Introduction To Demand Planning & Forecasting

CTL.SC1x -Supply Chain & Logistics FundamentalsIntroduction toDemand Planning& ForecastingMIT Center forTransportation & Logistics

Demand Process – Three Key QuestionsDemand Planningn What should we do to shape andcreate demand for our product?n n n What should we expect demand tobe given the demand plan in place?Product & PackagingPromotionsPricingPlaceDemand Forecastingn n n Strategic, Tactical, OperationalConsiders internal & external factorsBaseline, unbiased, & unconstrainedDemand ManagementHow do we prepare for and acton demand when it materializes?n n n Balances demand & supplySales & Operations Planning (S&OP)Bridges both sides of a firmMaterial adapted from Lapide, L. (2006) Course Notes, ESD.260 Logistics Systems.CTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics2

Forecasting LevelsLevelStrategicHorizon Business PlanningCapacity PlanningInvestment Strategies Brand PlansBudgetingSales PlanningManpower PlanningMonths/Weeks Short-term Capacity PlanningMaster PlanningInventory PlanningDays/Hours Transportation PlanningProduction PlanningInventory rposesMaterial adapted from Lapide, L. (2006) Course Notes, ESD.260 Logistics Systems.CTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics3

Agenda Forecasting TruismsSubjective vs. Objective ApproachesForecast QualityForecasting MetricsCTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics

Forecasting Truisms 1:Forecasts are always wrongCTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics5

1. Forecasts are always wrongWhy?n n n Demand is essentially a continuous variableEvery estimate has an “error band”Forecasts are highly disaggregatedw Typically SKU-Location-Time forecastsn Things happen . . .OK, so what can we do?n n n n Don’t fixate on the point valueUse range forecastsCapture error of forecastsUse buffer capacity or stockCTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics6

Forecasting Truisms 2:Aggregated forecastsare more accurateCTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics7

2. Aggregated forecasts are more accurate Aggregation by SKU, Time, Location, etc. Coefficient of Variation (CV)Definition: Standard Deviation / Mean σ/µProvides a relative measure of volatility or uncertaintyCV is non-negative and higher CV indicates higher volatilityn n n Red: µ 100, σ 45, CV 0.45Blue: µ 100, σ 1, CV 0.01200180Daily Demand160140120100806040202/26/113/28/11CTL.SC1x - Supply Chain and Logistics Fundamentals4/27/115/27/116/26/11Lesson: Demand Forecasting Basics7/26/118/25/118

Aggregating by SKU Coffee Cups and Lids @ the Sandwich Shopn n n LargeMediumSmall N(80, 30) N(450, 210) N(250, 110)CV 0.38CV 0.47CV 39/13/1310/13/1311/12/13CTL.SC1x - Supply Chain and Logistics Fundamentals12/12/131/11/142/10/143/12/14Lesson: Demand Forecasting Basics4/11/145/11/146/10/149

Aggregating by SKU What if I design cups with a common lid? Common Lid N(780, 239) CV 0.31n n µ (80 450 250) 780 units/dayσ sqrt(302 2102 1102) 239 units/dayLarge N(80, 30) CV 0.38Med. N(450, 210) CV 0.47Small N(250, 110) CV 0.44Lids N(780, 239) CV 145/11/146/10/14Example of Modularity or Parts Commonality Reduces the relative variability Increases forecasting accuracy Lowers safety stock requirementsCTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics10

Aggregating by TimeDaily Demand for Lids N(780, 239) CV 0.311,600Forecasts with longer timebuckets have betterforecast accuracy.The time bucket usedshould match the situation. /11/142/10/143/12/144/11/145/11/146/10/14Weekly Demand for Lids N(5458, 632) CV Monthly Demand for Lids N(21840, 1264) CV x - Supply Chain and Logistics Fundamentals5678Lesson: Demand Forecasting Basics111211

Aggregating by Locations Suppose we have three sandwich shopsn CV reduces as weaggregate over SKUs,time, or locations.Weekly lid demand at each N(5458, 632) CV 0.12 N(5458, 632) N(5458, 632) N(5458, 632) N(16374, 1095) What if demand is pooled at a common Distribution Center?n Weekly lid demand at DC N(16374, 1095) CV 0.07σCVind µCTL.SC1x - Supply Chain and Logistics FundamentalsCVaggCVindσ nσ µn µ nnLesson: Demand Forecasting Basics12

Forecasting Truisms 3:Shorter horizon forecastsare more accurateCTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics13

