The rolling horizon modeling workflow

• Set up your model with historical demand (typically in the Customer Orders table) and forecast data populated in the Customer Forecasts table. For example:
  • Assume a model with one year of historical demand data (from 1/1/2018 through 12/31/2018). This data is populated in weekly buckets in the Customer Orders and Customer Forecast input tables.
  • 3 months of future forecast data is also populated in the Customer Forecast table. This data is in weekly buckets from 1/1/2019 through 3/31/2019.

    

• Set your overall model horizon start and end dates.
  • Our model horizon is from the start of the historical data through the first month of the planning data (1/1/2018 to 1/31/2019)
• Enter the Rolling Horizon Start Date on the Rolling Horizon Modeling tab in Inventory Optimization options.
  • The future data begins on 1/1/2019, so this is the rolling horizon start date.
• If desired, you can populate future forecast means in the User Defined Customer Forecast Profile table.
• Run Safety Stock Optimization + Rolling Horizon Modeling. This sets the strategic inventory placement and sets the policy parameters for the planning horizon.
• Confirm the results in the Rolling Horizon Inventory Policy Summary output table.
  • In this table, summary results are populated by Date, based on the Forecast Time Bucket on the Safety Stock Optimization: Advanced tab in Inventory Optimization options.

    

• If needed, populate new future demand and forecast data.
•  Adjust the model horizon end date and the Rolling Horizon Start Date.
  • We'll now look at the second month of planning data, so we set the model horizon end date date to 2/28/2019 and the Rolling Horizon Start Date to 2/1/2019.
• Run Rolling Horizon Modeling. The results from the original Safety Stock Optimization run are used.
  You must have Safety Stock Optimization output populated if you are running the Rolling Horizon Modeling problem type.
• Confirm the new results in the Rolling Horizon Inventory Policy Summary output table.