POST GRADUATE DIPLOMA IN SUPPLY CHAIN MANAGEMENT

Manufacturing Planning & Control

(Forecasting & Demand Management)
Lecture - 3
 Topics to be Discussed

Processing Demand

Influencing Demand

How to Improve Forecast Accuracy

Long Term Forecasting

Short Term Forecasting

Summary
Scope of Demand Management
• Demand Management covers how a firm integrates
information from and about its customers, internal and
external to the firm, into the manufacturing planning and
control systems.
• It deals about how a firm integrates information from its
customers with information about the firms goals and
capabilities, to determine what should be produced in the
future.
• Demand Management is concerned with processing,
influencing, and anticipating demand
Demand Management, Forecast &
Plans
• In DM, FORECASTS of the quantities and timing of
customer demand are developed.

What do we actually plan to deliver to customers each

period is the output of the process. This is based on
marketing quotas, special sales incentives, etc. These
amounts will be based on inputs from many different
sources and not just quantitative forecasts.
Why Forecast and Plans are important

• A manufacturing manager cannot be held

responsible for not getting a forecast right,

• A manufacturing manager can and should be

held responsible for making their plans.
Responsibility of the MPC
• Providing the means for making as good a set of
executable plans as possible and then
Providing the information to execute them.

and when conditions change

• The control function should change the plans and
• The new plans should be executed faithfully.
Forecasting
• Predict the next number in the pattern:

a) 3.7, 3.7, 3.7, 3.7, 3.7, ?

b) 2.5, 4.5, 6.5, 8.5, 10.5, ?

c) 5.0, 7.5, 6.0, 4.5, 7.0, 9.5, 8.0, 6.5, ?

Forecasting

• Predict the next number in the pattern:

a) 3.7, 3.7, 3.7, 3.7, 3.7, 3.7

b) 2.5, 4.5, 6.5, 8.5, 10.5, 12.5

9.0
c) 5.0, 7.5, 6.0, 4.5, 7.0, 9.5, 8.0, 6.5,
Definition of the Forecasting Process

• The Art and Science of Predicting Future Events

• Forecasting vs. Predicting
• Based on Past Data
• Economic vs. Demand Forecasting
• Process of predicting a future event based on historical data
• Underlying basis of all business decisions
• Production
• Inventory
• Personnel
• Facilities
• Forecasting is the process of projecting the values of one or
more variables into the future.
 Why do we need to forecast?
In general, forecasts are almost always wrong. So,

Throughout the day we forecast very different

things such as weather, traffic, stock market, state
of our company from different perspectives.

Virtually every business attempt is based on

forecasting. Not all of them are derived from
sophisticated methods. However, “Best" educated
guesses about future are more valuable for
purpose of Planning than no forecasts and hence
no planning.
Importance of Forecasting in OM
Departments throughout the organization depend on
forecasts to formulate and execute their plans.

Finance needs forecasts to project cash flows and

capital requirements.

Human resources need forecasts to anticipate hiring

needs.

Production needs forecasts to plan production

levels, workforce, material requirements,
inventories, etc.
Importance of Forecasting in OM
Demand is not the only variable of interest to
forecasters.

Manufacturers also forecast worker

absenteeism, machine availability, material
costs, transportation and production lead
times, etc.

Besides demand, service providers are also

interested in forecasts of population, of other
demographic variables, of weather, etc.
Elements of Demand Forecasting

• Dynamic in Nature
• Consider Uncertainty
• Rely on Information contained in Past Data
• Applied to various time horizons
• short term
• medium term forecasts
• long term forecasts
 Steps in the Forecasting Process

• Determine the Use of the Forecast

• Select the Items to be Forecasted
• Determine a Suitable Time Horizon
• Select an appropriate Set of Forecasting Models
• Gather Relevant Data
• Conduct the Analysis
• Validate the Model - Assess its Accuracy
• Make the Forecast
• Implement the Results
 Independent Demand:
What a firm can do to manage it?

• Can take an active role to influence demand

• Can take a passive role and simply respond to

demand
Factors Influencing Demand
• EXTERNAL FACTORS over which management has little or no
control:
• Economic conditions
• Government regulation
• What the competition does
• Consumer behavior
• The Leading, Coincident and Lagging Indicators provide
forecasters with data on the external factors.
• Market Research also does this.

• INTERNAL FACTORS that management can control:

• Price
• Promotion
• Product
• Quality
• Reputation
OPERATIONS FORECASTING

• OPERATIONS MANAGERS are primarily concerned

with forecasting demand…
…for capacity planning.
(Effective capacity, capacity cushion, etc.)
…for production scheduling.
(Lot sizes, aggregate planning, etc.)
…to plan inventory needs.
(Order quantities, safety stock, etc.)
…to better match supply and demand.