3. Shorter horizon forecasts aremore accurate?212223241CTL.SC1x - Supply Chain and Logistics Fundamentals2345678Lesson: Demand Forecasting Basics910111214

3. Shorter horizon forecasts aremore accurate Postponed final customization tocloser time of consumptionRisk pooling of component (e.g.,ham) increases forecast accuracy.?212223241CTL.SC1x - Supply Chain and Logistics Fundamentals2345678Lesson: Demand Forecasting Basics910111215

Forecasting Truisms Forecasts are always wrongUse ranges & track forecast error Aggregated forecasts are more accurateè Risk pooling reduces CV Shorter time horizon forecasts are moreaccurateè Postpone customization until aslate as possibleè CTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics16

Subjective & Objective ApproachesCTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics17

Fundamental Forecasting ApproachesSubjectiveObjectiveJudgmentaln n n Causal / RelationalSales force surveysJury of expertsDelphi techniquesExperimentaln n n n n n Econometric ModelsLeading IndicatorsInput-Output ModelsTime SeriesCustomer surveysFocus group sessionsTest marketingn n n “Black Box” ApproachPast predicts the futureIdentify patternsOften times, you will need to use a combination of approachesCTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics

Forecasting QualityCTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics19

Cost of Forecasting vs Inaccuracyß Overly Naïve Models à ß Good Region à ß Excessive Causal Models à CostTotal CostCost of ErrorsIn ForecastCost of ForecastingForecast AccuracyCTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics

How do we determine if a forecast is good? What metrics should we use? Example - Which is a better forecast?n n Squares & triangles are different forecastsCircles are actual values11001000900timeCTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics

Accuracy versus Biasn n Accuracy - Closeness to actual observationsBias - Persistent tendency to over or under predictAccurateNot AccurateBiasedCTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting BasicsNot Biased

Forecasting MetricsCTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics23

Forecasting MetricsMean Deviation(MD)Mean AbsoluteDeviation (MAD)n etMD Mean SquaredError (MSE)Mean PercentError (MPE)n2tnnet AMPE t 1 tnMean AbsolutePercent Error (MAPE)MAPE At Actual value for obs. tFt Forecasted value for obs. tCTL.SC1x - Supply Chain and Logistics Fundamentalst 1nnRMSE t 1 etRoot MeanSquaredError (RMSE) eNotation:nMAD t 1nMSE et At – Ft2e tt 1netn At 1tnet Error for observation tn Number of observationsLesson: Demand Forecasting Basics

Example: Forecasting Bagels For the bagel forecast and actual valuesshown below, find the:n n n Mean Absolute Deviation (MAD)Root Mean Square of Error (RMSE)Mean Absolute Percent Error 5066Thursday5038Friday7586CTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basicsn etMAD t 1nnRMSE 2e tt 1netnMAPE At 1tn25

Example: Forecasting Bagels Solution:901. Graph it.2. Extend data table:w Error: et At-Ftw Abs[error] et w Sqr[error] e2w AbsPct[error] et/At 3. Sum and find meansForecastDaily Bagel Demand8070605040FtAtet et ay5066161625624.2%Thursday5038 -121214431.6%Friday75 et/At 86 11 11 121 12.8%Sum 054 634104%Mean 0 10.8 126.8 21%CTL.SC1x - Supply Chain and Logistics FundamentalsActual30MondayTuesdayWednesday ThursdayFridayMAD 54/5 10.8RMSE sqrt(126.8) 11.3MAPE 104%/5 21%Lesson: Demand Forecasting Basics26

Key Points from LessonCTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics27

Key Points Forecasting is a means not an end Forecasting Truismsn n n Forecasts are always wrongAggregated forecasts are more accurateShorter horizon forecasts are more accurate Subjective & Objective Approachesn n Judgmental & experimentalCausal & time series Forecasting metricsn n Capture both bias & accuracyMAD, RMSE, MAPECTL.SC1x - Supply Chain and Logistics FundamentalsLesson: Demand Forecasting Basics

CTL.SC1x -Supply Chain & Logistics FundamentalsQuestions, Comments, Suggestions?Use the Discussion!“Janie”Photo courtesy Yankee GoldenRetriever Rescue (www.ygrr.org)MIT Center forTransportation & Logisticscaplice@mit.edu

Forecasting is a means not an end Forecasting Truisms ! Forecasts are always wrong ! Aggregated forecasts are more accurate ! Shorter horizon forecasts are more accurate Subjective & Objective Approaches ! Judgmental & experimental ! Causal & time series Forecasting metrics ! Capture both bias & accuracy !