Forecasting

STEVEN CO JR., CIE, AAE

What is Forecasting?

 Forecasts are a basic input in the decision process of OM

 Forecasts provide information on future demand
 Businesses make plans for future operations based on anticipated
future demand
 For many businesses, anticipated demand is derived from forecasts
 Two important aspects of forecasts:
 Expected level of demand
 Degree of accuracy of a forecast
What is Forecasting?

 Forecasts are made with reference to a specific time horizon

 Some examples of uses of forecasts include:
 Accounting & Finance
 Human resources
 Marketing
 MIS
 Operations
 Product/service design
What is Forecasting?

 Forecasting is an important component of yield management

 Yield management refers to the percentage of capacity being used
 Two main uses of forecasts:
 Plan the system
 Plan the use of the system
Service Forecasting

 Service forecasting differs with manufacturing forecasting as it

presents unusual challenges:
 Special needs for short term records
 Needs differ greatly as function of industry and product
 Holidays and other calendar events
 Unusual events
Product Life Cycle

 Introduction > Growth > Maturity > Decline

 Introduction and growth require longer forecasts than maturity and
decline
Features of Forecasting

 Forecasting techniques generally assume that the same underlying

causal system that existed in the past will continue to exist in the
future.
 Forecasts are not perfect
 Forecasts for groups of items tend to be more accurate than
forecasts for individual items due to the cancelling effect of errors in
a group
 Forecast accuracy increases as time period covered increases
Elements of a Good Forecast

 Timely
 Accurate
 Reliable
 Expressed in meaningful units
 In writing
 Simple to understand and use
 Cost-effective
Steps in Forecasting

 Determine the purpose of the forecast

 Establish a time horizon
 Obtain, clean, and analyze appropriate data
 Select a forecasting technique
 Make the forecast
 Monitor the forecast errors
Forecasting Time Horizons

 Short-range
 Up to 1 year, generally less than 3 months
 Usually employ different methodologies
 Tend to be more accurate than longer-term forecasts
 Medium-range
 3 months to 3 years
 Deal with more comprehensive issues and support management
decisions regarding planning and products, plants and processes
 Long-range
 3+ years
Forecasting Approaches

 Judgmental/qualitative forecasts
 Rely on analysis of subjective inputs obtained from various sources
 Time-series forecasts
 Attempt to project past experiences into the future
 Associative models
 Use equations that consist of one or more explanatory variables that
can be used to predict demand
Qualitative Forecasts

 Executive opinions
 Small group of upper-level managers may meet and collectively
develop a forecast
 Salesforce opinions
 Members of sales staff or customer service staff develop the forecast
 Consumer surveys
 Consumers dictate demand through surveys
 Delphi method
 Managers and staff complete a series of questionnaires to achieve a
consensus forecast
Quantitative Forecasts

 Time-series models
 Uses a set of evenly spaced numerical data obtained by observing
response variable at regular time periods
 Forecast based only on past values, no other variables important
 Assumes that factors influencing past and present will continue
influence in future
 Associative model
 Used when changes in one or more independent variables can be used
to predict the changes in the dependent variable
Quantitative Forecasts

 Time-series models
 Naïve approach
 Moving averages
 Exponential smoothing
 Trend projection
 Associative model
 Linear regression
Quantitative Forecasts

 Components of Time-series models

 Trend
 persistent, overall upward or downward
 Changes due to population, technology, age, culture, etc.
 Typically several years duration
 Seasonal
 Regular pattern of up and down fluctuations
 Due to weather, customs, etc.
 Occurs within a single year
Quantitative Forecasts

 Components of Time-series models

 Cyclical
 Repeating up and down movements
 Affected by business cycle, political, and economic factors
 Multiple years duration
 Often causal or associative relationships
 Random
 Erratic, unsystematic, residual fluctuations
 Due to random variation or unforeseen events
 Short duration and nonrepeating
Naïve Approach

 Assumes demand in the next period is the same as the demand in

the most recent period
 Example: If January sales were 50, February sales will be 50
 Sometimes cost-effective and efficient
 Good starting point
Moving Average Method

