Chapter 15

Demand Management and Forecasting

McGraw-Hill/Irwin © 2011 The McGraw-Hill Companies, All Rights Reserved

Learning Objectives

      Understand the role of forecasting as a basis for supply chain planning.

Compare the differences between independent and dependent demand.

Identify the basic components of independent demand: average, trend, seasonal, and random variation.

Describe the common qualitative forecasting techniques such as the Delphi method and Collaborative Forecasting.

Show how to make a time series forecast using regression, moving averages, and exponential smoothing.

Use decomposition to forecast when trend and seasonality is present.

15-2

Characteristics of Forecasts

  Guessing at the future: educated guessing game   Seldom correct No perfect forecast Objective is to minimize forecast errors    It is only a tool used to set: Production plan and budgets Work schedules   Forecasts are more accurate in aggregation Long-term forecasts are less accurate than short-term forecasts  Forecasts are means to an end

15-3

Demand Management

LO 2    

Strategic forecasts

: forecasts used to help set the strategy of how demand will be met

Tactical forecasts

day basis : forecasted needed for how a firm operates processes on a day-to The purpose of demand management is to coordinate and control all sources of demand   Two basic sources of demand

Dependent demand

services : the demand for a product or service caused by the demand for other products or

Independent demand

other products : the demand for a product or service that cannot be derived directly from that of

15-4

LO 1

Demand Management

Continued

  Not much a firm can do about dependent demand It is demand that must be met    There is a lot a firm can do about independent demand Take an active role to influence demand  Offer incentive to customers  Wage campaigns to sell products Take a passive role and respond to demand  Especially if at full capacity  High cost of advertisement

15-5

Types of Forecasts

   Basic types of forecasts Quantitative — use historical data    Time series analysis Causal relationships Simulation Qualitative — based on subjective estimates/opinion LO 1  Time series analysis is based on the idea that data relating to past demand  can be used to predict future demand Primary focus of this chapter

15-6

Components of Demand

LO 3  Average demand for a period of time  Trend  Seasonal element  Cyclical elements  Random variation  Autocorrelation

15-7

LO 3

Common Types of Trends

15-8

Time Series Analysis

LO 5 

Short term

: forecast under three  months Tactical decisions 

Medium term

: three months to two  years Capturing seasonal effects 

Long term

: forecast longer than two   years Detecting general trends Identifying major turning points

15-9

LO 5

A Guide to Selecting an Appropriate Forecasting Method

15-10

LO 5

Pick Forecasting Model Based On

 Time horizon to forecast  Data availability  Accuracy required  Size of forecasting budget  Availability of qualified personnel

15-11

LO 5

Linear Regression Analysis

TC

 

FC



VC TC

 80000  75

X

Regression: functional relationship between two or more correlated variables      

Y = a + bX Y a b

is the value of the dependent variable is the Y intercept is the slope

X

is the independent variable  It is used to predict one variable given the other Assumes data falls in a straight line

15-12

LO 5

Example 15.1: The Data and Least Squares Regression Line

15-13

LO 5

Example 15.1: Equations and Calculating Totals

15-14

LO 5

Example 15.1: Calculating the Forecast

Y Y

13

Y

14

Y

15

Y

16     

a



bx

441 .

6 441 .

6 441 .

6 441 .

6     359 .

359 .

359 359 .

.

6 6 6 6             5 , 116 .

4 5 , 476 .

0 5 , 835 .

6 6 , 195 .

2

15-15

Calculating the Forecast

160 155 150 145 180 175 170 165 Week Sales Forecast 1 2 3

Week

4 5

Y

= 143.5 + 6.3

x

What is forecast for

x

=100?

