Production Management Demand Management

Production Management Demand Management

UNIVERSITY X

Introduction

lDemand Management is the process of forecasting, influencing, or controlling demand for a product or service with the objective of satisfying that demand through effective direction, utilization and integration with the supply process

Types of Demand

lGenerally, two types of demand are defined

lIndependent demand – Demand is not derived directly from that of other products

lDependent demand – Demand is derived from that of other products

Forecasting

lForecasting strives to predict future demand through the use of processes, techniques, past knowledge, related variables, etc

lForecasts and Targets

lForecasts and Sales Plans

Types of Forecasting

lForecasting may be done in any of the following ways

–    Qualitative

–    Time Series Analysis

–    Causal Relationships

Qualitative Forecasting

lThis method is usually subjective and based on opinions, intuition, etc

lFactors influencing forecast may include

–    Experience of market

–    General intuitive evaluation of variables assumed to be correlated

–    Personal opinion or judgment

lUsed in cases such as new product demand forecasting or very long term forecasting

Time Series Forecasting

lUsually the most commonly deployed method of forecasting

lPredict future based on past data

lMay be adjusted for seasonal effects, outliers, even causal relationships

Time Series Forecasting

lCommon techniques used include

–    Simple Moving Average

–    Weighted Moving Average

–    Exponential Smoothing

Simple Moving Average

lAverage of pre-determined number of past period data is taken as the forecast for the next period

lTends to smooth out abnormal periods of demand making smoother operations possible

lIs slow in reflecting trends

lNumber of periods chosen may impact accuracy of forecast

Weighted Moving Average

lAverage of pre-determined number of past period data with weights according to relative importance is taken as the forecast for the next period

lMakes it possible to concentrate on particular periods where data is more reflective of actual demand patterns

lCan be used to eliminate outliers

Weighted Moving Average

lChoosing Weights

–    Depends on the particular product or industry

–    Recent periods usually are weighted more

–    Seasonal trends may be weighted higher to reflect true demand

Exponential Smoothing

lIf the market is such that recent data is usually more indicative than distant data, exponential smoothing may be used to forecast future demand

–    A smoothing constant ? is used depending on relative importance of recent trends

–    The forecast for the last period, the actual demand for the last period and the smoothing constant are used to predict the next period’s demand

F_t = F_t-1 + ? (A_t-1 – F_t-1)

Exponential Smoothing

lChoosing the appropriate value for ? determines the accuracy of the forecast

–    If the demand is stable, then a small ? is used to lessen the effects of random or short term change

–    If the demand is highly variable, then a large ? is used to ensure the most recent trend is followed

–    In some models, the value of ? is changed based on the magnitude of the forecast error

Seasonality

lSimilar patterns or trends observed in past demand data when compared over fixed time periods may indicate the existence of seasonality

Linear Regression

lUsed in causal relationship based forecasting or even time series based forecasting where time is considered the predictive variable

–    Before using, it should be checked whether past data points appear linear when plotted

–    The regression line is of the form y = a + bx

where,              y = value of dependent variable

a = intercept with vertical or Y axis

b = slope

x = predictive variable

Linear Regression

lApplications may be in the following areas

–    Long range forecasting

–    Product family forecasting

Least Squares Method

lThe least squares method tries to fit the line to the data such that the sum of the squares of the vertical distance between each observed data point and its predicted point on the line is minimized

lThe equation for linear regression is,

Y = a + bx

However, in this case Y is the computed dependent variable

Linear Regression

lCausal relationship based forecasting

Questions