seasons and then multiplying it
by the seasonal index for that season.
 Use of Seasonal Index
Example: Monthly sales of Compaq laptop computers at Des
Moines distributor for 2001 – 2003 are shown below. Develop a
monthly forecast for 2004 if total demand for said year is estimated
No. of Seasons
to be 1,200 units. Average Monthly
Average
Sales Demand 2001-2003 Monthly Seasonal Demand
Month 2001 2002 2003 Demand
 Demand = Index Forecast
Jan 80 85 105 90 94 .957 96
Feb 70 85 85 80 94 .851 85
Mar 80 93 82 85 94 .904 90
Apr 90 95 115 100 94 1.064 106
May 113 125 131 123 94 1.309 131
Jun 110 115 120 115 1,128 = 94 1.223 122
Jul 100 102 113 105 12 months 94 1.117 112
Aug 88 102 110 100 94 1.064 106
Sep 85 90 95 90 94 .957 96
Oct 77 78 85 80 94 .851 85
Nov 75 82 83 80 94 .851 85
Dec 82 78 80 80 94 .851 85
 Using Seasonal Relatives
 Incorporating seasonality is useful when data has both trend and seasonality,
and can be accomplished by the following steps:
1) Obtain trend estimates for desired periods using a trend equation.
2) Add seasonality to the trend estimates by multiplying these trend estimates by
the corresponding seasonal relative.
 Example : A furniture maker wants to predict quarterly demand for a certain
loveseat for periods 15 and 16, w/c happens to be the 2nd and 3rd quarters of
a particular year. The trend portion of demand is projected using the
equation yt = 124 + 7.5 t. Quarter relatives are Q1 = 1.20, Q2 = 1.10,
Q3 = 0.75, and Q4 = 0.95.
Solution : At t = 15, y15 = 124 + 7.5(15) = 236.5
At t = 16, y16 = 124 + 7.5(16) = 244.0
Multiplying the trend values above by the appropriate quarter relative
yields the 2nd and 3rd quarter forecasts,
Period 15 : 236.5(1.10) = 260.15
Period 16 : 244.0(0.75) = 183.00
 Computing Seasonal Relatives
 Get the Centered Moving Average
 same computations as those for a moving average forecast but
positioned in the middle of the periods used to compute the moving
average Three-Period Seasonal
Period Demand Centered Average Relative _____
1 40
2 46 42.67 1.08
3 42
MA3= 40 + 46 + 42 Ratio = 46 / 42.67
3
 Compute the Seasonal Relative
 the ratio of the demand at period 2 to the CMA at that point
 Achieve a reasonable estimate of seasonality for any season
 compute seasonal ratios for a number of seasons
 average these ratios
 Computing Seasonal Relatives
Example : A parking lot manager has the following record of the
number of cars per day in the lot for the past three weeks.
Estimate a Friday relative as well as a Tuesday relative.

Week 1 Week 2 Week3

Day Volume Day Volume Day Volume
Tues 67 Tues 60 Tues 64
Wed 75 Wed 73 Wed 76
Thurs 82 Thurs 85 Thurs 87
Fri 98 Fri 99 Fri 96
Sat 90 Sat 86 Sat 88
Sun 36 Sun 40 Sun 44
Mon 55 Mon 52 Mon 50
 Computing Seasonal Relatives
Solution : Since there are seven days (seasons) per
week, a seven-period centered moving average is used.
Centered Daily Ratios =
Day Volume Moving Total MA7 Volume /MA
Tue 67
Wed 75
Thurs 82
Fri 98 71.86 98  71.86 = 1.36 (Friday)
Sat 90 70.86 90  70.86 = 1.27
Sun 36 70.57 36  70.57 = 0.51
Mon 55 503  7 = 71.00 55  71.00 = 0.77
Tue 60 496  7 = 71.14 60  71.14 = 0.84 (Tuesday)