Y

= 143.5 + 6.3(100) = 774

15-16

LO 6

Decomposition of a Time Series



Time series

: chronologically ordered data that may contain one or more components of demand 

Decomposition

: identifying and separating the time series data into these components    Seasonal variation

Additive

: the seasonal amount is constant

Multiplicative

: the seasonal variation is a percentage of demand

15-17

LO 6

Additive and Multiplicative Seasonal Variation Superimposed on Changing Trend

15-18

Example 15.3: The Data and Hand Fitting

Y



a



bx

 170  55

x

LO 6

15-19

LO 5

Example 15.3: Computing Seasonal Factors and Computing Forecast

15-20

Decomposition Using Least Squares Regression

LO 6  Determine the seasonal factor  Deseasonalize the original data  Develop a least squares regression line for the deseasonalized data  Project the regression line through the period of the forecast  Create the final forecast by adjusting the regression line by the seasonal factor

15-21

LO 6

Steps 1-3 Deseasonalized Demand

15-22

LO 6

Steps 4 – 5

15-23

Simple Moving Average

LO 5  Useful when demand is neither growing nor declining rapidly and does not have seasonal characteristics  Moving averages can be centered or used to predict the following period    Important to select the best period Longer gives more smoothing/less sensitive Shorter reacts quicker to trends

15-24

Simple Moving Average Formula

F t



A t

 1 

A t

 2 

A t

 3   

A t



n n

LO 5

F t

 Forecast for the coming period n  Number of periods to be averaged

A t

 2 ,

A t

 3 and

A t

 1  Actual occurrence in the past period

A t



n

 Actual occurrence s two periods ago, three periods ago, and so on up to n periods ago

15-25

LO 5

Forecast Demand Based on a Three- and a Nine-Week Simple Moving Average

15-26

Forecast Demand Based on a Three- and a Nine-Week Simple Moving Average

15-27

Weighted Moving Average

  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 All the weights must sum to one if fractions Otherwise, weights can be real numbers. If so divide by sum of weights: 

F t

 

w i



D t

 1   

w i

 LO 5 F = w A t 1 t -1 + w A 2 t - 2 + w A 3 t -3 + ...+ w A n t - n

15-28

WMA Example

Question: Given the weekly demand information and weights of 0.6, 0.1, and 0.3, what is the weighted moving average forecast for the 5 th period or week?

Week 1 2 3 4 Demand 820 775 680 655

F

5 = (0.6)(655)+(0.1)(680)+(0.3)(755)= 688

15-29

Choosing Weights

LO 5  Experience and trial-and-error are the simplest ways  Generally, the most recent past is the best indicator  When data are seasonal, weights should be established accordingly

15-30

Exponential Smoothing

LO 5  Most used of all forecasting techniques  Integral part of all computerized forecasting programs  Widely used in retail and service        Widely accepted because… Exponential models are surprisingly accurate Formulating an exponential model is relatively easy The user can understand how the model works Little computation is required to use the model Computer storage requirements are small Tests for accuracy are easy to compute

15-31

Exponential Smoothing Model

F t = F t-1

 (

A t-1 - F t-1

) Where:

F t F t

 1

A t

 1  = Forecast value for the coming

t

time period = Forecast value in 1 past time period = Actual occurrence in the past 1 time period = 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 LO 5

15-32

LO 5

Exponential Smoothing Example (



=0.20)

Week 1 2 3 4 5 6 7 8 9 10 Demand 820 775 680 655 750 802 798 689 775

0.2

820.00

820.00

811.00

784.80

758.84

757.07

766.06

772.45

755.76

759.61

820  820  0 .

2 0 .

2  820     820 820   820 820   0 .

0 .

2 2   775   45  820    820  811  811  9 .

0 .

2 .

2     811 680  131  811    811  26 .

2  784 .

8

15-33

ES Example (



=0.10, 0.60)

Week 1 6 7 8 9 10 2 3 4 5 Demand 820 775 680 655 750 802 798 689 775  

0.1

820.00

820.00

815.50

801.95

787.26

783.53

785.38

786.64

776.88

776.69

 

0.6

820.00

820.00

793.00

725.20

683.08

723.23

770.49

787.00

728.20

756.28

15-34

850 800

ES Example (



=0.10, 0.60)

Note how the smaller alpha results in a smoother line in this example 750 700 650 600 1 2 3 4 5

Week

6 7 8 9 Demand Alpha=0.1

Alpha=0.6

10

15-35

Trend Effects in Exponential Smoothing

LO 5  An trend in data causes the exponential forecast to always lag the actual data  Can be corrected somewhat by adding in a trend adjustment  To correct the trend, we need two   smoothing constants Smoothing constant alpha (  ) Trend smoothing constant delta ( δ)

