1

PRODUCTION AND OPERATIONS

MANAGEMENT MGT-303

Forecasting
 The outline of the session
2

 What is Forecasting?
 Types of Forecasts
 Seven Steps in Forecasting System
 An Overview of Forecasting Techniques
 Qualitative Methods of Forecasting
 Quantitative Methods of Forecasting
 A Case Study: The Computer Club Warehouse
Problem
 Applying Time-Series Forecasting to the Case Study
 Learning objectives
3

At the end of forecasting session, you should be able to

1. Describe some important types of forecasting applications.
2. Identify two common measures of the accuracy of forecasting
methods.
3. Adjust forecasting data to consider seasonal patterns.
4. Describe several forecasting methods that use the pattern of
historical data to forecast a future value.
5. Apply these methods with the Excel software.
6. Compare these methods to identify the conditions when each is
particularly suitable.
7. Describe several forecasting methods that use expert judgment.
 What is Forecasting?
4

 Forecasting is a process of estimating the

unknown.

 Forecasting Time Horizons

• Short-range forecast
• Medium-range forecast
• Long-range forecast
 Forecasting System
5

1. Determine what the forecast is for.

2. Select the items for the forecast.

3. Select the time horizon.

4. Select the forecast model type.

5. Gather data to be input into the model.

6. Make the forecast.

7. Verify and implement the results.

 Approaches to Forecasting
6

Forecasting Methods

Quantitative Qualitative

Time-series
Models Associative Personal Assessment
Panel consensus

Model
Delphi method
Sales Force composite
Market Research

Last-value Method
Moving Averages
Exponential
Linear
Smoothing
Trend Projection
Regression
 Forecasting Methods Currently in Use
7

 Qualitative Methods
• Personal Assessment
• Panel Consensus/Jury of executive opinion
• Delphi Method
• Sales Force Composite
• Market Research/Market survey
 Forecasting Methods Currently in Use
8

 Quantitative Methods
- There are two subcategories:

-Time Series Methods: seek to identify historical patterns

(using time as reference) and the forecast using a time-based
extrapolation of those patterns.

- Explanatory (Causal) Methods: seek to identify the

relationships that led to (caused) observed outcomes in the past
and then forecast by applying those relationships to the future.
 An overview of Quantitative Forecasting
Techniques
9

Time Series Forecasting Methods

 Example: Forecasting at
10
Fastchips
 Fastchips is a leading producer of microprocessors.
 Six months ago, it launched the sales of its latest
microprocessor.
 Month-by-month sales (in thousands) over the initial
six months have been

17 25 24 26 30 28

Question: What is the forecast for next month’s sales?

 The Last-Value Forecasting Method
11

The last-value forecasting method ignores all

data points in a time series except the last one.

Forecast = Last value

Fastchips: Month-by-month sales (in thousands)

over the initial six months:

17 25 24 26 30 28
Forecast = 28
 The Averaging Forecasting Method
12

The averaging forecasting method uses all the

data points in the time series and simply
averages these points.
Forecast = Average of all data to date

Fastchips: Month-by-month sales (in

thousands) over the initial six months:
17 25 24 26 30 28
Forecast = (17+25+24+26+30+28) / 6 = 25
 The Moving-Average Forecasting Method
13

The moving-average forecasting method averages the data for

only the most recent time periods.

n = Number of recent periods to consider as relevant for

forecasting
Forecast = Average of last n values

Fastchips: Month-by-month sales (in thousands) over the initial

six months:
17 25 24 26 30 28

Forecast (n=3) = (26+30+28) / 3 = 28

 The Exponential Smoothing Forecasting
Method
14

 The exponential smoothing forecasting method

provides a more sophisticated version of the
moving-average method.
 It gives the greatest weight to the last month and
then progressively smaller weights to the older
months.
 Exponential smoothing with trend adjusts
exponential smoothing by also directly considering
any current upward or downward trend in sales
 Measuring the Forecast (error) Accuracy
15

 The mean absolute deviation (called MAD) measures the

average forecasting error.

MAD = (Sum of forecasting errors) / (Number of forecasts)

 The mean square error (often abbreviated MSE)

measures the average of the square of the forecasting error.

MSE = (Sum of square of forecasting errors) / (Number of forecasts).

 The MSE increases the weight of large errors relative to

the weight of small errors.
 A Case Study: The Computer Club Warehouse
(CCW)
16
 The Computer Club Warehouse (CCW) sells computer products at
bargain prices by taking telephone orders (as well as website and
fax orders) directly from customers.
 Products include computers, peripherals, supplies, software, and
computer furniture.
 The CCW call center is never closed. It is staffed by dozens of
agents to take and process customer orders.
 A large number of telephone trunks are provided for incoming
calls. If an agent is not free when a call arrives, it is placed on
hold. If all the trunks are in use (called saturation), the call
receives a busy signal.
 An accurate forecast of the demand for agents is needed.

Question: How should the demand for agents be forecasted?

 Average Daily Call Volume (3 Years of Data)
17

A B C D E
1 CCW's Average Daily Call Volume
2
3 Year Quarter Call Volume
4 1 1 6,809
5 1 2 6,465
6 1 3 6,569
7 1 4 8,266
8 2 1 7,257
9 2 2 7,064
10 2 3 7,784
11 2 4 8,724
12 3 1 6,992
13 3 2 6,822
14 3 3 7,949
15 3 4 9,650
 Applying Time based (time-series) Forecasting
Models to the Case Study
18

Data collected over a period of time is usually called a time-series.