 Series of arithmetic means

 Used if little or no trend
 Used often for smoothing
 Provides overall impression of data over time

σ 𝑑𝑒𝑚𝑎𝑛𝑑 𝑖𝑛 𝑝𝑟𝑒𝑣𝑖𝑜𝑢𝑠 𝑛 𝑝𝑒𝑟𝑖𝑜𝑑𝑠

𝑀𝑜𝑣𝑖𝑛𝑔 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 =
𝑛
Moving Average Method

Sample problem:
 Donnalyn’s Office Space Rental services wants a 3-month moving
average forecast, including a forecast for August for office space
rentals.
Month Actual sales
Jan 10
Feb 12
Mar 13
Apr 16
May 19
Jun 23
Jul 26
Moving Average Method

Answer:

Month Actual sales 3-month Moving Average

Jan 10 -
Feb 12 -
Mar 13 -
Apr 16 (10+12+13)/3 = 11.6667
May 19 (12 + 13 + 16)/3 = 13.6667
Jun 23 (13 + 16 + 19)/3 = 16
Jul 26 (16 + 19 + 23)/3 = 19.3333
Aug (19 + 23 + 26)/3 = 22.6667
Moving Average Method

 Potential problems with Moving Average

 Increasing n smooths the forecast but makes it less sensitive to changes
 Do not forecast trends well
 Require extensive historical data
Weighted Moving Average
Method
 Used when trend is present
 Older data are usually less important
 Weights based on experience and intuition

σ(𝑤𝑒𝑖𝑔ℎ𝑡 𝑓𝑜𝑟 𝑝𝑒𝑟𝑖𝑜𝑑 𝑛)(𝑑𝑒𝑚𝑎𝑛𝑑 𝑖𝑛 𝑝𝑒𝑟𝑖𝑜𝑑 𝑛)

𝑀𝑜𝑣𝑖𝑛𝑔 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 =
σ 𝑤𝑒𝑖𝑔ℎ𝑡𝑠
Weighted Moving Average
Method
Sample problem:
 Donnalyn’s Office Space Rental services wants to forecast sales by
weighing the past months, with more weight given to recent data to
make them more significant.
Month Actual sales
Weights applied:
Jan 10 3 – last month
Feb 12 2 – 2 months ago
Mar 13 1 – 3 months ago

Apr 16 6 = sum of weights

May 19
Jun 23
Jul 26
 Weighted Moving Average
Method
Answer:
Month Actual 3-month Moving Average Weights applied:
sales 3 – last month
Jan 10 - 2 – 2 months ago
Feb 12 - 1 – 3 months ago

Mar 13 - 6 = sum of weights

Apr 16 [(3*13)+(2*12)+(1*10)]/6 = 12.1667
May 19 [(3*16)+(2*13)+(1*12)]/6 = 14.3333
Jun 23 [(3*19)+(2*16)+(1*13)]/6 = 17
Jul 26 [(3*23)+(2*19)+(1*16)]/6 = 20.5
Aug [(3*26)+(2*23)+(1*19)]/6 = 23.8333
Exponential Smoothing
 Form of weighted moving average
 Weights decline exponentially
 Most recent data weighted most
 Requires smoothing constant (𝛼)
 Ranges from 0 to 1, subjectively chosen
 Involves little record keeping of past data

𝐹𝑡 = 𝐹𝑡−1 + 𝛼(𝐴𝑡−1 − 𝐹𝑡−1 )

 Where:
 Ft = new forecast
 Ft-1 = previous forecast
 At-1 = previous actual
Exponential Smoothing
 Impact of using different 𝛼
 Choose high values when underlying average is likely to change
 Choose low values when underlying average is stable
Exponential Smoothing
Sample problem:
 In February, a car dealer predicted March demand for 142 cars.
Actual March demand was 153 cars. Using a smoothing constant
chosen by the management of α = .20, the dealer wants to forecast
April demand using the exponential smoothing model.
Exponential Smoothing
Answer:
 Using 𝐹𝑡 = 𝐹𝑡−1 + 𝛼(𝐴𝑡−1 − 𝐹𝑡−1 )

