POM - J. Galvn*PRODUCTION AND OPERATIONS MANAGEMENTCh. 5: Forecasting

POM - J. Galvn

POM - J. Galvn*Learning ObjectivesUnderstand techniques to foresee the future

POM - J. Galvn

POM - J. Galvn*What is Forecasting?Process of predicting a future eventUnderlying basis of all business decisionsProductionInventoryPersonnelFacilities

POM - J. Galvn

POM - J. Galvn*Short-range forecastUp to 1 year; usually < 3 monthsJob scheduling, worker assignmentsMedium-range forecast3 months to 3 yearsSales & production planning, budgetingLong-range forecast3+ yearsNew product planning, facility locationTypes of Forecasts by Time Horizon

POM - J. Galvn

POM - J. Galvn*Short-term vs. Longer-term ForecastingMedium/long range forecasts deal with more comprehensive issues and support management decisions regarding planning and products, plants and processes.Short-term forecasting usually employs different methodologies than longer-term forecastingShort-term forecasts tend to be more accurate than longer-term forecasts.

POM - J. Galvn

POM - J. Galvn*Influence of Product Life CycleStages of introduction & growth require longer forecasts than maturity and declineForecasts useful in projecting staffing levels, inventory levels, and factory capacity as product passes through stages

POM - J. Galvn

POM - J. Galvn*Types of ForecastsEconomic forecastsAddress business cyclee.g., inflation rate, money supply etc.Technological forecastsPredict technological changePredict new product salesDemand forecastsPredict existing product sales

POM - J. Galvn

POM - J. Galvn*Seven Steps in ForecastingDetermine the use of the forecastSelect the items to be forecastDetermine the time horizon of the forecastSelect the forecasting model(s)Gather the dataMake the forecastValidate and implement results

POM - J. Galvn

POM - J. Galvn*Realities of ForecastingForecasts are seldom perfectMost forecasting methods assume that there is some underlying stability in the systemBoth product family and aggregated product forecasts are more accurate than individual product forecasts

POM - J. Galvn

POM - J. Galvn*Forecasting ApproachesUsed when situation is stable & historical data existExisting productsCurrent technologyInvolves mathematical techniquese.g., forecasting sales of color televisionsQuantitative MethodsUsed when situation is vague & little data existNew productsNew technologyInvolves intuition, experiencee.g., forecasting sales on InternetQualitative Methods

POM - J. Galvn

POM - J. Galvn*Overview of Qualitative MethodsJury of executive opinionPool opinions of high-level executives, sometimes augment by statistical modelsSales force compositeestimates from individual salespersons are reviewed for reasonableness, then aggregatedDelphi methodPanel of experts, queried iterativelyConsumer Market SurveyAsk the customer

POM - J. Galvn

POM - J. Galvn*Involves small group of high-level managersGroup estimates demand by working togetherCombines managerial experience with statistical modelsRelatively quickGroup-think disadvantage 1995 Corel Corp.Jury of Executive Opinion

POM - J. Galvn

POM - J. Galvn*Sales Force CompositeEach salesperson projects their salesCombined at district & national levelsSales reps know customers wantsTends to be overly optimistic

POM - J. Galvn

POM - J. Galvn*Delphi MethodIterative group process3 types of peopleDecision makersStaffRespondentsReduces group-think(Sales?)(What will sales be? survey)(Sales will be 45, 50, 55)(Sales will be 50!)

POM - J. Galvn

POM - J. Galvn*Consumer Market SurveyAsk customers about purchasing plansWhat consumers say, and what they actually do are often differentSometimes difficult to answer

POM - J. Galvn

POM - J. Galvn*5-22Overview of Quantitative ApproachesNave approachMoving averagesExponential smoothingTrend projection

Linear regressionTime-series ModelsCausal models

POM - J. Galvn

POM - J. Galvn*Quantitative Forecasting Methods (Non-Naive)

POM - J. Galvn

POM - J. Galvn*Set of evenly spaced numerical dataObtained by observing response variable at regular time periodsForecast based only on past valuesAssumes that factors influencing past, present, & future will continue ExampleYear:19931994199519961997Sales:78.763.589.793.292.1What is a Time Series?

POM - J. Galvn

POM - J. Galvn*Time Series Components

POM - J. Galvn

POM - J. Galvn*Persistent, overall upward or downward patternDue to population, technology etc.Several years duration Trend Component

POM - J. Galvn

POM - J. Galvn*Repeating up & down movementsDue to interactions of factors influencing economyUsually 2-10 years duration Cyclical Component

POM - J. Galvn

POM - J. Galvn*Regular pattern of up & down fluctuationsDue to weather, customs etc.Occurs within 1 year Seasonal Component

POM - J. Galvn

POM - J. Galvn*Erratic, unsystematic, residual fluctuationsDue to random variation or unforeseen eventsUnion strikeTornadoShort duration & nonrepeating Random Component

POM - J. Galvn

POM - J. Galvn*Any observed value in a time series is the product (or sum) of time series componentsMultiplicative modelYi = Ti Si Ci Ri (if quarterly or mo. data)Additive modelYi = Ti + Si + Ci + Ri (if quarterly or mo. data)General Time Series Models

POM - J. Galvn

POM - J. Galvn*Naive ApproachAssumes demand in next period is the same as demand in most recent periode.g., If May sales were 48, then June sales will be 48Sometimes cost effective & efficient

