Docstoc

Collaborative Planning_ Forecasting_ and Replenishment

Document Sample
Collaborative Planning_ Forecasting_ and Replenishment Powered By Docstoc
					    Demand Forecasting in a
        Supply Chain




Frank Davis   Demand Forecasting in a Supply Chain   7/25/2002   1
           Why do you forecast?
• Who is involved in forecasting?
  – Marketing – Why? Do they influence
    forecast?
  – Production – Why? How do they influence?
  – Distribution – Why? How do they influence?
  – Channel Members – Why? How do they
    influence?
  – Suppliers – Why? How do they influence?

   Frank Davis   Demand Forecasting in a Supply Chain   7/25/2002   2
   Characteristics of Forecasts
• Why are forecasts always wrong? Does this
  mean we should not forecast? What does it
  mean?
• Why are long-term forecasts less accurate than
  short-term forecasts?
• Why are aggregate forecasts typically more
  accurate than disaggregate forecasts? Are
  there cases when this would not be the case?
• Who needs to make forecasts? Should
  everyone in the supply chain use the same
  forecast?

   Frank Davis   Demand Forecasting in a Supply Chain   7/25/2002   3
                   When do you use?
•   Qualitative – subjective judgment call
•   Time series – crank the numbers
•   Causal – correlate with known variables
•   Simulation – combine various methods
•   How do you determine which method to use?
•   Would you use the same method to forecast
    – the outcome of the Miami football game
    – the amount of Coke to produce and
    – staffing for a hospital emergency room?
     Frank Davis    Demand Forecasting in a Supply Chain   7/25/2002   4
                 Basis for a forecast
• What do you use for a forecast?
• How do you forecast score of upcoming football
  game?
  – Qualitative - Poll sportscasters (experts)
  – Time series - Look at scores of last 10 games
       • Level of scoring
       • Trends
       • Does schedule make difference
  – Causal – look at player and coaching data
  – Combination – time series plus modification by player
    and coaching data

   Frank Davis     Demand Forecasting in a Supply Chain   7/25/2002   5
                 Basic Questions
• What is the objective of forecasting?
   – Why is the forecast horizon important?
• Should all groups use the same forecast?
• Should demand forecasts be based on sales?
• Why is it important to identify the factors that
  influence demand?
• When would you have different forecasts for
  different customer segments?
• Which forecasting method is best?
• How do you determine how good a specific
  method is?
   Frank Davis   Demand Forecasting in a Supply Chain   7/25/2002   6
                Forecasting method
• Static – once level, trend and seasonal
  factors determined keep using the same
  formula
• Adaptive – new data may reveal
  something about level, trend and seasonal
  changes – recalculate new formula each
  time


  Frank Davis     Demand Forecasting in a Supply Chain   7/25/2002   7
                 Time Series - Static
• Information needed
  – Demand level
  – Demand trend
  – Cyclical effect
  – Error
• Modify forecast by causal factors



   Frank Davis     Demand Forecasting in a Supply Chain   7/25/2002   8
    Static time series forecasting
•   Four steps
    –   Deseasonalize know demands to prime regression model
        •   Must cover full cycle (each season)
        •   Average for each season if odd number of seasons
        •   See eq. 4.2 for even number of seasons
    –   Use deseasonalized demand to calculate level and trend
        •   use regression to calculate intercept and trend
    –   Use intercept and trend to forecast deseasonalized demand
    –   Calculate Seasonal Factor (actual demand/forecast)
    –   Calculate average seasonal factor
    –   Calculate seasonal forecast
        •   Use average seasonal factor to adjust trend [(level + trends * period) * average
            seasonal factor]
    –   See if forecast is good
        •   How big is your error?
        •   Is the forecast bias? (positive or negative)
        •   How much confidence can you have in forecast?
    –   Spreadsheet to illustrate class problem (Save this to you hard drive so you can
        work on it.)