PGDSCM
Manufacturing, Planning & Control
Operations Forecasting

Most operations forecasting is time-series. (Historic

data is correlated with time.)
• Independent variable is time.
• Time is independent of the data being forecast.
• Dependent variable is demand.
• EG: Using the past years of monthly demand data for your
product or service to make a forecast for next month’s
demand.
 Forecasting and Demand Planning
• Poor forecasting can result in poor inventory and staffing decisions, resulting in
part shortages, inadequate customer service, and many customer complaints.
• Many firms integrate forecasting with value chain and capacity management
systems to make better operational decisions.
• Accurate forecasts are needed throughout the value chain, and are used by all
functional areas of the organization, including accounting, finance, marketing,
operations, and distribution.
• One of the biggest problems with forecasting systems is that they are driven by
different departmental needs and incentive systems.
• Demand planning software systems integrate marketing, inventory, sales,
operations planning, and financial data.
Basic Approach to Demand Forecasting

• Understand the objectives of forecasting

• Integrate demand planning and forecasting
• Identify major factors that influence the demand
forecast
• Understand and identify customer segments
• Determine the appropriate forecasting technique
• Establish performance and error measures for the
forecast
Role of Forecasting in a Supply Chain

• The basis for all strategic and planning decisions in a supply

chain
• Used for both push and pull processes
• Examples:
• Production: scheduling, inventory, aggregate planning
• Marketing: sales force allocation, promotions, new
production introduction
• Finance: plant/equipment investment, budgetary
planning
• Personnel: workforce planning, hiring, layoffs
• All of these decisions are interrelated
Characteristics of Forecasts

• All forecasts are wrong (rarely correct). The

best we can hope for is to reduce the amount
of error. Forecasts should include expected
value and measure of error.
• Long-term forecasts are less accurate than
short-term forecasts (forecast horizon is
important)
• Aggregate forecasts are more accurate than
disaggregate forecasts
Benefits of Forecasts

• Forecasts reduce the uncertainty in our

decision making.
• Forecasts aid us in planning.
• They allow us to plan for contingencies.
 Forecasting at Unilever

• Unilever is an Anglo-Dutch multinational corporation that

owns many of the world's consumer product brands in
foods, beverages, cleaning agents and personal care
products.
• Their demand forecasting blends historical shipment
data with promotional data and current order data.
• Statistical forecasts are adjusted based on planned
product promotions.
• Forecasts are frequently reviewed and adjusted with the
most recent point-of-sale data.
• This has enabled Unilever to reduce its inventory, and
improved its scheduling and customer service.
Types of Forecasts by Time Horizon
Quantitative
• Short-range forecast methods
• Usually < 3 months
• Job scheduling, worker assignments

• Medium-range forecast
Detailed
• 3 months to 2 years use of
• Sales/production planning system

• Long-range forecast
• > 2 years
• New product planning

Design Qualitative
of system Methods
Characteristics of Long Term Forecasts

• Single or multi-year horizon

• Monthly or annual time bucket

• Aggregate units
• Input to “long term” decisions

• Accuracy generally more important than short term

forecasts
• Tend to use expensive & time consuming methods
 WHICH METHOD FOR LONG-TERM
DEMAND FORECASTING?
• Long Term (Beyond 2 years)
• Causal & Qualitative (judgment) Models are typically used.
Purposes include…
• Location decisions
• Capacity decisions
• Layout and Process decisions
• Most forecasting for strategic decisions is long-term
forecasting.
 WHICH METHOD TO USE FOR
MEDIUM-TERM DEMAND FORECASTING?

Medium Term (3 months–2 years)

• Purpose:
• capacity planning
• Causal Models are the best to use.
• Regression is common.
• Qualitative (Judgment) models are also helpful.
• Executive opinion, Market Research, Sales force estimates
 WHICH METHOD TO USE FOR
SHORT-TERM DEMAND FORECASTING?
Short Term: (Up to three months)
• Purpose:
• Production scheduling
• Inventory planning
• Method:
• Time Series Forecasting is the most commonly used
forecasting technique.
• It is inexpensive and easy to do.
• Some judgment models can be used.
• Sales-force estimates, executive opinion
• Again, good judgment is ALWAYS important.
 Demand Forecasting Summary
Time Horizon

Short Term Medium Term Long Term

Application (0–3 months) (3 months–2 yrs) (over 2 years)
• Individual products • Total sales • Total sales
Forecast or services • Groups of products or
Focus services

• Inventory Mgt. • Staff planning • Facility location

• Final assembly • Production planning • Capacity planning
Decision scheduling • Aggregate Prod. • Process
• Workforce scheduling management
Area scheduling • Purchasing
• Master Prod. • Distribution
scheduling

Forecasting • Time series • Causal • Causal

Technique • Causal • Judgment • Judgment
• Judgment
 Long term/short term characteristics
Long term forecasts Short term forecasts

• Single or multi-year horizon • Weekly or monthly horizon

• Monthly or annual time bucket • Daily & weekly time bucket

• Aggregate units (e.g., product/ service • Detailed units (e.g., SKU)

categories)
• Input to “short term” decisions
• Input to “long term” decisions
• Inexpensive & quick methods
• Expensive & time consuming • Accuracy importance
methods • Trumpet of doom
• Accuracy importance
• Trumpet of doom
Forecasting During the Life Cycle
Introduction Growth Maturity Decline

Qualitative models Quantitative models

- Executive judgment
- Time series analysis
- Market research
- Regression analysis
-Survey of sales force
-Delphi method
Sales

Time
 Types of Forecasts
• Qualitative (Judgmental)

