AN AUTOMATED SALES FORECASTING SYSTEM
Sales Forecasting Example 80000 60000 40000 20000 0 Jun-91 Jun-92 Jun-93 Jun-94 Jun-95 Jun-96 Date
Italian
Italian German
Italian German From Bavaria
Sales Forecasting Example 80000 60000 40000 20000 0 Jun-91 Jun-92 Jun-93 Jun-94 Jun-95 Jun-96 Date
Sales Forecasting Example 80000 60000 40000 20000 0 Jun-91 Jun-92 Jun-93 Jun-94 Jun-95 Jun-96 Date
REQUIREMENTS Regular weekly and monthly forecasts
REQUIREMENTS Regular weekly and monthly forecasts Input directly into production scheduling
REQUIREMENTS Regular weekly and monthly forecasts Input directly into production scheduling Automated
THERE IS NO FOOLPROOF WAY OF AUTOMATICALLY PRODUCING PERFECT FORECASTS
THE SAS SOLUTION SAS/ACCESS SAS/ETS PROC ARIMA PROC FORECAST PROC AUTOREG PROC EXPAND SAS/BASE SAS/AF SAS/EIS SAS/GRAPH
THE SAS SOLUTION SAS/ACCESS SAS/ETS PROC ARIMA PROC FORECAST PROC AUTOREG PROC EXPAND SAS/BASE SAS/AF SAS/EIS SAS/GRAPH CLIENT/SERVER ARCHITECTURE
SYSTEM STRUCTURE Batch job running under UNIX Reads sales Calculates and stores forecasts Produces some reports
SYSTEM STRUCTURE Batch job running under UNIX Reads sales Calculates and stores forecasts Produces some reports Online system running under Windows Viewing & editing forecasts Setting of parameters and options Flexible report generation Exporting the forecasts to other systems
WHY GO TO ALL THIS EFFORT?
WHY GO TO ALL THIS EFFORT? Reduced Inventory Levels
WHY GO TO ALL THIS EFFORT? Reduced Inventory Levels Reduced Out-of-Stock
WHY GO TO ALL THIS EFFORT? Reduced Inventory Levels Reduced Out-of-Stock Reduced Old/Obsolete Stock
WHY GO TO ALL THIS EFFORT? Reduced Inventory Levels Reduced Out-of-Stock Reduced Old/Obsolete Stock Reduced Distribution Costs
WHY GO TO ALL THIS EFFORT? Reduced Inventory Levels Reduced Out-of-Stock Reduced Old/Obsolete Stock Reduced Distribution Costs Improved Cash Flow Planning
WHY GO TO ALL THIS EFFORT? Reduced Inventory Levels Reduced Out-of-Stock Reduced Old/Obsolete Stock Reduced Distribution Costs Improved Cash Flow Planning Highlighting of Problem Sales Areas
METHODOLOGY
METHODOLOGY Bottom up
Sales Customer Example 250 200 150 100 50 0 Jan-92 Jan-93 Jan-94 Jan-95 Jan-96 Date
METHODOLOGY Bottom up Top down
Sales Forecasting Example 80000 60000 40000 20000 0 Jun-91 Jun-92 Jun-93 Jun-94 Jun-95 Jun-96 Date
ISSUES AND PROBLEMS
ISSUES AND PROBLEMS New products or products with minimal sales history
Sales Product Example 1000 900 800 700 600 500 400 300 200 100 0 Jan-92 Jan-93 Jan-94 Jan-95 Jan-96 Date
ISSUES AND PROBLEMS New products or products with minimal sales history Read Budgets
ISSUES AND PROBLEMS New products or products with minimal sales history Read Budgets Mathematically Derive
ISSUES AND PROBLEMS New products or products with minimal sales history Read Budgets Mathematically Derive» Date of first sale
ISSUES AND PROBLEMS New products or products with minimal sales history Read Budgets Mathematically Derive» Date of first sale» Expected sales for the first 12 months
ISSUES AND PROBLEMS New products or products with minimal sales history Read Budgets Mathematically Derive» Date of first sale» Expected sales for the first 12 months» Trend
ISSUES AND PROBLEMS New products or products with minimal sales history Read Budgets Mathematically Derive» Date of first sale» Expected sales for the first 12 months» Trend» Similar product
ISSUES AND PROBLEMS New products or products with minimal sales history Read Budgets Mathematically Derive» Date of first sale» Expected sales for the first 12 months» Trend» Similar product» Seasonal breakdown (known)
ISSUES AND PROBLEMS New products or products with minimal sales history Read Budgets Mathematically Derive» Date of first sale» Expected sales for the first 12 months» Trend» Similar product» Seasonal breakdown (known) Deseasonalise the data and use simple models (eg Exponential Smoothing)
ISSUES AND PROBLEMS New products or products with minimal sales history Product succession
ISSUES AND PROBLEMS New products or products with minimal sales history Product succession Some forecasts need to be adjusted
FORECASTS TO BE ADJUSTED Reasons:
FORECASTS TO BE ADJUSTED Reasons: Significant future events
FORECASTS TO BE ADJUSTED Reasons: Significant future events Forecasts are not politically acceptable
FORECASTS TO BE ADJUSTED Reasons: Significant future events Forecasts are not politically acceptable System overreacts to short term trends