𝐴𝑝𝑟𝑖𝑙 𝑓𝑜𝑟𝑒𝑐𝑎𝑠𝑡 = 142 + 0.20 153 − 142 = 144.20 𝑐𝑎𝑟𝑠

Choosing the Smoothing Constant
 The objective is to obtain the most accurate forecast no matter the
technique
 This is done by selecting the model with the lowest forecast error

𝐹𝑜𝑟𝑒𝑐𝑎𝑠𝑡 𝑒𝑟𝑟𝑜𝑟 = 𝐴𝑡 − 𝐹𝑡
Common Measures of Error
 Mean Absolute Deviation (MAD)
 How much the forecast missed the target
σ 𝐴𝑐𝑡𝑢𝑎𝑙 − 𝐹𝑜𝑟𝑒𝑐𝑎𝑠𝑡
𝑀𝐴𝐷 =
𝑛
 Mean Squared Error (MSE)
 The square of how much the forecast missed the target
σ(𝐹𝑜𝑟𝑒𝑐𝑎𝑠𝑡 𝑒𝑟𝑟𝑜𝑟𝑠)2
𝑀𝑆𝐸 =
𝑛
 Mean Absolute Percent Error (MAPE)
 The average percent error
σ𝑛𝑖=1(100) 𝐴𝑐𝑡𝑢𝑎𝑙𝑖 − 𝐹𝑜𝑟𝑒𝑐𝑎𝑠𝑡𝑖 /𝐴𝑐𝑡𝑢𝑎𝑙𝑖
𝑀𝐴𝑃𝐸 =
𝑛
Common Measures of Error
Sample problem:
 During the past 8 quarters, the Port of Baltimore has unloaded large
quantities of grain from ships. The port’s operations manager wants
to test the use of exponential smoothing to see how well the
technique works in predicting tonnage unloaded. He guesses that
the forecast of grain unloaded in the first quarter was 175 tons. Two
values of α are to be examined: .10 and .50.
Common Measures of Error
Common Measures of Error
Answer:
Common Measures of Error
Answer:
 MAD:
 For = 0.10
 82.45/8 = 10.3063
 For = 0.50
 98.62/8 = 12.3275
Common Measures of Error
Answer:
Absolute Error2 Absolute Error2
deviation for = deviation for =
0.10 0.50
5 25 5 25.00
7.5 56.25 9.5 90.25
15.75 248.06 13.75 189.06
1.82 3.31 9.12 83.17
16.64 276.89 19.56 382.59
29.98 898.80 24.78 614.05
1.98 3.92 12.61 159.01
3.78 14.29 4.3 18.49
Total 1526.52 Total 1561.63
Common Measures of Error
Answer:
 MSE:
 For = 0.10
 1526.54/8 = 190.8175
 For = 0.50
 1561.63/8 = 195.2038
Common Measures of Error
Answer:
Actual Absolute (100)Error/Actual Absolute (100)Error/Actual
deviation for = deviation for =
0.10 0.50
180 5 2.78 5 2.78
168 7.5 5.65 9.5 4.46
159 15.75 8.65 13.75 9.91
175 1.82 5.21 9.12 1.04
190 16.64 10.29 19.56 8.76
205 29.98 12.09 24.78 14.62
180 1.98 7.01 12.61 1.10
182 3.78 2.36 4.3 2.08
Total 44.75 Total 54.04
Common Measures of Error
Answer:
 MAPE:
 For = 0.10
 44.75/8 = 5.59%
 For = 0.50
 54.04/8 = 6.76%
Common Measures of Error
Answer:
 For = 0.10
 MAD = 10.3063
 MSE = 190.8175
 MAPE = 5.59%
 For = 0.50
 MAD = 12.3275
 MSE = 195.2038
 MAPE = 6.76%
 Therefore, = 0.10 is better
Exponential Smoothing with Trend
Adjustment
 When a trend is used during exponential smoothing
 Forecast including trend (FIT) = Exponentially smoothed forecast (F)
+ Exponentially smoothed trend (T)
𝐹𝑡 = 𝛼 𝐴𝑡−1 + 1 − 𝛼 𝐹𝑡−1 + 𝑇𝑡−1

