SCM350: OPERATIONS MANAGEMENT

Instructor: Dr. Mohammed Hejazi

Ch.3 Forecasting

1
 q Forecast  – a  statement  about  the  future  value  of  
a  variable  of  interest
q We  make  forecasts  about  such  things  as  
weather,  demand,  and  resource  availability
q Forecasts  are  important  to  making  informed  
decisions

2
Types of Forecasting

q Economic  forecasts
q Address  business  cycle  – inflation  rate,  money  
supply,  housing  starts,  etc.
q Technological  forecasts
q Predict  rate  of  technological  progress
q Impacts  development  of  new  products
q Demand  forecasts
q Predict  sales  of  existing  products  and  services

3
1. Short-­range  forecast
▶ Up  to  1  year,  generally  less  than  3  months
▶ Purchasing,  job  scheduling,  workforce  levels,  job  
assignments,  production  levels
2. Medium-­range  forecast
▶ 3  months  to  3  years
▶ Sales  and  production  planning,  budgeting
3. Long-­range  forecast
▶ 3+ years
▶ New  product  planning,  facility  location,  research  and  
development
4
Elements of Good Forecasting

Timely

Reliable Accurate

Written

5
Steps in the Forecasting Process

“The  forecast”

Step  6  Monitor  the  forecast

Step  5  Prepare  the  forecast
Step  4  Gather  and  analyze  data
Step  3  Select  a  forecasting  technique
Step  2  Establish  a  time  horizon
Step  1  Determine  purpose  of  forecast
6
Forecast Accuracy and Control

q Allowances  should  be  made  for  forecast  errors

q It  is  important  to  provide  an  indication  of  the  extent  to  

which  the  forecast  might  deviate  from  the  value  of  the  
variable  that  actually  occurs

q Forecast  errors  should  be  monitored

q Error  =  Actual  – Forecast

q If  errors  fall  beyond  acceptable  bounds,  corrective  action  

may  be  necessary

7
q Mean absolute deviation (MAD) The average absolute forecast
error.

q Mean squared error (MSE) The average of squared forecast

errors.

q Mean absolute percent error (MAPE) The average

absolute percent error.

MAD weights all errors
evenly

MSE weights errors according

to their squared values

MAPE weights errors

according to relative error
 Actual Forecast (A-­F)  
Period
(A) (F) Error |Error| Error2 [|Error|/Actual]x100

1 107 110

2 125 121

3 115 112

4 118 120

5 108 109

Sum

MAD MSE MAPE

 Actual Forecast (A-­F)  
Period
(A) (F) Error |Error| Error2 [|Error|/Actual]x100

1 107 110 -­3 3 9 2.80%

2 125 121 4 4 16 3.20%

3 115 112 3 3 9 2.61%

4 118 120 -­2 2 4 1.69%

5 108 109 1 1 1 0.93%

Sum 13 39 11.23%

n  =  5 n-­1  =  4 n  =  5

MAD MSE MAPE

=  2.6 =  9.75 =  2.25%

 Features Common to All
Forecasts
q Techniques  assume  some  underlying  causal  system  
that  existed  in  the  past  will  persist  into  the  future

q Forecasts  are  not  perfect

q Forecasts  for  groups  of  items  are  more  accurate  than  

those  for  individual  items

q Forecast  accuracy  decreases  as  the  forecasting  

horizon  increases
Forecasting Approaches/Methods
Product Life Cycle
Forecasting During the Life Cycle
1. Qualitative Forecasting Methods
q Qualitative  Forecasting
q Qualitative  techniques  permit  the  inclusion  of  :
q Human  factors

q Personal  opinions
q Hunches

q Used  when  situation  is  vague  and  little  data  exist

q New  products
q New  technology

q These  factors  are  difficult,  or  impossible,  to  quantify

q Quantitative  Forecasting
q These  techniques  rely  on  hard data

q Quantitative  techniques  involve  either  the  projection  of  historical  data  or  the  
development  of  associative  methods  that  attempt  to  use  causal  variables to  make  a  
forecast
 Executive Opinion

q Involves  small  group  of  high-­level  experts  

and  managers
q Group  estimates  demand  by  working  
together
q Combines  managerial  experience  with  
statistical  models
q Relatively  quick
Slide  19
 Sales Force Composite/
Estimates
q Each  salesperson  projects  his  or  her  
sales
q Combined  at  district  and  national  
levels
q Sales  reps  know  customers’  wants
q May  be  overly  optimistic
 Market Survey

q Ask  customers  about  purchasing  

plans
q Useful  for  demand  and  product  
design  and  planning
q What  consumers  say,  and  what  they  
actually  do  may  be  different
q May  be  overly  optimistic
Summary of Qualitative Forecast methods