POM - J. Galvn

POM - J. Galvn*MA is a series of arithmetic means Used if little or no trendUsed often for smoothingProvides overall impression of data over timeEquationMoving Average Method

POM - J. Galvn

POM - J. Galvn*YearSales02468939495969798ActualMoving Average Graph

POM - J. Galvn

POM - J. Galvn*Increasing n makes forecast less sensitive to changesDo not forecast trend wellRequire much historical data 1984-1994 T/Maker Co.Disadvantages of Moving Average Method

POM - J. Galvn

POM - J. Galvn*Used for forecasting linear trend lineAssumes relationship between response variable, Y, and time, X, is a linear function

Estimated by least squares methodMinimizes sum of squared errorsLinear Trend Projection

POM - J. Galvn

POM - J. Galvn*b > 0b < 0aaYTime, XLinear Trend Projection Model

POM - J. Galvn

POM - J. Galvn*TimeScatter Diagram

POM - J. Galvn

POM - J. Galvn*Least Squares EquationsSlope:

POM - J. Galvn

POM - J. Galvn*Multiplicative Seasonal ModelFind average historical demand for each season by summing the demand for that season in each year, and dividing by the number of years for which you have data.Compute the average demand over all seasons by dividing the total average annual demand by the number of seasons.Compute a seasonal index by dividing that seasons historical demand (from step 1) by the average demand over all seasons.Estimate next years total demandDivide this estimate of total demand by the number of seasons then multiply it by the seasonal index for that season. This provides the seasonal forecast.

POM - J. Galvn

POM - J. Galvn*YXii=+abShows linear relationship between dependent & explanatory variablesExample: Sales & advertising (not time)Dependent (response) variableIndependent (explanatory) variable Slope Y-intercept^Linear Regression Model

POM - J. Galvn

POM - J. Galvn*YXYai+^iiErrorLinear Regression Model

POM - J. Galvn

POM - J. Galvn*Linear Regression EquationsSlope:

POM - J. Galvn

POM - J. Galvn*Slope (b)Estimated Y changes by b for each 1 unit increase in XIf b = 2, then sales (Y) is expected to increase by 2 for each 1 unit increase in advertising (X)Y-intercept (a)Average value of Y when X = 0If a = 4, then average sales (Y) is expected to be 4 when advertising (X) is 0Interpretation of Coefficients

POM - J. Galvn

POM - J. Galvn*Answers: how strong is the linear relationship between the variables?Coefficient of correlation Sample correlation coefficient denoted rValues range from -1 to +1Measures degree of associationUsed mainly for understandingCorrelation

POM - J. Galvn

POM - J. Galvn*-1.0+1.00Perfect Positive CorrelationIncreasing degree of negative correlation-.5+.5Perfect Negative CorrelationNo CorrelationIncreasing degree of positive correlationCoefficient of Correlation Values

POM - J. Galvn

POM - J. Galvn*Coefficient of Correlation and Regression Model

POM - J. Galvn

POM - J. Galvn*You want to achieve:No pattern or direction in forecast errorError = (Yi - Yi) = (Actual - Forecast)Seen in plots of errors over timeSmallest forecast errorMean square error (MSE)Mean absolute deviation (MAD)^Guidelines for Selecting Forecasting Model

POM - J. Galvn

POM - J. Galvn*Pattern of Forecast Error

POM - J. Galvn

POM - J. Galvn*Measures how well forecast is predicting actual valuesRatio of running sum of forecast errors (RSFE) to mean absolute deviation (MAD)Good tracking signal has low valuesShould be within upper and lower control limitsTracking Signal

POM - J. Galvn

POM - J. Galvn*Tracking Signal Plot

POM - J. Galvn

POM - J. Galvn*Forecasting in the Service SectorPresents unusual challengesspecial need for short term recordsneeds differ greatly as function of industry and productissues of holidays and calendarunusual events

POM - J. Galvn

POM - J. Galvn*Forecasting example

SALES DURING LAST YEAR

LAST YEARReal salesSpring200Summer350Fall 300Winter150TOTAL ANNUAL SALES1000

ESTIMATION: Annual increase of sales10,00%

What are the estimated seasonal sales amount for next year?

POM - J. Galvn

POM - J. Galvn*Forecasting example (II)

LAST YEARPastsalesAverage sales for each seasonSeasonal factorTotal past annual sales/n of seasonsPast sales/Avg. SalesSpring2002500,8Summer3502501,4Fall 3002501,2Winter1502500,6TOTAL ANNUAL SALES10001000

POM - J. Galvn

POM - J. Galvn*Forecasting example (III)

NEXT YEARSALES1100(10% increase)NEXT YEARAverage sales for each seasonSeasonal factorNext year's seasonal forecastTotal estimated annual sales/n of seasonsAs calculatedAvg.sales*FactorSpring?2750,8220Summer?2751,4385Fall ?2751,2330Winter?2750,6165TOTAL ANNUAL SALES110011001100

POM - J. Galvn

*5*9*10*11*12*13*14*15*16*17*18*19*20*21*22*23*24*25*26*27*28*29*30*31*32*37*39*66*67*68*69*73*74*75*76*78*82*84*85*86*87*93*108*109