Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

Embed Size (px)

Citation preview

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    1/38

    Forecasting

    N.K.Agarwal

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    2/38

    All supply chain decisions based on estimates offuture demands

    Historical demand information can be used to

    forecast future demands For push/pull philosophy of supply chain

    Push processes are performed in anticipation of demand

    Pull processes performed in response to the customerdemand

    Dell orders components for computers in anticipation ofcustomer demand, while

    Assembly is performed in response to a customerdemand

    Forecasting

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    3/38

    When individual stages in the supply chain maketheir independent forecast of demand, there is alwaysa mismatch between the supply and demand

    Collaborative forecast for the entire chain partnerstends to be much more accurate

    Decisions for functions like Production, Marketing,Finance, Personnel are best taken based oncollaborative forecast

    Mature products with stable demand are usuallyeasiest to forecast Staple products like food grains, sugar at superbazars

    Forecasting

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    4/38

    Forecasting and accompanying managerial decisionsare extremely difficult when either the supply of rawmaterials or the demand for the finished product is

    highly variable Fashion garments, high tech products etc.

    Good forecasting is important for products with shortlife cycle, like fashion goods

    Products with a long life cycle have less significanteffect from forecasting errors

    Forecasting

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    5/38

    Forecasts are always wrong and should include boththe expected value and a measure of the forecasterror

    Long term forecasts are usually less accurate thanshort term forecasts

    The greater the degree of aggregation , the moreaccurate is the forecast Easier to forecast the GNP in a year of a country within 2%

    accuracy than the annual revenue of a company

    The greater up the supply chain a company is, thegreater the distortion of information they receive Bullwhip effect

    Forecasting- Characteristics

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    6/38

    Bullwhip Effect

    Tier 2

    Suppliers

    Tier 1

    SuppliersProducer Distributor Customers

    Ordering

    Amount ofinventory=

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    7/38

    Companies need to first Identify the factors that influence the future demand, and

    then

    Ascertain the relationship between these factors and futuredemand

    Some of the factors that need to be looked into Past demand

    Lead time of products

    Planned advertising or marketing efforts State of economy

    Planned price discounts

    Action competitors have taken

    Forecasting- Components

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    8/38

    Demand Forecasting -

    Basic Forecasting Six step approach for effective forecasting

    Understand the objective of forecasting

    Integrate demand planning and forecasting throughout the

    supply chain

    Understand and identify customer segments

    Identify the major factors that influence the demand forecast

    Determine the appropriate forecasting technique

    Establish performance and error measures for the forecast

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    9/38

    Qualitative Method Qualitative forecasting methods are primarily subjective and

    rely on human judgment

    Most appropriate when there is little historical dataavailable or when experts have market intelligence that iscritical in making forecast

    Used to forecast future demand for long term in a newindustry

    Time Series Use historical demand to forecast Method appropriate when the demand pattern does not

    vary significantly from one year to the next

    Forecasting- Methods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    10/38

    Causal Method assumes that the demand forecast is highlycorrelated with certain factors in the environment

    State of economy, interest rates etc.

    Used to determine the impact of price promotions ondemand

    Simulation Methods imitate the consumer choices that give rise to

    demand to arrive at a forecast

    Simulation is used to combine time series and causal

    methods to find answers to Impact of price promotion, competitors stores coming upin the vicinity etc.

    Forecast demand for higher fare seats when thereare no seats available at economy class fare

    Modeling makes use of computers

    Forecasting- Methods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    11/38

    Time Series Forecasting Methods

    A time series is a time-ordered sequence ofobservations taken at regular intervals over a periodof time

    Data may be measurement of demand, earnings, profits,outputs etc.

    Analysis of time series data requires identification ofthe underlying behaviour of the series

    Done by plotting the data with time and examining for somepattern

    Trend, Seasonal variations, Cycles, and Random orIrregular variations ( errors)

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    12/38

    Trend Refers to gradual, long term, upward or downward

    movement in the data over time

    Changes in income, population etc.

    Seasonality Refers to short term fairly regular variations related to factors

    such as weather, holidays, vacations etc.