• Quantitative
• Time Series Analysis
• Causal Relationships
• Simulation
 JUDGMENT METHODS (QUALITATIVE)
• Judgment methods rely on the opinions of experts, or on
the judgment and experience of people in the best
position to know.
• Much of market research is qualitative.
• Surveys of customer preferences and intentions to buy

• Judgment methods are best for medium or long-term

forecasting.
• Some qualitative methods can take considerable time
to obtain, and they can be expensive.
 JUDGMENT METHODS
• Sales force estimates: Forecasts are made by a company’s sales-force
members who have first-hand interaction with customers.
• Executive opinion (Executive intuition): The opinions, experience, and
technical knowledge of experienced managers are summarized to arrive
at a single forecast.
• Market research: A systematic approach to determining consumer
interest in a service or product through data-gathering surveys.
(Quantitative methods are often applied to this data.)
• Delphi method: A process of gaining consensus from a group of experts,
usually external to the organization, while maintaining individual
anonymity. (Survey-feedback-survey method)
• Experts are drawn from across the industry, government,
public and private organizations.
 Using Judgment Forecasting
• Judgment forecasting is clearly needed when
numerical data are not available for quantitative
forecasting approaches.
• It should also be used, even when quantitative data is
available, as an additional forecasting tool.
• Good judgment quite often is better than all the statistics in
the world.

• Guidelines for the use of any type of forecasting:

• Adjust forecasts when you have access to important
contextual knowledge.
• Make adjustments to compensate for specific events, such as
advertising campaigns, the actions of competitors, changes in the
economic situation, and international developments.
 Qualitative Methods

Executive Judgment Grass Roots

Historical analogy Qualitative Market Research

Methods

Delphi Method Panel Consensus

 Delphi Method
l. Choose the experts to participate representing a variety of
knowledgeable people in different areas
2. Through a questionnaire (or E-mail), obtain forecasts (and any
premises or qualifications for the forecasts) from all participants
3. Summarize the results and redistribute them to the
participants along with appropriate new questions
4. Summarize again, refining forecasts and conditions, and again
develop new questions
5. Repeat Step 4 as necessary and distribute the final results to all
participants
 Quantitative Forecasting Models

• Both Pattern Based and Correlational Models rest on the assumption

that the relationships of the past will continue into the Future
• Both can Mathematically Characterize the Probabilistic Nature of the
Forecast
• Both Use Information from Relevant Time Frames
 Components of Demand

• Average demand for a period of time

• Trend
• Seasonal element
• Cyclical elements
• Random variation
• Autocorrelation
Pattern Based Analyses
• Definition
• Identifying an underlying pattern in historical data, describe it in
mathematical terms, and then extrapolate it into the future
• Uses a “Time Series” of Past Data
 Time Series Variation

• Time Series of Demand Data Typically Contain Four

Components of Variation About the Mean or
Average
• Pattern Based Forecasting Needs to
Mathematically Characterize Each of these
 Finding Components of Demand
Seasonal variation

x Linear
x x
x x
x x Trend
Sales

x x
x x x
x
x
xx
x xx x x
x
x
x x x x x x
x x x x x x
x x x
x xxxxx
x
x x

1 2 3 4

Year
 Time Series Methods
• TIME-SERIES is a commonly used statistical
approach that relies on historical data for short-
term forecasting.
• Types of Time Series forecasting models:
(Listed in order of increasing complexity)
• NAIVE FORECASTING
• MOVING AVERAGES
• TREND PROJECTIONS (Good for long-term)
• EXPONENTIAL SMOOTHING
• BOX JENKINS
 CAUSAL METHODS
• Causal methods are used when historical data are available and a
relationship between the variable and other external or internal
factors can be identified.
• Causal methods use historical data on independent variables, such as
promotional campaigns, economic conditions, and competitors’ actions, to
predict demand (dependent variable)
• May be short, medium, or long term, depending on the model.
• LINEAR & NON-LINEAR REGRESSION (Short & Medium term)
• ECONOMETRICS (Good for long term)
• INTENTION-TO-BUY & ANTICIPATION SURVEYS (Short-term)
• INPUT-OUTPUT MODELS (Good for long-term)
• LEADING INDICATORS (Only for long-term)
• These are indicators that precede economic change.
(unemployment, inventory changes, building permits, money
supply, etc.)
 Linear Regression
• Linear Regression is a causal method in which one variable
(dependent variable) is related to one or more independent
variables using a linear equation.
• Dependent variable: The variable to be forecasted.
• In demand forecasting, demand would be the dependent variable
• Data plot must be linear in order to use Linear Regression
• Independent variables are assumed to have a correlation with the
dependent variable being forecast.
• The Independent Variable is some variable to which demand appears to
be related. It can be time or some other variable.
• If the Independent variable is time, then linear regression becomes a
Time-Series method of forecasting.
• NO “Cause and Effect” should be assumed, even though it is called a
Causal Method!
 Time Series Analysis

• Time series forecasting models try to predict the future based on

past data
• One can pick models based on:
1. Time horizon to forecast
2. Data availability
3. Accuracy required
4. Size of forecasting budget
5. Availability of qualified personnel
 Basic Time-Series Patterns
There are five basic patterns of most time series.
a. Horizontal: Over time the data fluctuates around a constant
mean.
b. Trend: The systematic increase (or decrease) in the data over
time.
c. Seasonal: A repeatable pattern of increases and decreases in
demand that relates to a specific period, such as the time of day,
week, month, or season.
d. Cyclical: Less predictable, gradual increases and decreases over
longer periods of time (years or decades), such as the business
cycle. No consistent time frame.
e. Random: A variation in demand that has no pattern.
The data cannot be used for forecasting.
Patterns of Demand
Horizontal Trend

Seasonal Cyclical
 Rules for Time-Series Forecasting

• RULE 1: Plot your data to see if it has a pattern.