q Forecasts  that  use  subjective  inputs  such  as  opinions  from  consumer  
surveys,  sales  staff,  managers,  executives,  and  experts
q Executive  opinions
q a  small  group  of  upper-­level  managers  may  meet  and  collectively  develop  a  
forecast
q Sales  force  opinions
q members  of  the  sales  or  customer  service  staff  can  be  good  sources  of  
information  due  to  their  direct  contact  with  customers  and  may  be  aware  of  plans  
customers  may  be  considering  for  the  future
q Consumer  surveys
q since  consumers  ultimately  determine  demand,  it  makes  sense  to  solicit  input  
from  them
q consumer  surveys  typically  represent  a  sample of  consumer  opinions
q Other  approaches
q managers  may  solicit  opinions  from  other  managers  or  staff  people  or  outside  
experts  to  help  with  developing  a  forecast.  
q the  Delphi  method  is  an  iterative  process  intended  to  achieve  a  consensus
2. Quantitative Forecasting Methods

Slide  23
Quantitative Methods

q Used  when  situation  is  ‘stable’  and  

historical  data  exist
q Existing  products
q Current  technology
q Involves  mathematical  techniques
q e.g.,  forecasting  sales  of  HTC  
smartphones  
Overview of Quantitative Approaches

1. Naive  approach

2. Moving  averages Time-­series  

models
3. Exponential  smoothing

4. Trend  projection
Associative  
5. Linear  regression model
 Time-­Series Forecasting
q Set  of  evenly  spaced  numerical  data

q Obtained  by  observing  response  variable  

at  regular  time  periods

q Forecast  based  only  on  past  values,  no  other  

variables  important

q Assumes  that  factors  influencing  past  and  

present  will  continue  influence  in  future
Time-­Series Components

Trend Cycle

Seasonal Variation
 Trend Component

q Persistent,  overall  upward  or  downward  

pattern

q Changes  due  to  population,  technology,  

age,  culture,  etc.

q Typically  several  years  duration  

 Seasonal Component

q Regular  pattern  of  up  and  down  fluctuations

q Due  to  weather,  customs,  etc.

q Occurs  within  a  single  year  

PERIOD  LENGTH “SEASON”  LENGTH NUMBER  OF  “SEASONS”  IN  

PATTERN
Week Day 7
Month Week 4  – 4.5
Month Day 28  – 31
Year Quarter 4
Year Month 12
Year Week 52
 Cycle Component

q Repeating  up  and  down  movements

q Affected  by  business  cycle,  political,  and  economic  

factors

q Multiple  years  duration

q Often  causal  or  

associative  
relationships

0 5 10 15 20
 Variation Component
q Short  duration  and  nonrepeating  
q Irregular  variation  Caused
q by  unusual  circumstances,  not
reflective  of  typical  behavior.

q Random  variations  
q Residual  variations  after  all  other  behaviors  
are  accounted  for.
q Erratic,  unsystematic,  ‘residual’  fluctuations.
Trends and Seasonality

• Trend

– A long-term upward or downward movement in data

• Population shifts

• Changing income

• Seasonality

– Short-term, fairly regular variations related to the calendar or

time of day

– Restaurants, service call centers, and theaters all experience

seasonal demand
Cycles and Variations
q Cycle
q Wavelike  variations  lasting  more  than  one  year
q These  are  often  related  to  a  variety  of  economic,  political,  or  even  agricultural  
conditions

q Irregular  variation
q Due  to  unusual  circumstances  that  do  not  reflect  typical  behavior

q Labor  strike

q Weather  event

q Random  Variation
q Residual  variation  that  remains  after  all  other  behaviors  have  been  accounted  for
Components of Demand

Slide  34
 1.Naive Approach
q Assumes  demand  in  next  
period  is  the  same  as  
demand  in  most  recent  period
q e.g.,  If  January  sales  were  68,  then  February  sales  
will  be  68
q Sometimes  cost  effective  and  efficient
q Can  be  good  starting  point
q Weakness:  doesn’t  smooth  the  data.  
q Data  smoothing:  attempts  to  capture  
important patterns in  the  data,  while  leaving  out noise
Naïve (Cont.)

q Check  if  the  resulting  accuracy  is  acceptable

q The  higher  the  accuracy,  often  the  higher  the  cost.

q Do  we  really  need  our  forecast  that  accurate?  Is  it  worth  the  
additional  resources?

qWhy  do  you  need  forecasts  for?  How  critical  they  are  for  
operations?