𝑇𝑡 = 𝛽 𝐹𝑡 − 𝐹𝑡−1 + 1 − 𝛽 𝑇𝑡−1
 Steps to compute:
 Compute Ft
 Compute Tt
 Calculate the forecast Ft + Tt
Exponential Smoothing with Trend
Adjustment
Sample problem:
 A large Portland manufacturer wants to forecast demand for a
piece of pollution-control equipment. A review of past sales, as
shown below, indicates that an increasing trend is present.
Smoothing constants are assigned the values of α = .2 and β = .4.
The firm assumes the initial forecast for month 1 (F1) was 11 units and
the trend over the period (T1) was 2 units. Determine the 10th month
forecast.
Exponential Smoothing with Trend
Adjustment
Sample problem:
Month Actual demand
1 12
2 17
3 20
4 19
5 24
6 21
7 31
8 28
9 36
10
Exponential Smoothing with Trend
Adjustment
Answer:
Month Actual demand Ft Tt FITt
1 12 11 2 13
2 17 12.8000 1.9200 14.7200
3 20 15.1760 2.1024 17.2784
4 19 17.8227 2.3201 20.1428
5 24 19.9143 2.2287 22.1430
6 21 22.5144 2.3773 24.8916
7 31 24.1133 2.0659 26.1792
8 28 27.1434 2.4516 29.5950
9 36 29.2760 2.3240 31.6000
10 32.4800 2.6760 35.1560
Least Squares Method

 Trend projections
 Fit a trend line to historical data points to project into medium to long-
range
 Linear trends can be found using the least squares method
𝑦ො = 𝑎 + 𝑏𝑥
 Where:
 𝑦ො = computed value of the variable to be predicted
 a = y-axis intercept
 b = slope of the regression line
 x = the independent variable
Least Squares Method

 Least squares method requirement

 We always plot the data to ensure a linear relationship
 We do not predict time periods far beyond the database
 Deviations around the least squares line are assumed to be random
Least Squares Method

 Equations to calculate the regression variables

𝑦ො = 𝑎 + 𝑏𝑥

σ 𝑥𝑦 − 𝑛𝑥ҧ 𝑦ത
𝑏=
σ 𝑥 2 − 𝑛𝑥ҧ 2

𝑎 = 𝑦ത − 𝑏𝑥ҧ
Least Squares Method
Sample problem:
 Demand for electric power at NY Edison over the period 2001 to
2009 is shown in the following table, in megawatts. The firm wants to
forecast 2010 demand by fitting a straight-line trend to these data.

Year Demand
2001 74
2002 79
2003 80
2004 90
2005 105
2006 142
2007 122
Least Squares Method
Answer:

Year Time period (x) Demand x2 xy

2001 1 74 1 74
2002 2 79 4 158
2003 3 80 9 240
2004 4 90 16 360
2005 5 105 25 525
2006 6 142 36 852
2007 7 122 49 854
Total 28 692 140 3063
Average 4 98.86
Least Squares Method
Answer:
σ 𝑥𝑦 − 𝑛𝑥ҧ 𝑦ത 3063 − (7)(4)(98.86)
𝑏= = = 10.54
σ 𝑥 2 − 𝑛𝑥ҧ 2 140 − (7)(4)2

𝑎 = 𝑦ത − 𝑏𝑥ҧ = 98.86 − 10.54 4 = 56.70

 Therefore, the trend line is:

𝑦ො = 56.70 + 10.54𝑥
 For 2010, demand will be 162.1
Least Squares Method
Seasonal Index

 The multiplicative seasonal model can adjust trend data for

seasonal variations in demand
 Steps:
 Find average historical demand for each season
 Compute the average demand over all seasons
 Compute a seasonal index for each season
 Estimate next year’s total demand
 Divide this estimate of total demand by the number of seasons, then
multiply by the seasonal index for that season
Seasonal Index
Sample problem:
 A Des Moines distributor of Sony laptop computers wants to develop
monthly indices for sales. Data from 2005 to 2007, by month are
available.
Seasonal Index
Answer:
 Compute for the seasonal index.