If the data has a pattern, you

can select an appropriate
forecasting model. Demand

Sales history by month.

If the data has no pattern,

you cannot do forecasting!
Demand

Sales history by month.

Rules for Time-Series Forecasting
• RULE 2: The number of periods of data depends
on how much confidence you want in the results
and the technique being used.
• The number of periods you need varies with the
forecasting technique. Some methods require less data;
some require lots of data.
• RULE 3: Time-Series Forecasting is SHORT-TERM
forecasting.
• Generally, you don’t forecast beyond the first unknown
period unless your historic data has a clear pattern.
(minimal variation/noise)
• Time-Series forecasting beyond the first unknown period
greatly increases forecast error and unreliability.
 TIME-SERIES METHODS

• NAÏVE FORECASTING: A time-series method whereby the forecast for

the next period is the known demand for the current period.
• It takes the most recent known period value and projects it to the
first forecast period.

• SIMPLE MOVING AVERAGES is a time-series method that averages

demand over a specified period “n” of time.
• It computes the average for the last “n” periods and uses that as
the forecast for the next period.
• By averaging, it removes the effects of random fluctuations, and it
is most useful when demand has no pronounced trend or
seasonal influences.
 Simple Moving Averages
The moving average method involves the use of as many periods of past
demand as desired or deemed appropriate. The stability of the demand
series generally determines how many periods.

This is a 4-period moving average.

WEEK DEMAND AVERAGE

1 20
2 23
3 21
4 24 22
5 25 23.25
6 ?

22 becomes the forecast for week #5 23.25 becomes the

forecast for week #6.
 Simple Moving Average Formula
• The simple moving average model assumes an average is
a good estimator of future behavior
• The formula for the simple moving average is:

A t-1 + A t-2 + A t-3 +...+A t- n

Ft =
n
Ft = Forecast for the coming period
N = Number of periods to be averaged
A t-1 = Actual occurrence in the past period for up to
“n” periods
Comparison of 3- and 6-Week Moving Average
Forecasts
3-week moving average 6-week moving average
forecast forecast

Patient Arrivals

Actual Data Historic Data

Week
A longer averaging period soothes the fluctuations.
Simple Moving Average Problem (1)

A t-1 + A t-2 + A t-3 +...+A t- n

Week Demand Ft =
1 650 n
2 678
3 720 Question: What are the 3-
4 785 week and 6-week moving
5 859 average forecasts for
6 920 demand?
7 850
Assume you only have 3
8 758
9 892
weeks and 6 weeks of actual
10 920
demand data for the
11 789
respective forecasts
12 844
Calculating the moving averages gives us:

Week Demand 3-Week 6-Week

1 650 F4=(650+678+720)/3
2 678 =682.67
3 720 F7=(650+678+720
+785+859+920)/6
4 785 682.67
=768.67
5 859 727.67
6 920 788.00
7 850 854.67 768.67
8 758 876.33 802.00
9 892 842.67 815.33
10 920 833.33 844.00
11 789 856.67 866.50
12 844 867.00 854.83
©The McGraw-Hill Companies, Inc., 2004
 Plotting the moving averages and comparing them shows how
the lines smooth out to reveal the overall upward trend in this
example

Demand 3-Week 6-Week

950
900
850
800
Dem and

750
700 Note how the 3-
650 Week is smoother
600 than the Demand,
550
500 and 6-Week is even
1 2 3 4 5 6 7 8 9 10 11 12 smoother
Week
 Simple Moving Average Problem (2) Data

Question: What is the 3

week moving average
Week Demand
forecast for this data?
1 820 Assume you only have 3
2 775 weeks and 5 weeks of
3 680 actual demand data for
4 655 the respective
5 620 forecasts
6 600
7 575
Simple Moving Average Problem (2) Solution

Week Demand 3-Week 5-Week

1 820 F4=(820+775+680)/3
=758.33
2 775
F6=(820+775+680
3 680 +655+620)/5
=710.00
4 655 758.33
5 620 703.33
6 600 651.67 710.00
7 575 625.00 666.00
 Double Moving Averages
(Averaging the averages)

n= 4 periods for the average

WEEK DEMAND SINGLE DOUBLE

AVG AVG
1 20
2 25
3 34
4 19 25
5 22 25
6 12 22
7 36 22 23
8 14 21 23
9 19 20 21
10 24 23 22
11 22 20 21
12 18 21 21
13 27 23 22

Averages are rounded to the nearest whole numbers.