36
2. Moving Average

• Assumes  an  average  is  a  good  estimator  of  future  behavior

– Used  if  little  or  no  trend

– Used  for  smoothing

A t + A t -1 + A t -2 + ... + A t -n +1
Ft +1 =
n
 Moving Average Method (cont.)
q MA  is  a  series  of  arithmetic  means  
q Assumes  an  average  is  a  good  estimator  of  future  behavior
q Used  if  little  or  no  trend
q Used  often  for  smoothing
q Provides  overall  impression  of  data  over  time

∑ demand in previous n periods

Moving average =
n
A t + A t -1 + A t -2 + ... + A t -n +1
Ft +1 =
n
Ft+1 = Forecast for the upcoming period, t+1
n = Number of periods to be averaged
At = Actual occurrence in period t
Moving Average Example

MONTH ACTUAL  SHED  SALES 3-­MONTH  MOVING  AVERAGE

January 10 10
February 12 12
March 13 13
(10 +  12 +  13)/3  =  11  2/3
April 16
(12  +  13  +  16)/3  =  13  2/3
May 19
(13  +  16  +  19)/3  =  16
June 23
(16  +  19  +  23)/3  =  19  1/3
July 26
(19  +  23  +  26)/3  =  22  2/3
August 30
(23  +  26  +  30)/3  =  26  1/3
September 28
(29  +  30  +  28)/3  =  28
October 18
(30  +  28  +  18)/3  =  25  1/3
November 16
(28  +  18  +  16)/3  =  20  2/3
December 14
Moving Average Ft +1 =
A t + A t -1 + A t -2 + ... + A t -n +1
n
You’re  manager  in  Amazon’s  electronics  
department.  You  want  to  forecast  ipod sales  for  
months  4-­6  using  a  3-­period  moving  average.

Sales Moving Average

Month (000) (n=3)
1 4 NA
2 6 NA
3 5 NA
4 ? (4+6+5)/3=5
5 ?
6 ?
Moving Average
Forecast for Month 5?

Sales Moving Average

Month (000) (n=3)
1 4 NA
2 6 NA
3 5 NA
4 3 5
5 ? (6+5+3)/3=4.667
6 ?
Moving Average -­-­ example
MONTH Demand Month Demand
January 89 July 223
February 57 August 286
March 144 September 212
April 221 October 275
May 177 November 188
June 280 December 312

q 3 month MA: (oct+nov+dec)/3=258.33

q 6 month MA: (jul+aug+…+dec)/6=249.33

q 12 month MA: (Jan+feb+…+dec)/12=205.33

Summary of Moving Averages
q Advantages of Moving Average Method

q Easily understood

q Easily computed

q Provides stable forecasts

q Disadvantages of Moving Average Method

q Requires saving lots of past data points: at least the N periods

used in the moving average computation

q Lags behind a trend

q Ignores complex relationships in data

What about Weighted Moving Averages?

q This  method  looks  at  past  data  and  tries  to  logically  attach  
importance  to  certain  data  over  other  data

q Weighting  factors  must  add  to  one

q Can  weight  recent  higher  than  older  or  specific  data  above  
others
q If  forecasting  staffing,  we  could  use  data  from  the  last  four  weeks  
where  Tuesdays  are  to  be  forecast.

q Weighting  on  Tuesdays  is:  T-­1 is  .25;;  T-­2 is  .20;; T-­3    is  .15;;  T-­4 is  .10  
and  Average  of  all  other  days  is  weighed  .30.
Weighted Moving Average

q Used  when  some  trend  might  be  present  

q Older  data  usually  less  important

q Weights  based  on  experience  and  intuition

((
Weighted   ∑ Weight for period n Demand in period n
moving   =
)( ))
average ∑ Weights
Weighted Moving Average

qGives more emphasis to recent data

qWeights
q decrease for older data

Ft +1 = w1A t + w 2 A t -1 + w 3A t -2 + ... + w n A t -n +1

Simple  moving
average  models
weight  all  previous
periods  equally
Weight MA Example
a) Compute  a  weighted  average  forecast  using  a  weight  of  .40  for  the  
most  recent  period,  .30  for  the  next  most  recent,  .20  for  the  next,  
and  .10  for  the  next.
b) If  the  actual  demand  for  period  6  is  39,  forecast  demand  for  period  
7  using  the  same  weights  as  in  part  a.
Period Demand
1 42
2 40
3 43
4 40
5 41
Exponential Smoothing

• A weighted averaging method that is based on the previous

forecast plus a percentage of the forecast error

Ft = Ft -1 + a ( At -1 - Ft -1 )

where
Ft = Forecast for period t
Ft -1 = Forecast for the previous period
a = Smoothing constant
At -1 = Actual demand or sales from the previous period