    Frank Davis         Demand Forecasting in a Supply Chain              7/25/2002            9
Relationship between Beginning,
  End and Average of Period
      25

      20

      15

      10

        5

        0
        1-Jan                    Ave                   31-Jan


Frank Davis     Demand Forecasting in a Supply Chain     7/25/2002   10
   Average for three Months

      30

      25

      20

      15                                                         Sales

      10

        5

        0
        1-Jan       1-Feb           1-Mar              1-Apr


Frank Davis     Demand Forecasting in a Supply Chain           7/25/2002   11
         Average for 4 months

      35
      30
      25
      20
                                                             Sales
      15
      10
        5
        0
       Jan-03   Feb-03     Mar-03      Apr-03     May-03


Frank Davis     Demand Forecasting in a Supply Chain       7/25/2002   12
           Seasonal Adjustments
• If one season how
  would you determine
  average demand for                    8000
                                        7000
  season                                6000
                                        5000
                                        4000
                                                                    Sales
                                        3000
                                        2000
                                        1000
                                           0
                                               Summer




   Frank Davis   Demand Forecasting in a Supply Chain   7/25/2002     13
            Seasonal Adjustment
• If you have an odd                    25000



  number of seasons in
                                        20000
  the cycle how would
  you determine                         15000

  average sales rate in                                                     Summer
                                                                            Fall

  the middle of the                     10000
                                                                            Winter



  cycle?
                                         5000




                                           0
                                                        Sales



   Frank Davis   Demand Forecasting in a Supply Chain           7/25/2002        14
           Seasonal Adjustments
• If you have a 4
  season cycle how
  would you calculate                   35000
                                        30000
  average sales for the                 25000
  middle of the fall                    20000
                                                                           Sales
                                        15000
  quarter?                              10000
                                                                           3-D Column 2


                                         5000
                                            0
                                             Spring     Fall   Spring




   Frank Davis   Demand Forecasting in a Supply Chain          7/25/2002            15
         How do you do Seasonal
              Adjustment?
• Odd number of seasons in cycle
  – See equations 4,2 bottom
• Even number of seasons in cycle
  – Equation 4.2 top
  – Why can’t you calculate first 2 seasons
  – Why can’t you calculate last 2 seasons




  Frank Davis   Demand Forecasting in a Supply Chain   7/25/2002   16
             Data hard to interpret
Quarter Demand for Natural Gas.com

  Year     Quarter   Period t   Demand
  1998       2          1          8000
  1998       3          2         13000
  1998       4          3         23000
  1999       1          4         34000
  1999       2          5         10000
  1999       3          6         18000
  1999       4          7         23000
  2000       1          8         38000
  2000       2          9         12000
  2000       3         10         13000
  2000       4         11         32000
  2001       1         12         41000



    Frank Davis      Demand Forecasting in a Supply Chain   7/25/2002   17
                 Plot helps see periods
                             Quarterly Demand

         50000
         40000
Demand




         30000
         20000
         10000
            0
                   1   2      3    4    5    6     7    8    9    10   11   12
                                            Quarter




     Frank Davis           Demand Forecasting in a Supply Chain        7/25/2002   18
  Adaptive Model Steps (adjust as
             you go)
• Initialize just like static
   – Level
   – Trend
   – Cyclical
• Forecast
   – Prior forecast adjusted by actual demand for
     period
• Estimate error
• Modify forecast based on prior error
   Frank Davis   Demand Forecasting in a Supply Chain   7/25/2002   19
            Adaptive Model Steps
• Moving average
   – Average of n preceding periods (Eq. 4.9)
   – Forecast equal to average of last n periods
   – When is the moving average appropriate?
• Exponential Smoothing
   – Forecast = α(prior forecast) + (1- α) last demand
        • Concept (adjust last forecast by current experience)
   – Use same approach on
        • Level
        • Trend – Holt’s model
        • Trend and Season – Winter’s model
• Which method is best?
• Can you modify forecast to reflect other casual factors?
    Frank Davis     Demand Forecasting in a Supply Chain   7/25/2002   20
       How do you determine best
         forecasting method?
• What is purpose of forecast? What are
  you trying to do?
• What method do you use to evaluate value
  of forecast?
• What do you look for?
  – Mean absolute error
  – Bias
• What do you do if error is too high?
   Frank Davis   Demand Forecasting in a Supply Chain   7/25/2002   21
       Expectations of this class
• When to forecast
• Different forecasting methods
• Forecasting horizon
• What do you forecast
• Availability of tools to assist you but you
  need to know how to evaluate each
  method
• How do you cope with forecast error?
    Frank Davis   Demand Forecasting in a Supply Chain   7/25/2002   22
• Do you need to know how to calculate
  each model?
  – Firms will have software packages
  – You need to understand them conceptually
  – Advanced classes will go into more detail
• You do need to know that the model is not
  as important as knowing how to check for
  accuracy of method – error testing

  Frank Davis   Demand Forecasting in a Supply Chain   7/25/2002   23

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:4
posted:1/23/2012
language:English
pages:23