ch 15 demand manmagement & forecasting-hk

Upload: shashank-gupta

Post on 05-Apr-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    1/20

    1

    Chapter 15. Demand Management and Forecasting

    Demand Management Independent vs. dependent demand

    Qualitative Techniques Market research, focus groups, Delphi technique,

    Quantitative Techniques 1. Time series based models

    2. Associative (causal) Models

    Accuracy and Control of Forecast (errors)

    1. Measuring and comparing forecast errors usingMAD,MAPE, MSE, RMSE

    2. Controlling Forecasting Process via Tracking Signal

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    2/20

    2

    A

    B(4) C(2)

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

    Dependent Demand:

    Raw Materials,Component parts,Sub-assemblies, etc.

    Independent Demand:Finished Goods

    Demand Management

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    3/20

    3

    DemandManagement

    Active role to influence demand. Examples from seasonal goods or services

    Campaigns, discounts, etc.

    Incentives to sales personnel

    Passive role, limited or no action taken. Why?

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    4/20

    4

    Qualitative Methods

    Grass RootsMarket Research

    Panel Consensus

    Executive Judgment

    Historical analogy

    Delphi Method

    Qualitative

    Methods

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    5/20

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    6/20

    6

    Quantitative forecasting methods

    Time series models

    Past predicts future

    Uses time series data

    Key variable: time (t)

    Easier to apply

    Less accurate Examples:

    Moving averages

    Exponential smoothing

    Causal models

    Examines potential cause=> effectrelationships

    Requires cross sectional data

    Key variables are usually denotedas X

    1, X

    2, X

    3,

    More difficult

    Takes more time

    But worth it since it providesinsight to the system (process)

    under study Examples:

    Various regression models

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    7/207

    Quantitative techniques

    Basic time series approaches i. Moving averages, simple &weighted

    ii. Exponential smoothing, simple & trend adjusted

    iii. Linear regression (linear trend model)

    iv. Techniques for seasonality and trend -

    Decomposition of time series

    Causal approach i. Simple Linear Regression

    ii. Multiple Linear Regression

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    8/208

    Finding Components of a Time Series

    1 2 3 4

    x

    x xx

    xx

    x xx

    xx x x x

    xxxxxx x x

    xx

    x x xx

    x

    xx

    x

    x

    xx

    xx

    xx

    x

    xx

    xx

    x

    x

    x

    Year

    Sales

    Seasonal variation

    Linear

    Trend

    15-8

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    9/209

    What to look for in a time series

    Trend - long-term movement in data

    Seasonality - short-term regular and repetitive variations in data

    Cyclical variations long(er) term, occasionally caused by

    unusual circumstances, (war, economic downturn, etc.)

    Autocorrelation denotes persistence of occurrence (momentumdriven)

    Random variations - caused by chance

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    10/20

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    11/2011

    Exponential Smoothing Models

    Simple exponential smoothingmodel

    Alpha is the smoothing constant

    Whenever appropriate more weightcan be given to the more recentdata (time periods)

    Double exponential smoothing(Holts model)

    Adds trend component,

    Tand delta(gamma) as the smoothing constant

    for trend

    Forecast Including Trend (FIT)

    Excel time!

    Problem 20, continued.

    )FA(FF 111 tttt

    10where

    )FITF(TT 11 tttt

    )FITA(FITF 111 tttt

    ttt TFFIT

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    12/2012

    Measuring Accuracy, Forecast Errors

    To compare different time series techniques or to select thebest set of initial values for the parameters, use a combinationof the the following four metrics:

    Mean Absolute Deviation Most popular but

    Mean Absolut Percent Error Should be used in tandem with MAD

    Mean Square Error

    Root Mean Square Error

    n

    FA

    =MAD1

    n

    i

    ii

    n

    i i

    ii

    n 1 A

    FA100=MAPE

    n

    FA

    =MSE 1

    2

    n

    i

    ii

    MSERMSE

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    13/2013

    Tracking Signal

    The Tracking Signal or TS is a measure that indicateswhether the forecast average is keeping pace with anygenuine upward or downward changes in demand.

    Depending on the number of MADs selected, the TS can beused like a quality control chart indicating when the modelis generating too much error in its forecasts.

    TS is a monitoring system.

    The TS formula is:

    DeviationAbsoluteMean

    ErrorsForecastofSumRunning=TS

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    14/2014

    Regression analysis

    Identify factors (independent variables) that can be used to

    predict the values for the forecast variable (e.g., sales).

    Regression applied to causal data requires different kinds ofdata

    Regression applied to time series data is also know astrend line analysis

    We will use Excel (Tools/Data analysis) to obtain theregression line and all relevant statistics.

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    15/2015

    A simple regression example

    The first example applies regression to time series data.

    Whenever possible, plot and observe the data.

    The scatter plot shows a linear relation between advertising andsales. So the following regression model is suggested by the data,which refers to the true relationship between the entire population ofadvertising and sales values.

    Other common formats are:

    ii 110i XY

    tbaYXbaY

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    16/2016

    Decomposition of a Time Series

    Demand has both trend and seasonal components.

    View data via Excel.

    1. Compute overall average

    2. Compute average of the same seasons of each cycle (e.g., year)

    3. Compute seasonal indexes (seasonal averages / overall avg.)

    4. Deseasonalize data (actual values /seasonal indexes)

    5. Apply regression to deseasonalized data6. Compute (project) deseasonalized forecasts using the regressionequation

    7. Reseasonalize the forecasts by multiplying them with theseasonal indexes.

    Excel time.

    Problem 21.

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    17/2017

    Multiple regression

    Most regression problems involve more than one independent

    variable. If each independent variables varies in a linear manner with Y, the

    estimated regression function in this case is:

    Where b0 is the intercept (also called constant)

    The optimal values for the bi (slopes) can again be found using theleast squares method

    kkbbbb XXXY

    22110

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    18/2018

    Steps in multiple regression analysis

    1. Hypotheses for testing whether a general linear model is useful ispredicting Y:

    1. Ho :1 = 2 = 3 = ... = k = 0 (means there is NOTHING useful)

    2. HA : At least one of the parameters in Ho is nonzero.

    2. Test statistic: F-statistic = MSR / MSE

    3. If the model is deemed adequate (passes the F-test; rejected H0 )

    then go to step 4 (otherwise, none of variables have any impact on Y )4. Conduct t-tests (significance tests) on parameters (slopes).

    5. Remove the most insignificant independentvariable, re-run theregression, and go to step 4.

    6. Repeat steps 4 & 5 until all remaining independent variable

    parameters (slopes) are significant, then go to step 77. If the intercept (0 ) is insignificant then remove it, run regression one

    more time.

    Excel time!

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    19/2019

    What Forecasters Should Do

    Determine what elements of historical data provide repeatable

    patterns and utilize this to make extrapolations. Make a list of the possible independent variables that may have

    influenced the historical data and may influence futureoutcomes.

    Statistically correlate the independent variables to the outcome

    history using regression analysis to validate their importanceand to calibrate their effects.

    Make estimates of forecast error wherever possible using MADor standard deviation measures.

    Make clear presentations of the results and assumptions andlisten to feedback.

  • 7/31/2019 Ch 15 Demand Manmagement & Forecasting-HK

    20/2020

    Forecasting

    Always remember that you (managers) are decision makers and

    sound decisions are based on good forecasts

    Suggested problems:

    2, 3, 4, 7, 11, 12, 14, 17, 20, 21, 27