WEIGHTED MOVING AVERAGES
Weighted moving average method: A time-series method in which each
historical data point can have its own weight. (The sum of the weights equals 1.0)
It allows the forecaster to give more weight to the more recent data or the
more relevant data.
Important in trend or cyclical data

WEEK DEMAND Wt. Wt. AVERAGE

1 20 .1
2 23 .2
3 21 .3
4 24 .4 22.5

(20*0.1)+(23*0.2)+(21*0.3)+(24*0.4)=22.5
 Weighted Moving Average Formula

While the moving average formula implies an equal

weight being placed on each value that is being averaged,
the weighted moving average permits an unequal
weighting on prior time periods

The formula for the moving average is:

Ft = w 1 A t -1 + w 2 A t - 2 + w 3 A t -3 + ...+ w n A t - n
n
wt = weight given to time period “t”
occurrence (weights must add to one)
w
i=1
i =1
Weighted Moving Average Problem (1) Data
Question: Given the weekly demand and weights, what is the forecast for the 4th
period or Week 4?

Week Demand Weights:

1 650
2 678 t-1 .5
3 720 t-2 .3
4 t-3 .2

Note that the weights place more emphasis on the most recent data,
that is time period “t-1”
Weighted Moving Average Problem (1) Solution

Week Demand Forecast

1 650
2 678
3 720
4 693.4

F4 = 0.5(720)+0.3(678)+0.2(650)=693.4
Weighted Moving Average Problem (2) Data

Question: Given the weekly demand information and

weights, what is the weighted moving average forecast of the
5th period or week?

Week Demand Weights:

1 820 t-1 .7
2 775
t-2 .2
3 680
t-3 .1
4 655
Weighted Moving Average Problem (2) Solution

Week Demand Forecast

1 820
2 775
3 680
4 655
5 672

F5 = (0.1)(755)+(0.2)(680)+(0.7)(655)= 672
Some pros/cons
1. Simple (+)

2. Designated weights of history (-)

3. History cut-off beyond m periods (-)

Other Time Series Methods
• EXPONENTIAL SMOOTHING is a complex form of weighted moving
averages. It is good for trends and cyclical data.
• It is the most frequently used formal forecasting method because of its
simplicity and the small amount of data needed to support it.
• Double and Triple Exponential Smoothing are used for highly fluctuating
data.
• BOX JENKINS
• This is probably the most complex but often the most accurate of the
time-series methods for all data patterns.
• If you really want to know: It is named after the statisticians George Box and Gwilym
Jenkins. It applies autoregressive integrated moving average (ARIMA) models to find
the best fit of a time series to past values of this time series, in order to make
forecasts.
 Exponential Smoothing Model

Ft = Ft-1 + a(At-1 - Ft-1)

Where :
Ft  Forcast va lue for the coming t time period
Ft - 1  Forecast v alue in 1 past time period
At - 1  Actual occurance in the past t tim e period
a  Alpha smoothing constant
• Premise: The most recent observations might have the
highest predictive value
• Therefore, we should give more weight to the more
recent time periods when forecasting
Exponential Smoothing Problem (1) Data
Week Demand Question: Given the weekly
demand data, what are the
1 820 exponential smoothing forecasts
2 775 for periods 2-10 using a=0.10
3 680 and a=0.60?
4 655 Assume F1=D1
5 750
6 802
7 798
8 689
9 775
10
Answer: The respective alphas columns denote the forecast values. Note that
you can only forecast one time period into the future.

Week Demand 0.1 0.6

1 820 820.00 820.00
2 775 820.00 820.00
3 680 815.50 820.00
4 655 801.95 817.30
5 750 787.26 808.09
6 802 783.53 795.59
7 798 785.38 788.35
8 689 786.64 786.57
9 775 776.88 786.61
10 776.69 780.77
Exponential Smoothing Problem (1) Plotting
Note how that the smaller alpha results in a smoother line in this example

850
800
Demand
De m and
750
700 0.1
650
600 0.6
550
500
1 2 3 4 5 6 7 8 9 10
Week
Exponential Smoothing Problem (2) Data

Question: What are the

Week Demand exponential smoothing
1 820 forecasts for periods 2-5 using
2 775 a =0.5?
3 680
4 655 Assume F1=D1
5
 Exponential Smoothing Problem (2) Solution

F1=820+(0.5)(820-820)=820 F3=820+(0.5)(775-820)=797.75

Week Demand 0.5

1 820 820.00
2 775 820.00
3 680 797.50
4 655 738.75
5 696.88
Seasonal Patterns
• An easy way to account for seasonal effects is to use one
of the techniques already described, but to limit the
data used to those time periods in the same season.
• EG: May, June, July, August for the past five years.

• If the weighted moving-average method is used, higher

weights are placed on the more recent periods
belonging to the same season.