Wed 73 494  7 = 70.57 73  70.57 = 1.03

Thurs 85 497  7 = 71.14 85  71.14 = 1.19
Fri 99 498 etc. 70.71 99  70.71 = 1.40 (Friday)
 Computing Seasonal Relatives
Solution : Add the Friday seasonal ratios and average
them to estimate the Friday relative. Do the same for
the Tuesday ratios to obtain the relative for Tuesday.
Sat 86 494 71.29 86  71.29 = 1.21
Sun 40 498 71.71 40  71.71 = 0.56
Mon 52 495 72.00 52  72.00 = 0.72
Tue 64 499 71.57 64  71.57 = 0.89 (Tuesday)
Wed 76 502 71.86 76  71.86 = 1.06
Thurs 87 504 72.43 87  72.43 = 1.20
Fri 96 501 72.14 96  72.14 = 1.33 (Friday)
Sat 88 503
Sun 44 507

Mon 50 505
Friday Relative = (1.36 + 1.40 + 1.33)/3 = 1.36
Tuesday Relative = (0.84 + 0.89) / 2 = 0.87
 TECHNIQUES FOR CYCLES
 CYCLES are patterns in the data that occur every several years.
[Forecasting them is difficult because it is very hard to predict the
turning points that indicate a new cycle is beginning.]
 The best way to predict business cycles is finding a leading variable
with which the data series seems to correlate. Examples are
 Birthrates “lead” college enrollments by about 18 years, so changes in
birthrates is a good predictor of swings in enrollment.
 Housing construction permits are excellent leading variable for household
appliances and services.
 ASSOCIATIVE FORECASTING METHODS consider variables that
are related to the quantity being predicted. This approach is more
powerful than the time-series methods that use only the historic
values for the forecasted variable.
1. Regression Analysis
2. Correlation Analysis
 Associative Forecasting Techniques
 The essence is to develop the best statistical relationship between the
variable of interest or being forecast (w/c is the Dependent Variable)
and the related variables that can be used to predict values of the
variable of interest (called Predictor Variables).
 For example, the sales of IBM PCs might be related to IBM’s
advertising budget, the company’s prices, competitors’ prices and
promotional strategies, and even the nation’s economy and
unemployment rates. In this case PC sales would be the dependent
variable and the others would be called the independent variables.
 The primary method of analysis is known as regression. It is a
technique for fitting a line to a set of points.
a) Simple Linear Regression – linear relationship between 2 variables
b) Curvilinear Regression – non linear relationships are present
c) Multiple Regression – involves more than one predictor variable
 Linear Regression Analysis
 A straight-line mathematical model to describe the functional relationships between
independent and dependent variables.
 Simple Linear Regression : to obtain an equation of a straight line that minimizes
the sum of squared vertical deviations of data points from the line.
This least squares line has the equation: Computed
y Relationship
yc = a + bx
where,
yc = predicted (dependent) variable
x = predictor (independent) variable
b = slope of the line
a = y-intercept
n xy -  x y 0 Predictor variable
x

b = n x2 – ( x)2 y

y - b  x
a = n y
_ _ b= t
or y - bx
n = no. of paired observations 0 The line intersects the y-axis where y =a. the slope of the line = b. x
 Linear Regression Analysis
Example : Nodel Construction Company renovates old homes in Orono,
Maine. Over time, the company has found that its dollar volume of
renovation work is dependent on the Orono area payroll.
The following table lists Nodel’s revenues (in hundred thousand dollars) and the
amount of money earned by wage earners (in $ dundreds of millions) in Orono
during the past six years. A linear relationship was found after plotting the data,
so management wants to establish a mathematical model to help predict sales.
Sales, y Payroll, x x2 xy n xy -  x y
2.0 1 1 2.0 b = n x2 – ( x)2 =
3.0 3 9 9.0
2.5 4 16 10.0 6(51.5) – (18)(15.0) = 0.25
2.0 2 4 4.0 6(80) – (18)2
2.0 1 1 2.0
3.5
________ 7
________ 49
________ 24.5
_________ y - b  x
a = n or
 y = 15.0  x = 18  x2 = 80  xy = 51.5
_ _
The least squares line, therefore, is yc = 1.75 + 0.25x y – b x
or Sales = 1.75 + 0.25 (payroll). So, if the local
chamber of commerce predicts Orano area payroll will be = 15 - .25(18)
$600 million next year, Nodel can estimate its sales to be 6 6
= 1.75 + .25(6) = 3.25 or Sales = $325,000. = 1.75
 Linear Regression Analysis
 Three (3) conditions for an indicator to be valid:
1) The relationship between the movement s of an indicator and the
movements of the variable should have a logical explanation.
2) Movements of the indicator must precede movements of the dependent
variable by enough time so that the forecast isn’t outdated before it can
be acted upon.
3) A fairly high correlation should exist between the two variables