𝑎𝑣𝑒𝑟𝑎𝑔𝑒 2005 − 2007 𝑚𝑜𝑛𝑡ℎ𝑙𝑦 𝑑𝑒𝑚𝑎𝑛𝑑

𝑆𝑒𝑎𝑠𝑜𝑛𝑎𝑙 𝑖𝑛𝑑𝑒𝑥 =
𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑚𝑜𝑛𝑡ℎ𝑙𝑦 𝑑𝑒𝑚𝑎𝑛𝑑

90
𝑆𝑒𝑎𝑠𝑜𝑛𝑎𝑙 𝑖𝑛𝑑𝑒𝑥 𝐽𝑎𝑛𝑢𝑎𝑟𝑦 = = 0.957
94
Seasonal Index
Answer:
Seasonal Index
Answer:
 For 2008 forecast:
 If expected annual demand is 1200:

1200
𝐽𝑎𝑛𝑢𝑎𝑟𝑦 2008 = ∗ 0.957 = 96
12

1200
𝐹𝑒𝑏𝑟𝑢𝑎𝑟𝑦 2008 = ∗ 0.851 = 85
12
Associative Forecasting

 Used when changes in one or more independent variables can be

used to predict the changes in the dependent variable
 Most common technique is the linear regression analysis
 The same method as with least squares method is used
Associative Forecasting

𝑦ො = 𝑎 + 𝑏𝑥

σ 𝑥𝑦 − 𝑛𝑥ҧ 𝑦ത
𝑏=
σ 𝑥 2 − 𝑛𝑥ҧ 2

𝑎 = 𝑦ത − 𝑏𝑥ҧ
 Where:
 𝑦ො = computed value of the variable to be predicted
 a = y-axis intercept
 b = slope of the regression line
 x = the independent variable
Associative Forecasting
Sample problem:
 Nodel Construction Company renovates old homes in West
Bloomfield, Michigan. Over time, the company has found that its
dollar volume of renovation work is dependent on the West
Bloomfield area payroll. Management wants to establish a
mathematical relationship to help predict sales.

Sales (in million $) Local payroll (billion $)

2.0 1
3.0 3
2.5 4
2.0 2
2.0 1
3.5 7
Associative Forecasting
Answer:

Sales (y) Local payroll (x) x2 xy

2.0 1 1 2.0
3.0 3 9 9.0
2.5 4 16 10.0
2.0 2 4 4.0
2.0 1 1 2.0
3.5 7 49 24.5
Total: 15 Total: 18 Total: 80 Total: 51.5
Average: 2.5 Average: 3
Associative Forecasting
Answer:
σ 𝑥𝑦 − 𝑛𝑥ҧ 𝑦ത 51.5 − (6)(3)(2.5)
𝑏= = = 0.25
σ 𝑥 2 − 𝑛𝑥ҧ 2 80 − (6)(3)2

𝑎 = 𝑦ത − 𝑏𝑥ҧ = 2.5 − 0.25 3 = 1.75

 Therefore, the trend line is:

𝑦ො = 1.75 + 0.25𝑥
 If payroll next year is estimated to be $6 billion, then sales will be
$3,250,000
Associative Forecasting
Answer:
Standard Error of Estimate

 A forecast is just a point estimate of a future value

 This point is actually the mean of a probability distribution

σ(𝑦 − 𝑦𝑐 )2
𝑆𝑦,𝑥 =
𝑛−2
 Where:
 y = y-value of each data point
 𝑦𝑐 = computed value of the dependent variable from the regression
equation
 n = number of data points
Standard Error of Estimate

σ 𝑦 2 − 𝑎 σ 𝑦 − 𝑏 σ 𝑥𝑦
𝑆𝑦,𝑥 =
𝑛−2
 Computationally easier to use
 We use the standard error to set up prediction intervals around the
point estimate
Standard Error of Estimate

 Using the previous problem:

σ 𝑦 2 − 𝑎 σ 𝑦 − 𝑏 σ 𝑥𝑦 39.5 − (1.75) 15 − (0.25)(51.5)
𝑆𝑦,𝑥 = = = 0.306
𝑛−2 6−2
 The standard error of estimate is $306,000 in sales
Correlation

 How strong is the linear relationship between variables?