• Multiplicative seasonal method is a method whereby

seasonal factors are multiplied by an estimate of average
demand to arrive at a seasonal forecast.
 Seasonal Adjustments
• Applied to Moving Averages and Time Series
Regression
• First, Calculate a Seasonal Index (SI) Factor for Each
Relevant Time Period (day, week, month, quarter)
• Each Seasonal Period’s SI is Calculated by
Averaging the Ratio of its Actual Demand to the
Forecast Demand for all Corresponding Periods
Seasonal Adjustments

• Forecast for Future Periods is Calculated by

Multiplying the Unadjusted Moving Average or Time
Series Forecast for a given Period by the
Corresponding Seasonal Index for that Period
• i.e. if the SMA forecast for the month of March is 27
and the SI for March is 1.125, then
• Emar = 27*1.125 = 30.375
Seasonal Adjustment Example
Seasonal Adjustments

Sales Demand

Monthly Overall SI Adjusted

Month 1993 1994 Seasonal Index
Average Average Forecast

Jan 80 100 90.00 94.00 0.96 86.17

Feb 75 85 80.00 94.00 0.85 68.09
Mar 80 90 85.00 94.00 0.90 76.86
Apr 90 110 100.00 94.00 1.06 106.38
May 115 131 123.00 94.00 1.31 160.95
Jun 110 120 115.00 94.00 1.22 140.69
Jul 100 110 105.00 94.00 1.12 117.29
Aug 90 110 100.00 94.00 1.06 106.38
Sep 85 95 90.00 94.00 0.96 86.17
Oct 75 85 80.00 94.00 0.85 68.09
Nov 75 85 80.00 94.00 0.85 68.09
Dec 80 80 80.00 94.00 0.85 68.09

Average 87.92 100.08

Expected Demand for 1995 = 1153.23

Seasonal Adjustments Example Graph
Seasonal Adjusted Forecasting 1993

1994
170
SI Adjusted
Forecast
150
Overall
Average
130

110

90

70

50
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
 Evaluating Forecast Accuracy

• Use of Residuals Analyses

• Residuals are the Difference Between the Forecast and the Actual Demand for
a Given Period
• Assessed by Several Measures
• Mean Absolute Deviation - MAD
• Mean Squared Error - MSE
• Tracking Signal
 The MAD Statistic to Determine Forecasting Error

1 MAD  0.8 standard deviation

n

A
t=1
t - Ft
1 standard deviation  1.25 MAD
MAD =
n

• The ideal MAD is zero which would mean

there is no forecasting error

• The larger the MAD, the less the accurate

the resulting model
 MAD Problem Data

Question: What is the MAD value given the

forecast values in the table below?

Month Sales Forecast

1 220 n/a
2 250 255
3 210 205
4 300 320
5 325 315
 MAD Problem Solution

Month Sales Forecast Abs Error

1 220 n/a
2 250 255 5
3 210 205 5
4 300 320 20
5 325 315 10

40
n
Note that by itself, the MAD
A
t=1
t - Ft
40 only lets us know the mean
MAD = = = 10 error in a set of forecasts
n 4
Evaluating Forecast Accuracy
Mean Absolute Deviation - MAD
• Exponentially Smoothed MAD
• MADt = aMAD|Dt - Forecastt| + (1- aMAD)MADt-1
Evaluating Forecast Accuracy
Mean Squared Error - MSE
• MSE = ((Di - Forecasti)2)/n
Time Time
Actual Squared
Period Series Series
Demand Error
Forecast Residual
1 12 12.16 -0.16 0.03
2 13 12.13 0.87 0.76
3 10 12.09 -2.09 4.39
4 11 12.06 -1.06 1.13
5 10 12.03 -2.03 4.12
6 14 12.00 2.00 4.01
7 16 11.97 4.03 16.28
8 15 11.93 3.07 9.40
9 13 11.90 1.10 1.21
10 8 11.87 -3.87 14.97
11 10 11.84 -1.84 3.37
12 12 11.80 0.20 0.04
13 9 11.77 -2.77 7.69
14 13 11.74 1.26 1.59
15 13 11.71 1.29 1.67

MSE = 4.71
RMSE = 2.17
 Tracking Signal Formula

• The Tracking Signal or TS is a measure that indicates

whether the forecast average is keeping pace with any
genuine upward or downward changes in demand.
• Depending on the number of MAD’s selected, the TS
can be used like a quality control chart indicating
when the model is generating too much error in its
forecasts.
• The TS formula is:

RSFE Running sum of forecast errors

TS = =
MAD Mean absolute deviation
Evaluating Forecast Accuracy
Tracking Signal
• Tracking Signal = Running Sum of Forecast Error /
MAD = RSFE/MAD
Time Time
Actual Tracking
Period Series Series RSFE MAD
Demand Signal
Forecast Residual
1 12 12.16 -0.16 -0.16 0.03 -5.00
2 13 12.13 0.87 0.72 0.20 3.58
3 10 12.09 -2.09 -1.38 0.58 -2.38
4 11 12.06 -1.06 -2.44 0.68 -3.61
5 10 12.03 -2.03 -4.47 0.95 -4.72
6 14 12.00 2.00 -2.47 1.16 -2.13
7 16 11.97 4.03 1.57 1.73 0.90
8 15 11.93 3.07 4.63 2.00 2.32
9 13 11.90 1.10 5.73 1.82 3.15
10 8 11.87 -3.87 1.86 2.23 0.84
11 10 11.84 -1.84 0.03 2.15 0.01
12 12 11.80 0.20 0.22 1.76 0.13
13 9 11.77 -2.77 -2.55 1.96 -1.30
14 13 11.74 1.26 -1.29 1.82 -0.71
15 13 11.71 1.29 0.00 1.72 0.00
 Old man winters
Winters method used to forecast one period into the future
See how method detects patterns & adapts to market changes over time