FORECASTS TO BE ADJUSTED Reasons: Significant future events Forecasts are not politically acceptable System overreacts to short term trends Solutions: Online editing of forecasts
FORECASTS TO BE ADJUSTED Reasons: Significant future events Forecasts are not politically acceptable System overreacts to short term trends Solutions Online editing of forecasts Quickly find potential problems
FORECASTS TO BE ADJUSTED Reasons: Significant future events Forecasts are not politically acceptable System overreacts to short term trends Solutions Online editing of forecasts Quickly find potential problems» Displaying forecasts
FORECASTS TO BE ADJUSTED Reasons: Significant future events Forecasts are not politically acceptable System overreacts to short term trends Solutions Online editing of forecasts Quickly find potential problems» Displaying forecasts» Analysis of how successful previous forecasts have been
FORECASTS TO BE ADJUSTED Reasons: Significant future events Forecasts are not politically acceptable System overreacts to short term trends Solutions Online editing of forecasts Quickly find potential problems» Displaying forecasts» Analysis of how successful previous forecasts have been Specify model(s) for future forecasts
ISSUES AND PROBLEMS New products or products with minimal sales history Product succession Some forecasts needed to be adjusted The selection criteria of the best model
ISSUES AND PROBLEMS New products or products with minimal sales history Product succession Some forecasts needed to be adjusted The selection criteria of the best model Mean Absolute Percentage Error (MAPE) Root Mean Squared Error (RMSE) Akaike Information Criterion (AIC)
ISSUES AND PROBLEMS New products or products with minimal sales history Product succession Some forecasts needed to be adjusted The selection criteria of the best model Freak Sales
ISSUES AND PROBLEMS New products or products with minimal sales history Product succession Some forecasts needed to be adjusted The selection criteria of the best model Freak Sales Tuning the system
ISSUES AND PROBLEMS New products or products with minimal sales history Product succession Some forecasts needed to be adjusted The selection criteria of the best model Freak Sales Tuning the system The number of products to forecast
ISSUES AND PROBLEMS New products or products with minimal sales history Product succession Some forecasts needed to be adjusted The selection criteria of the best model Freak Sales Tuning the system The number of products to forecast The number of models (and their variations) to try
ISSUES AND PROBLEMS New products or products with minimal sales history Product succession Some forecasts needed to be adjusted The selection criteria of the best model Freak Sales Tuning the system The number of products to forecast The number of models (and their variations) to try The number of periods of sales history to use as input
ISSUES AND PROBLEMS New products or products with minimal sales history Product succession Some forecasts needed to be adjusted The selection criteria of the best model Freak Sales Tuning the system The number of products to forecast The number of models (and their variations) to try The number of periods of sales history to use as input The speed and availability of the machine
STRENGTHS CONCLUSION
CONCLUSION STRENGTHS Saves time
CONCLUSION STRENGTHS Saves time Generally produces good forecasts
CONCLUSION STRENGTHS Saves time Generally produces good forecasts Unbiased & Apolitical
CONCLUSION STRENGTHS Saves time Generally produces good forecasts Unbiased & Apolitical Input into budgeting
CONCLUSION STRENGTHS Saves time Generally produces good forecasts Unbiased & Apolitical Input into budgeting Forecasts can be exported directly to other systems
CONCLUSION STRENGTHS Saves time Generally produces good forecasts Unbiased & Apolitical Input into budgeting Forecasts can be exported directly to other systems WEAKNESSES
CONCLUSION STRENGTHS Saves time Generally produces good forecasts Unbiased & Apolitical Input into budgeting Forecasts can be exported directly to other systems WEAKNESSES Doesnt produce perfect forecasts every time
CONCLUSION STRENGTHS Saves time Generally produces good forecasts Unbiased & Apolitical Input into budgeting Forecasts can be exported directly to other systems WEAKNESSES Doesnt produce perfect forecasts every time Doesnt know the future