 Basic assumptions in the use of linear regression:

1) Variations around the line are random (no patterns such as cycles or
trends exist when the line and data are plotted).
2) Deviations around the line should be normally distributed. (A
concentration of values close to the line with a small proportion of larger
deviation supports the assumption of normality.)
3) Predictions are being made only within the range of observed values.
 Correlation Analysis
 COEFFICIENT of CORRELATION (r) measures the strength and direction of
relationship between two variables. Correlation can range from –1.00 to +1.00
 r = +1.00 indicates that changes in one variable are always matched by changes in
the other
 r = -1.00 indicates that increases in one variable are matched by decreases in the
other
 r 0 indicates little linear relationship between two variables

 Correlation Formula is r =__ n (xy) – (  x)(  y)

 n ( ) – (  x)2   n ( y2 ) – (  y)2
x2
 COEFFICIENT of DETERMINATION (r2) provides a measure of the amount
of variation in the dependent variable about the mean that is explained by the
independent variable in the regression equation.
 The possible values of r2 range from 0 to 1.00. The closer r2 is to 1.00, the greater the
percentage of explained variation. A high value of r2, say 0.80 or more would indicated
that the independent variable is a good predictor of values of the dependent variable.
 A low value, say 0.25 or less, would indicate a poor predictor,and a value between
0.25 and 0.80 would indicate a moderate predictor.
 Correlation Analysis
Example : Using the earlier data for Nodel Construction
Company, assess the relationship between the firm’s renovation
sales and payroll in its hometown of Orono, Maine.

Sales,y Payroll, x x2 xy y2
2.0 1 1 2.0 4.0
3.0 3 9 9.0 9.0
2.5 4 16 10.0 6.25
2.0 2 4 4.0 4.0
2.0 1 1 2.0 4.0
3.5 7 49 24.5 12.25
________ ________ ________ _________ _________
 y = 15.0  x = 18  x2 = 80  xy = 51.5  y2 = 39.5

r= n (xy) – (  x)(  y)
 n ( x2 ) – (  x)2   n ( y2 ) – (  y)2

r= 6 (51.5) – ( 18 )( 15.0 ) = 0.901

 6 ( 80 ) – ( 18 )2   6 (39.5 ) – ( 15.0 )2
This r of .901 appears to be a significant correlation and helps to confirm the
closeness of the relationship between the two variables. With r 2 = .81,
payroll is a good predictor of Nodel Construction Company’s renovation sales.
 ACCURACY of FORECASTS
 Accuracy is based on historical error performance of a
forecast.
 Forecast error is the difference between the actual value
and the value predicted for a given period. Hence, Error =
Actual – Forecast or (et = At – Ft)
 Two commonly used measures for summarizing historical
errors are:
Mean Absolute Deviation (MAD) =  Actual - Forecast 
n
Mean Squared Error (MSE) =  (Actual - Forecast )2
n-1
 Forecast Accuracy is a significant factor in choosing
forecasting alternatives (by determining which yields the
lowest MAD or MSE for a given set of data).
 Example: The manager of a large manufacturer of industrial
pumps must choose between two alternative forecasting
techniques to prepare forecasts for a six month period. Using MAD
and MSE as criteria, which technique is better?