Old Man Winters in Action

600.00

500.00

400.00
Volum e

Actual
300.00
Forecast
200.00

100.00

0.00
0 20 40 60 80 100
Tim e
 Key to Winters method
• Winters is an exponential smoothing method

• Smoothing is based on a key idea

• For each component (which are?), a portion of difference
between estimate & actual is due to randomness & certain
portion due to real change
Smoothing in action...
• New estimate = old estimate + (some percentage)(error)

• Smoothes out peaks & valleys (i.e., randomness) of actual

Bernie’s insight…
…or what is focus forecasting?
• An intuitive & successful idea

• Regularly use a # of different methods to generate forecasts

• Maintain historical accuracy information on each method

• Use the most accurate method to generate “official”

forecasts
 Filtering
• An important feature of computer-based forecasting
systems
• Large amounts of data – impractical to manually review all

1. For data input errors (e.g., typos, scanner errors)

• If |“actual” - forecast| > limit, then report

2. For unacceptable forecast errors (e.g., warranting

management attention)
• If average absolute error > limit, then report
• If average error (i.e., bias) > limit, then report
 Demand Management
(Bill of Materials- BOM)
Independent Demand:
Finished Goods

A Dependent Demand:
Raw Materials,
Component parts,
B(4) C(2) Sub-assemblies, etc.

D(2) E(1) D(3) F(2)

Customer Order Decoupling Point
• Can be looked at as the point at which demand changes from
independent to dependent. It is the point at which the firm, as
opposed to the customer, becomes responsible for determining the
timing and quantity of material to be purchased, made, or finished.

• Engineered to order Suppliers

• Made to order Raw Materials inventory
• Assemble to order WIP
• Made to stock Finished Goods
 Decoupling Points & Lead Time

Make-to-Stock (MTS)
Short Finished Goods
Components/Subassemblies
Lead Time

Assemble to Order (ATO)

Raw Materials
Make to Order (MTO)
Long Suppliers Engineer to Order
 Demand Uncertainty…
how is it dealt with?
• MTS – Safety stocks of end items.
• ATO – Forecast product mix and calculate expected
components and sub-assemblies. Safety stock carried in
these items.
• MTO – Uncertainty involves the level of company resources
that will be required to complete the engineering and
produce the product once the requirements are
determined. May carry some raw materials.
 Web-Based Forecasting: CPFR
• Collaborative Planning, Forecasting, and Replenishment (CPFR) a
Web-based tool used to coordinate demand forecasting, production
and purchase planning, and inventory replenishment between
supply chain trading partners.
• Used to integrate the multi-tier or n-Tier supply chain, including
manufacturers, distributors and retailers.
• CPFR’s objective is to exchange selected internal information to
provide for a reliable, longer term future views of demand in the
supply chain.
• CPFR uses a cyclic and iterative approach to derive consensus
forecasts.
Web-Based Forecasting:
Steps in CPFR
1. Creation of a front-end partnership agreement.
2. Joint business planning
3. Development of demand forecasts
4. Sharing forecasts
5. Inventory replenishment
Correlational Forecasting

• Assumes an Outcome is Dependent an Existing

Relationship Between the Demand Variable and
Some other Independent Variable(s)
• Demand Variable is Dependent Variable
• Other Related Variables are Independent Variables
• Generally Expressed as a Multiple Linear Regression
Model
• Y =  + X1+ X2+ X2+ . . . nXn+ i
 Simple Linear Regression Model
Y

The simple linear regression a

model seeks to fit a line
through various data over time
0 1 2 3 4 5 x (Time)

Yt = a + bx Is the linear regression model

Yt is the regressed forecast value or dependent variable in the

model, a is the intercept value of the the regression line, and b is
similar to the slope of the regression line. However, since it is
calculated with the variability of the data in mind, its formulation
is not as straight forward as our usual notion of slope.
 Simple Linear Regression Formulas for
Calculating “a” and “b”

a = y - bx

 xy - n(y)(x)
b= 2 2
 x - n(x )
 Simple Linear Regression Problem Data

Question: Given the data below, what is the simple linear regression model
that can be used to predict sales in future weeks?

Week Sales
1 150
2 157
3 162
4 166
5 177
Answer: First, using the linear regression formulas, we can
compute “a” and “b”

Week Week*Week Sales Week*Sales

1 1 150 150
2 4 157 314
3 9 162 486
4 16 166 664
5 25 177 885
3 55 162.4 2499
Average Sum Average Sum

b=
 xy - n( y)(x) 2499 - 5(162.4)(3) 63
=  = 6.3
 x - n(x )
2 2
55  5(9 ) 10

a = y - bx = 162.4 - (6.3)(3) = 143.5

The resulting regression model
is: Yt = 143.5+6.3x
Now if we plot the regression generated forecasts against the
actual sales we obtain the following chart:
180
175
170
165
160 Sales
Sales