Components of a Time Series:

- Seasonal Component: these are regular fluctuations within a
complete time period (a day, a week, a month etc.).
- Trend Component: is the underlying movement in the data – it
may be upward, downward or stationary.
- Cyclical Component: are long term fluctuations in the data and
are similar to seasonal factors. They can be difficult to identify unless
long series of data is available.
- Irregular Component: are random or residual factors, like
unpredictable element to the data. Something like unusual weather
conditions affecting holiday sales.
 Considering Seasonal Effects
19
 When there are seasonal patterns in the data, they can be
addressed by forecasting methods that use seasonal factors.

 The seasonal factor for any period of a year (a quarter, a month,

etc.) measures how that period compares to the overall average
for an entire year.
Seasonal factor = (Average for the period) / (Overall average)

 It is easier to analyze data and detect new trends if the data are
first adjusted to remove the seasonal patterns.
Seasonally adjusted data = (Actual call volume) / (Seasonal
factor)
 Calculation of Seasonal Factors for CCW
20
 Excel Template for Calculating Seasonal
Factors
21

A B C D E F G
1 Estimating Seasonal Factors for CCW
2
3 True
4 Year Quarter Value Type of Seasonality
5 1 1 6,809 Quarterly
6 1 2 6,465
7 1 3 6,569
8 1 4 8,266 Estimate for
9 2 1 7,257 Quarter Seasonal Factor
10 2 2 7,064 1 0.9323
11 2 3 7,784 2 0.9010
12 2 4 8,724 3 0.9873
13 3 1 6,992 4 1.1794
14 3 2 6,822
15 3 3 7,949
16 3 4 9,650
 Seasonally Adjusted Time Series for CCW
22

A B C D E F
1 Seasonally Adjusted Time Series for CCW
2
3 Seasonal Actual Seasonally Adjusted
4 Year Quarter Factor Call Volume Call Volume
5 1 1 0.93 6,809 7,322
6 1 2 0.90 6,465 7,183
7 1 3 0.99 6,569 6,635
8 1 4 1.18 8,266 7,005
9 2 1 0.93 7,257 7,803
10 2 2 0.90 7,064 7,849
11 2 3 0.99 7,784 7,863
12 2 4 1.18 8,724 7,393
13 3 1 0.93 6,992 7,518
14 3 2 0.90 6,822 7,580
15 3 3 0.99 7,949 8,029
16 3 4 1.18 9,650 8,178
 Outline for Forecasting Call Volume
23

1. Select a time-series forecasting method.

2. Apply this method to the seasonally adjusted time

series to obtain a forecast of the seasonally
adjusted call volume for the next time period.

3. Multiply this forecast by the corresponding

seasonal factor to obtain a forecast of the actual
call volume (without seasonal adjustment).
 The Last-Value Forecasting Method
24

 The last-value forecasting method ignores all

data points in a time series except the last one.

Forecast = Last value

 The last-value forecasting method is sometimes
called the naïve method, because statisticians
consider it naïve to use just a sample size of one
when other data are available.
 However, when conditions are changing rapidly, it
may be that the last value is the only relevant data
point.
 Measuring the Forecast Accuracy
(error) for Last-Value Method with
25 Seasonality
 The mean absolute deviation (MAD)

MAD = (Sum of forecasting errors) / (Number of

forecasts)
= 3246/11 = 295

 The mean square error (MSE)

MSE = (Sum of square of forecasting errors) / (Number of

forecasts).
= (124)2+(543)2+ … +(175)2/11
= 145909
 The Averaging Forecasting Method
26

 The averaging forecasting method uses all the data points

in the time series and simply averages these points.

Forecast = Average of all data to date

 The averaging forecasting method is a good one to use when

conditions are very stable.

 However, the averaging method is very slow to respond to

changing conditions. It places the same weight on all the
data, even though the older values may be less representative
of current conditions than the last value observed.
 The Moving-Average Forecasting Method
27
 The moving-average forecasting method averages the data for only the
most recent time periods.
n = Number of recent periods to consider as relevant for
forecasting
Forecast = Average of last n values

 The moving-average forecasting method is a good one to use when

conditions don’t change much over the number of time periods included in
the average.

 However, the moving-average method is slow to respond to changing

conditions. It places the same weight on each of the last n values even
though the older values may be less representative of current conditions
than the last value observed.
 The Moving-Average Method Applied to CCW
28
A B C D E F G H I J K
1 Moving Average Forecasting Method with Seasonality for CCW
2
3 Seasonally Seasonally
4 True Adjusted Adjusted Actual Forecasting Number of previous
5 Year Quarter Value Value Forecast Forecast Error periods to consider
6 1 1 6,809 7,322 n= 4
7 1 2 6,465 7,183
8 1 3 6,569 6,635 Type of Seasonality
9 1 4 8,266 7,005 Quarterly
10 2 1 7,257 7,803 7,036 6,544 713
11 2 2 7,064 7,849 7,157 6,441 623 Quarter Seasonal Factor
12 2 3 7,784 7,863 7,323 7,250 534 1 0.93
13 2 4 8,724 7,393 7,630 9,003 279 2 0.90
14 3 1 6,992 7,518 7,727 7,186 194 3 0.99
15 3 2 6,822 7,580 7,656 6,890 68 4 1.18
16 3 3 7,949 8,029 7,589 7,513 436
17 3 4 9,650 8,178 7,630 9,004 646
18 4 1 7,826 7,279
19 4 2
20 4 3
21 4 4
22 5 1
23 5 2
24 5 3
25 5 4 Mean Absolute Deviation
26 6 1 MAD = 437
27 6 2
28 6 3 Mean Square Error
29 6 4 MSE = 238,816