 Coefficient of correlation, r, measures degree of association (range
from -1 to +1)

𝑛 σ 𝑥𝑦 − σ 𝑥 σ 𝑦
𝑟=
[𝑛 σ 𝑥 2 − (σ 𝑥)2 ][𝑛 σ 𝑦 2 − (σ 𝑦)2 ]
 Where:
 y = y-value of each data point
 𝑦𝑐 = computed value of the dependent variable from the regression
equation
 n = number of data points
Correlation

 Coefficient of determination, r2, measures percent of change in y

predicted by change in x (values range from 0 to 1)
 From the previous example:

(6)(51.5) − (18)(15)
𝑟= = 0.9014
[(6)(80) − (18)2 ][(6)(39.5) − (15)2 ]

𝑟 2 = (0.9014)2 = 0.8125
Multiple Regression Analysis

 If more than one independent variable is to be used in the model,

linear regression can be extended to multiple regression.

𝑦ො = 𝑎 + 𝑏1 𝑥1 + 𝑏2 𝑥2 + ⋯ + 𝑏𝑛 𝑥𝑛
 Generally complex and done on computers
Monitoring and Controlling
Forecasts
 Tracking Signal
 Measures how well the forecast is predicting actual values
 Ratio of running sum of forecast errors (RSFE) to mean absolute deviation (MAD)
 Good tracking signal has low values
 If forecasts are continually high or low, the forecast has a bias error

σ(𝐴𝑐𝑡𝑢𝑎𝑙 𝑑𝑒𝑚𝑎𝑛𝑑 𝑖𝑛 𝑝𝑒𝑟𝑖𝑜𝑑 𝑖 − 𝐹𝑜𝑟𝑒𝑐𝑎𝑠𝑡 𝑑𝑒𝑚𝑎𝑛𝑑 𝑖𝑛 𝑝𝑒𝑟𝑖𝑜𝑑 𝑖)

𝑇𝑟𝑎𝑐𝑘𝑖𝑛𝑔 𝑆𝑖𝑔𝑛𝑎𝑙 =
σ 𝐴𝑐𝑡𝑢𝑎𝑙 − 𝐹𝑜𝑟𝑒𝑐𝑎𝑠𝑡
𝑛
Monitoring and Controlling
Forecasts
 How do firms decide what the upper and lower tracking limits should be?
 There is no single answer. That is, limits not so low as to be triggered with every small
forecast error and not so high as to allow bad forecasts to be regularly overlooked
Monitoring and Controlling
Forecasts
Sample problem:
 Alex’s cake shop wants to evaluate the performance of its cake forecast.
Monitoring and Controlling
Forecasts
Answer:
Monitoring and Controlling
Forecasts
Answer:
Quarter Tracking Signal
1 -10/10 = -1
2 -15/7.5 = -2
3 0/10 = 0
4 -10/10 = -1
5 5/11 = 0.5
6 35/14.2 = 2.5

 The variation of the tracking signal between -2.0 and +2.5 is within
acceptable limits
Adaptive Forecasting

 It is possible to use the computer to continually monitor forecast

error and adjust the value of and β coefficients used in
exponential smoothing to continually minimize forecast error. This is
called adaptive smoothing.
Focus Forecasting

 Developed at American Hardware Supply

 Based on two principles:
 Sophisticated forecasting models are not always better than simple
ones
 There is no single technique that should be used for all products or
services
 Uses historical data to test multiple forecasting models for individual
items
 The forecasting model with the lowest error is used to forecast the
next demand
Thank you!