1. Forecasting Demand

2. Problems with Forecasts Forecasts are Usually Wrong. Every Forecast Should Include an Estimate of Error. Forecasts are More Accurate with Sample or Groups of Items. Forecasts are More Accurate for Nearer Time Periods.

3. Types of Demand Dependent (Derived demand – MRP – No Need to Forecast) Independent (Not Directly Related to Demand for Another Item – Must be Forecast)

4. Forecasting Methods Qualitative – Judgmental, Executive Opinion - Internal Opinions - Delphi Method - Surveys Quantitative - Extrinsic or Causal - Intrinsic or Time Series

5. Extrinsic Methods Seek Relation between Sales and Economic Indicators (Especially Leading Indicators) Example: Door Lock Demand & Housing Starts MonthHousing StartsDoor Locks January4000350 February5000450 March3000300 April6000550 May7000720

6. Scatter Diagram – Door Lock Sales vs. Housing Starts

7. Intrinsic or Time Series Forecasts Based on Past Demand Patterns or Components: Average or Level Trend Seasonal Cyclical (Omit) Random (Cannot Forecast)

8. Intrinsic - Level Demand Simple or Arithmetic Mean E.g. F 5 = (103 + 121 + 130 + 150) / 4 = 126 Moving Average – Discard Old Data Weighted Average F t+1 =  t D t +  t-1 D t-1 + Etc.  = Weight between 0 and 1,   i  D = Actual Demand t = Current Time Period (t=4) E.g. F 5 =.4(150)+.3(130)+.2(121)+.1(103) = 133.5

9. Intrinsic - Level Demand Exponential Smoothing Weighted Average New Forecast =  (Latest Demand) + (1-  ) (Previous Forecast) F t+1 =  D t + (1-  )F t F t is Old Forecast from Last Period E.g. F 5 = (.2)(150) + (.8)(115) = 122

10. Intrinsic - Trends Trend is Predictable Long Term Increase or Decrease in Demand E.g.January103 February121 March130 April 150 If Trend Continues, Averages are Too Low Forecasting Techniques: - Regression (Least Squares) - Adjusted Exponential Smoothing

11. Scatter Diagram of Demand vs. Month Number

12. Intrinsic - Trends Simple Regression: One Independent Variable E.g. F t = a + bt (t is Time, a & b are Constants) F 5 = 88.5 + (15)(5) = 163.5 Multiple Regression: Multiple Independent Variables E.g. F t = a + b 1 t + b 2 i (i is base index) F 5 = 81 + (12.83)(5) + (16.67)(1.05) = 162.6 We Can Use Excel to Get a & b’s

13. Intrinsic - Trends & Exponential Smoothing 1.F t+1 =  D t + (1-  )F t = 122 2.Trend Factor = (F t+1 – F t ) = 122 - 115 = 7 T t+1 =  (F t+1 – F t ) + (1-  ) T t  = Weight between 0 and 1, Often =  T t = Old Trend (Or Use Trend Line Slope) E.g. T t+1 =.2(7) +.8(15) = 13.4

14. Intrinsic - Trends & Exponential Smoothing 1.F t+1 = 122 2.T t+1 =.2(7) +.8(15) = 13.4 3.A F t+1 = F t+1 + (Lag)(T t+1 ) Lag Can be (1/  ) = (1/.2) = 5 E.g. A F t+1 = 122 + (5)(13.4) =189 Can You Do a Forecast for June?

15. Intrinsic: Seasonal Demand Seasonal Demand: Definite, Dependable Reason for Heavy Demand at One Time, Light Demand at Another 1.Construct Base Series or Index from Historical Demand (Period Average Demand / Average Demand for All Periods) E.g. Base Index for April = 110 % 2.Divide All Demand by Appropriate Base 3.Forecast Using Any Method 4.Adjust Forecast by Multiplying by Appropriate Base

16. Evaluating Forecasts - MAD MAD is Mean Absolute Deviation Smaller the MAD, the Better MAD =  | D t – F t | / n D t = Actual Demand F t = Forecast n= Number of Periods

17. Evaluating Forecasts - MAD Example of MAD for May and June: MonthD t F t | D t – F t | May17212250 June19213260 110 MAD = 110 / 2 = 55

18. Evaluating Forecasts - Tracking Signal (TS) Tracking Signal (TS) Monitors Quality of Forecast Tracking Signal > ± 4 Means Forecast Should be Reviewed TS = Algebraic Sum of Forecast Errors / MAD

19. Evaluating Forecasts - Tracking Signal (TS) Example of TS for July and August: Month D t F t | D t – F t | D t – F t July190 180 1010 August180 184 4 -4 14 6 MAD= 14 / 2 = 7 TS = 6/7