Forecast Forecast
Month Demand Technique 1 e e e2 Technique 2 e e e2
1 492 488 4 4 16 495 -3 3 9
2 470 484 -14 14 156 482 -12 12 144
3 485 480 5 5 25 478 7 7 49
4 493 490 3 3 9 488 5 5 25
5 498 497 1 1 1 492 6 6 36
6 492 493 -1 ____1 1 493 -1 1 1
___ ____ ___ ____ ____
-2 28 208 2 34 264
Check that each forecast has an average error of approximately zero.
MAD1 =   e  = 28 MSE1 =  (e )2 = 208
= 4.67 = 41.6
n 6 n-1 6-1
MAD2 =   e  = 34 MSE2 =  (e )2 = 264 = 52.8
= 5.67
n 6 n-1 6-1
Technique 1 is superior in this comparison because its MAD and MSE are smaller.
 MONITORING and CONTROLLING FORECASTS
 It is necessary to monitor forecast errors to ensure that the forecast is
performing adequately.
 A forecast is generally deemed to perform adequately when the errors
exhibit only random variations.
 Forecasts can be monitored using either of the following:
Tracking Signal – measurement of how well the forecast is predicting
actual values. It is computed as the running sum of forecast errors (RSFE)
divided by the corresponding value of mean absolute deviation (MAD), or
Tracking = RSFE =  (Actual - Forecast ) , where MAD =  Actual - Forecast 
Signal MAD MAD n

Control Chart – monitoring approach that sets limits for individual

forecast errors (instead of cumulative errors). The limits are multiples of
the square root of MSE, that is, s =  MSE , where MSE =  (Actual – Forecast)2
n-1
 Tracking Signal
Example: Rick Carlson Bakery’s quarterly sales of croissants (in
thousands), as well as forecast demand error computations, are shown
below. Compute the tracking signal and determine whether forecasts are
performing adequately.
Cumulative
Absolute Absolute n = Tracking
Forecast Actual A – F = Forecast Forecast Cumulative Signal
Quarter Demand Demand Error RSFE  Error   Error  MAD (RSFE/MAD)
1 100 90 -10 -10 10 10 10.0 -10/10 = -1
2 100 95 -5 -15 5 15 7.5 -15/7.5 = -2
3 100 115 +15 0 15 30 10.0 0/10 = 0
4 110 100 -10 -10 10 40 10.0 -10/10 = -1
5 110 125 +15 +5 15 55 11.0 +5/11 = +0.5
6 110 140 +30 +35 30 85 14.2 +35/14.2 = +2.5

At the end of quarter 6, MAD =  Actual – Forecast = 85 = 14.2

n 6
And Tracking signal = RSFE = 35 = 2.5 MADs
MAD 14.2
This tracking signal is within acceptable limits. We see that it drifted
from –2.0 MADs to +2.5 MADs
 Tracking Signal & Control Chart
 Bias in forecasts is reflected by the cumulative forecast error and is
defined as the persistent tendency for forecasts to be greater or
less than the actual value of a time series.
 Tracking Signal values are compared to predetermined limits (or
called “action limits” based on judgment and experience. They often
range from 3 to 8; for the most part, limits of 4 shall be used
which are roughly comparable to three (3) standard deviation limits.
Values within the limits suggest – but do not guarantee – that the
forecast is performing adequately.
 MAD can be updated using exponential smoothing after an initial
value of MAD has been computed:
MADnew = MADold +  (Actual - Forecast  - MADold )
 Control Chart method assumes the following:
1.Forecast errors are randomly distributed around a mean of zero.
2.The distribution of errors is normal.
 CHOOSING A FORECASTING TECHNIQUE
Factors to consider in choosing a forecasting technique:
1. Cost – how much money is budgeted for generating the forecast?
2. Accuracy – generally speaking, the higher the accuracy, the higher
the cost, so it’s important to weigh the cost-accuracy trade-offs
carefully. The best forecast is not necessarily the most accurate or
the least costly; rather, it is some combination of accuracy and cost
deemed by management.
3. Availability of historical data
4. Availability of computers
5. Ability of decision makers to utilize certain techniques
6. Time needed to gather and analyze data and prepare the forecast
7. Any prior experience with a technique