Post on 20-Aug-2015
Adexa Supply Chain Planning Web Series:
Deciding When to Apply Statistical Forecasts
Presented by:
William Green –wgreen@adexa.com
2The Role of Statistical Forecasting
•Products should be divided into categories to define how the statistical forecast should be used. The 80/20 rule should be used to classify products based on volume , value and
ability to forecast accurately
•High value items should be managed directly by planners when it is shown that human intervention can add value.
•Statistical forecasts should be used when shipment history provides reliable data to predict the future, and as a check against the opinions of the various stakeholders when
human intervention is used
•The large number of lower value items making up less of the revenue should be managed with the statistical forecast and checked at an aggregated level.
•For items with sparse data or a lot of random errors, a statistical forecast at an aggregated level is the best way to go.
Demand Management Product Classification
• Products are classified based on volume/value and forecast accuracy
• Forecast accuracy metrics of MAPE, RMSE, etc.
Vo
lum
e/V
alu
e
ForecastError
High ValueVolatile
High ValuePredictable
Low ValueVolatile
Low ValuePredictable
High
Low
Low High
Statistical Engine
Using Hierarchies based on product classification
• Statistical Forecasting can be used at any level of the hierarchy
• Too High and trends are hidden for higher value and higher volume products. (Example: One product is trending up in a group going down)
• Too low and the noise hides the trend. There is a basic lack of data
Disaggregation
Product Hierarchy
Aggregation
Manage Low Volumes
High VolumeProducts
5The Basics of Statistical Forecasting
• The past history is used in order to understand the trend and level of the forecast over time
• What is learned from the history is used to project into the future
• If the organization does not manage the historical data, then the forecasts will not be good
• The business can use history and the forecast as a discussion of where the business is going
History and fitStat Fcst
• Next Part 2: Business Managers Overview of the most popular Statistical Forecasting Techniques
• For more information please feel free to visit:
www.adexa.com
• Send follow up questions to William Green at:
wgreen@adexa.com
• Please feel free to contact us at:
events@adexa.com
Thank You