    Variations can be daily, weekly or monthly

    Cycles Wave like variations of more than one years duration or

    which occur every year

    Business cycle related to economic, political oragricultural conditions

    Random variations Residual variations which are blips in the data caused by

    chance and unusual situations

    Time Series Forecasting Methods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    13/38

    Constant Trend

    Seasonal Trend

    Demand Patterns

    Time

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    14/38

    Pattern continuous when it is constant and does notconsistently increase or decrease

    Sales of a product in the mature stage of its life cycle

    may show this Linear pattern emerges when demand increases or

    decreases from one period to the next

    Sales of product in the growth stage of the product lifecycle shows increasing while in the decline stage show

    decreasing trend Cyclical pattern pertains to influence of seasonal factors

    Demand of woolen wears will be high in winter and lowduring summer

    Quantitative Methods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    15/38

    Forecasts in time series methods based on averagessmoothened through averaging

    Three techniques used for Averaging

    Naive Forecasts

    Simplest method

    Assumption of demand for the next period based on theactual demand in the most recent period

    Moving Average method Simple moving average

    Weighted moving average

    Time Series Forecasting Methods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    16/38

    Simple Moving Average (SMA) Forecasts for the next month is the arithmetic average of the

    actual sales for a specific number of recent past time periods

    SMA =Sum of demands for all periods/Chosen number ofperiods

    SMA = in

    =1/n =(D1+D2+D3Dn)/n,

    where , n=the chosen number of periods,

    i= 1 is the oldest period in the n-period average

    i= n is the most recent period

    D1= the demand in the i th period

    Time Series Forecasting Methods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    17/38

    Weighted Moving Average (WMA)

    A weighted average of past sales is the forecast for the nexttime period

    A WMA allows for varying, not equal weightage of olddemands

    WMA= in

    =1 Ci Di ,

    where Di is the demand during time period i, Ci is theweight given to that demand and n is the chosen

    number of periods

    Also 0 Ci 1 , and in

    =1 Ci =1

    Time Series Forecasting Methods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    18/38

    Exponential Smoothing Models Forecasted sales for the last period modified by information

    about the forecast error of the last periods

    Modification of the last years forecasts are the forecast forthe next time periods

    Weight assigned to a previous periods demanddecreases exponentially as that data gets older

    Recent demand data receive a higher weight than doesthe older demand data

    Normally only three items of data are required

    This periods forecast, the actual demand for this periodand which is referred to as smoothening constant andhaving a value between 0 and 1

    Time Series Forecasting Methods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    19/38

    Formula used is

    Next periods forecast = This period's forecast + ( thisperiods actual demand this periods forecast)

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

    Where Ft = Forecast for this period (t)

    Ft-1 = Forecast for the previous period (t-1)

    At-1

    = Actual demand for the previous period ( t-1)

    = Smoothening constant

    Smoothening constant selection is a matter of judgment

    Commonly used values range between 0.05 0.5

    Time Series Forecasting Methods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    20/38

    Regression Analysis A forecasting technique that establishes a relationship

    between variables- one dependent and others independent

    Only one independent variable in simple regression

    Population, advertising expenses affecting sales More than one independent variable in multiple regression

    Population, income and sales force affecting sales

    It involves fitting a straight line equation ( in simple linearregression analysis) to explain sales fluctuations in terms ofrelated and presumable causal variables

    Three major steps in regression analysis

    Identifying variables which are causally related to thefirms sales

    Determine / estimate the values of these variablesrelated to sales

    Derive the sales forecast from these estimates

    Common Time Series Models

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    21/38

    A linear regression assumes the relationship betweendependent and independent variables a straight line( known a simple linear regression analysis)

    A curvilinear relationship is a non-linear regressionproducing a curve

    Common Time Series Models

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    22/38

    Forecasting- Adaptive Method

    Adaptive method uses more sophisticated approachcompared to static methods

    Popular models used in this method

    Holts Model

    This is a Trend corrected Exponential smoothened model

    Appropriate when demand is assumed to have a level and atrend but no seasonality

    Systematic component of demand = Level + Trend In period t, given estimate of level Lt and trend Tt, the

    forecast for future periods is expressed as

    Ft+1 = Lt + Tt and Ft+n = Lt+ nTt

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    23/38

    Forecasting- Adaptive Method

    After observing for Period t, the estimate for level and trendis corrected as

    Lt+1 = Dt+1 + (1- )(Lt + Tt)

    Tt+1 = (Lt+1 Lt) + (1- )Tt , Where is a smoothening constant for the level, and is

    a smoothening constant for trend and varies from 0 to 1like

    Winters Model

    Trend and Seasonality Corrected Exponential Smoothenedmodel

    Method appropriate when the demand is assumed to have alevel, trend and a seasonal factor

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    24/38

    Systematic component of the demand= ( Level + trend) x seasonal factor

    Assume periodicity of demand to be p, initial estimates oflevel L0, trend T0 and seasonal factors ( S1.Sn)

    In period t, the forecast for future periods is given by

    Ft+1 = (Lt + Tt)* St+1, and Ft+l = (Lt + lTt)*St+l On observing the demand for period t+1, the estimates for

    level, trend and seasonal factors are revised as

    Lt+1 = (Dt+1 /St+1) + (1- )(Lt-Tt)