155 Forecast
150
145
140
135
1 2 3 4 5
Period
Statistical Assumptions of Multiple
Linear Regression
• The Error Term (the residual i) is Normally Distributed
• There is no Serial Correlation Among Error Terms
• Magnitude of the Error Term is Independent of the Size of
Any of the Independent Variables - Xi
• Assumptions Can be Tested Through Analyses of the
Residuals - i
Major Statistical Problems of Multiple
Linear Regression
• Multicolinarity
• Use of Time-Lagged Independent Variables
• Both of These Problems Result in Models with Potentially
Valid Predictions, but the Reliability of the  Coefficients
is Questionable
Finding The Right Technique
1. Select a variety of forecasting techniques
2. Use historic data with each technique to forecast demand for the
most recent known demand period.
3. See which forecasting technique gives you the most accurate
forecast.
4. Use that technique to forecast the unknown period of demand.
5. Repeat this selection process each time you need to forecast
demand. Use the latest demand data.
This process is called FOCUS FORECASTING.
Forecasting As a Process
The forecast process itself, typically done on a monthly basis, consists
of structured steps. They often are facilitated by someone who might
be called a demand manager, forecast analyst, or demand/supply
planner.

It is not simply a matter of running a computer model!

Update Prepare Consensus

historic data Initial meetings &
Forecasts collaboration

Finalize Review by
Revise
and Operating
Communicate Committee forecasts
6 5 4
 Some Principles For The Forecasting Process

Forecasting is being done in virtually every company. The challenge is to do

it better than the competition.
 Better forecasts result in better customer service and lower costs, as well as better
relationships with suppliers and customers.
The forecast must make sense based on the big picture, economic outlook,
market share, and so on. Context is critical.
The best way to improve forecast accuracy is to focus on reducing forecast
error.
Whenever possible, forecast aggregate levels. Forecast in detail only where
necessary.
Use Judgment! Far more can be gained by people collaborating,
communicating well, and using judgment, than by using the most advanced
forecasting technique or model.
 Question Bowl

Which of the following is a classification of a basic type of

forecasting?
a. Transportation method
b. Simulation
c. Linear programming
d. All of the above
e. None of the above

Answer: b. Simulation (There are four types including Qualitative,

Time Series Analysis, Causal Relationships, and Simulation.)
 Question Bowl

Which of the following is an example of a “Qualitative” type

of forecasting technique or model?
a. Grass roots
b. Market research
c. Panel consensus
d. All of the above
e. None of the above

Answer: d. All of the above (Also includes Historical

Analogy and Delphi Method.)
 Question Bowl

Which of the following is an example of a “Time Series

Analysis” type of forecasting technique or model?
a. Simulation
b. Exponential smoothing
c. Panel consensus
d. All of the above
e. None of the above

Answer: b. Exponential smoothing (Also includes Simple Moving Average,

Weighted Moving Average, Regression Analysis, Box Jenkins, Shiskin Time Series,
and Trend Projections.)
 Question Bowl

Which of the following is a reason why a firm should choose

a particular forecasting model?
a. Time horizon to forecast
b. Data availability
c. Accuracy required
d. Size of forecasting budget
e. All of the above

Answer: e. All of the above (Also should include “availability of

qualified personnel” .)
 Question Bowl

Which of the following are ways to choose weights in a

Weighted Moving Average forecasting model?
a. Cost
b. Experience
c. Trial and error
d. Only b and c above
e. None of the above

Answer: d. Only b and c above

 Question Bowl

Which of the following are reasons why the Exponential

Smoothing model has been a well accepted forecasting
methodology?
a. It is accurate
b. It is easy to use
c. Computer storage requirements are small
d. All of the above
e. None of the above

Answer: d. All of the above

 Question Bowl

The value for alpha or α must be between which of the

following when used in an Exponential Smoothing
model?
a. 1 to 10
b. 1 to 2
c. 0 to 1
d. -1 to 1
e. Any number at all

Answer: c. 0 to 1
 Question Bowl

Which of the following are sources of error in forecasts?

a. Bias
b. Random
c. Employing the wrong trend line
d. All of the above
e. None of the above

Answer: d. All of the above

 Question Bowl

Which of the following would be the “best” MAD values in

an analysis of the accuracy of a forecasting model?
a. 1000
b. 100
c. 10
d. 1
e. 0

Answer: e. 0
 Question Bowl

If a Least Squares model is: Y=25+5x, and x is equal to 10,

what is the forecast value using this model?
a. 100
b. 75
c. 50
d. 25
e. None of the above

Answer: b. 75 (Y=25+5(10)=75)
 Question Bowl

Which of the following are examples of seasonal variation?

a. Additive
b. Least squares
c. Standard error of the estimate
d. Decomposition
e. None of the above

Answer: a. Additive (The other type is of seasonal

variation is Multiplicative.)
Concluding Principles
• Data capture must not be limited to sales but should include
domain info such as knowledge, trends, systems
performance
• Forecasting models should not be more complicated than
necessary. Simple models work just as good.
• Forecast from different sources must be reconciled and
made consistent with firm plans and constraints.
 Concluding Principles

• Input data and output forecasts should be routinely

monitored for quality and appropriateness.
• Information on sources of variation should be incorporated
into the forecasting system.
• Forecast from different sources must be reconciled and
made consistent with firm plans and constraints.