15-36

LO 5

Exponential Forecasts versus Actual Demand over Time Showing the Forecast Lag

15-37

Trend Effects Equations

FIT t

F t T t

  

F t



T t FIT t T t

 1   1    

F t

 

A t

 1 

FIT t FIT t

 1   1  LO 5 F t  The exponentia T t  The exponentia lly lly smoothed smoothed FIT t  The forecast including FIT t 1  The forecast including trend trend forecast for made for period t for the period trend for period t prior t period A t 1  The actual demand   Smoothing constant for the prior period   Smoothing constant

15-38

Forecast Error

 Sources of errors     Projecting the past into the future Wrong relationships Wrong information (data) Errors outside of our control  Goal is to minimize the errors

15-39

Forecast Error

LO 5 

Bias errors

: when a consistent mistake is made 

Random errors

: errors that cannot be explained by the forecast model being used     Measures of error Mean absolute deviation (MAD) Mean absolute percent error (MAPE) Tracking signal

15-40

LO 5

The MAD Statistic to Determine Forecasting Error

 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 = n  t=1 A - F t t n 1 MAD  0.8 standard deviation 1 standard deviation  1.25 MAD

15-41

Example: Find the MAD

Month Sales 1 220 2 3 4 5 250 210 300 325 Forecast Abs Error

— —

255 205 320 315

5 5 20 10 Total = 40 MAD = n  t=1 A - F t t n = 40 = 10 4 Note that by itself, the MAD only lets us know the mean error in a set of forecasts

15-42

Tracking Signal

LO 5  The tracking signal (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 TS = RSFE = MAD Running sum of forecast errors Mean absolute deviation

15-43

LO 5

Computing the MAD, the RSFE, and the TS from Forecast and Actual Data

15-44

Example: Tracking Signal

+ 4 3 2 1

TS

0 - 1 - 2 - 3

- 4 - 5 - 6

Period 5 6 1 2 3 4 Forecast 250 325 400 350 375 450 Demand 200 250 325 300 325 400 Error - 50 - 75 - 75 - 50 - 50 - 50 |E| 50 75 75 50 50 50 RSFE - 50 - 125 - 200 - 250 - 300 - 350

1 2 3 4 5

Sum |E| 50 125 200 250 300 350 MAD 50.0

62.5

66.7

62.5

60.0

58.3

TS - 1 - 2 - 3 - 4 - 5 - 6 Out of Control

6 Period

15-45

Causal Relationship Forecasting

LO 5 

Causal relationship forecasting

: using independent variables other than time to  predict future demand The independent variable must be a leading indicator  Must find those occurrences that are really the causes

15-46

Qualitative Techniques in Forecasting

 Qualitative forecasting techniques take advantage of the knowledge of experts  Most useful when the product is new or there is little experience with selling into a new region LO 4  The following are samples of qualitative       forecasting techniques Executive judgment Grass roots Market research Panel consensus Historical analogy Delphi method

15-47

Qualitative Methods

Executive Judgment • Used for new products introduction • Decisions are broader and at a higher level Historical analogy • Existing product used as model for

another

• Example: buying CDs on Internet put

you on mailing list for related products

Delphi Method • Based on expert opinion • Experts asked question anonymously • Goes thru several rounds of questioning • Results tabulated, iterated until a

consensus is reached

Qualitative Methods Grass Roots • Builds forecast by adding

successively from bottom

• Those closest to customer know

better

Market Research • Consumer surveys and interviews • Used to improve existing products Panel Consensus • Open meetings with free exchange of

ideas

• Power play possibilities

15-48

Web-Based Forecasting: (CPFR)

LO 5 

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   Objective is to exchange selected internal information to provide for a reliable, longer term future views of demand CPFR uses a cyclic and iterative approach to derive consensus forecasts

15-49

Web-Based Forecasting: Steps in CPFR

LO 5  Creation of a front-end partnership agreement  Joint business planning  Development of demand forecasts  Sharing forecasts  Inventory replenishment

15-50