    Tt+1 = (Lt+1 Lt) + (1- )*Tt St+p+1 = (Dt+1 / Lt+1) + (1- )*St+1,

    Where is a smoothening constant for seasonal factorvarying from 0 -1

    Forecasting- Adaptive Method

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    25/38

    Measure of Forecasting Errors

    Managers perform a thorough error analysis on aforecast to

    Determine whether the current forecasting method is

    accurately predicting the systematic components of demand A method consistently giving positive error can indicate

    over prediction by the method and manager can makenecessary corrections

    Estimate forecast error as any contingency plan must

    account for such an error

    Contracting with an outsource agency , even thoughmore expensive, to supply shortfalls in the order onurgent basis

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    26/38

    Measure of Forecasting Errors

    Forecasting Error is simply the difference betweenthe forecast and actual demand for a given period et = Ft At , where et = forecast error for the period t,

    At = actual demand for period t, and Ft = the forecast for theperiod t

    Mean Error (ME) = 1/n n

    t=1 et Cumulative Sum or Error (CFE) = nt=1 et CFE is useful in measuring the bias in a forecast

    Mean Absolute Deviation (MAD) = 1/n nt=1 | et | MAD is merely the average error for each forecast.

    Popular because it is easy to understand

    Mean Squared Error (MSE) = 1/n nt=1 et2

    Used as an estimate of the variance of the random error etwhich is 2

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    27/38

    Mean Absolute Percentage Error (MAPE)

    =1/n t=1 ( |et| / At) X100

    MAPE is useful for putting forecast performance in the

    proper perspective Forecast error of 100 when the actual demand is 200

    units results in larger percentage error than the erroroccurring when the demand was 1000 units

    Measure of Forecasting Errors

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    28/38

    Qualitative or Judgemental

    Methods Not based on quantitative numbers exclusively

    Based on judgment about the causal factors that underlinethe sales of particular products or services, and

    On opinions about the relative likelihood of these causalfactors being present in the future

    Useful when historical data are not available

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    29/38

    Executive Committee Consensus A committee of executives from different departments

    constituted and entrusted with the responsibility ofdeveloping a forecast

    Uses inputs from all parts of organisation and analystsanalyse data as required

    Such forecasts tend to be compromised ones, not reflectingthe extremes that might be present

    Most commonly used method of forecast

    Qualitative or Judgemental

    Methods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    30/38

    The Delphi Method Method seeks to remove the undesirable consequences of

    group thinking existing in committees

    Committee consists of experts from within and outside the

    organisation Expert in one aspect of the problem and no one

    conversant with all aspects of the issue

    Each expert makes independent predictions in the form ofbrief statements

    Coordinator edits and clarifies these statements Coordinator provides a series of questions in writing to the

    experts that includes feedback supplied by other experts

    Above repeated several times till consensus reached

    Qualitative or JudgementalMethods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    31/38

    Survey of Salesforce/ Field Expectation Method Individual members of the salesforce required to submit

    sales forecasts of their respective regions

    These combined to form total estimate of sales Estimates transformed into sales forecasts to ensure realistic

    estimates

    A popular method for companies having goodcommunication system and salesforce directly selling tocustomers

    Qualitative or JudgementalMethods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    32/38

    Survey of Customers/Users Expectation Method

    Estimates of future sales obtained directly from customersthrough survey

    Sales forecast determined by combining individualcustomers responses

    Method useful where customers are limited in number

    Qualitative or Judgemental

    Methods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    33/38

    Historical Analogy

    Estimates of future sales of product tied to knowledge of asimilar products sales

    Knowledge of one products sales during various stages ofits product life cycle applied to estimates of sale for a similarproduct

    Method useful for a new product

    Qualitative or Judgemental

    Methods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    34/38

    Market Surveys

    Questionnaires, telephone talks or field interviews form thebasis for predicting market demand for products

    Normally preferred for new products or existing products innew markets

    Qualitative or Judgemental

    Methods

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    35/38

    Demand Forecasting

    Forecasting is a key driver of virtually every designand planning decision made in both an enterpriseand a supply chain

    Collaborative forecasting taking all partners in thesupply chain give benefits an order of magnitudehigher than the cost

    Value of data depends upon where one is in thesupply chain

    Demand is not the same as sales

    True demand can be obtained by making adjustments for theunmet demands due to stock outs, competitors actions,

    pricing, promotions etc.

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    36/38

    References

    Supply Chain Management : Chopra / Meindl

    Logistics and Supply Chain Management :K. Shridhara Bhat

    Supply Chain Management : Rahul V. Altekar

    Google web site

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    37/38

    Thank You

  • 8/2/2019 Lect 04 LSCM Sterl (R0-July 16,09) Forecasting

    38/38