Algorithms Tab

The Algorithms Tab in the Forecast section of Advanced Business Planning provides the capability for forecasters to adjust the history used for forecasting. This is most applicable to adjusting history when unusual / unrepeatable events skew history and undermine the forecastability of the data. This feature is only available in the Forecast section of Advanced Planning.

This method can be done in conjunction with the remove outlier functionality.

Overview

The Algorithms tab is located within the Forecast section of Advanced Business Planning.

The default chart view of the algorithm tab is a year-on-year bar chart with the adjusted history as the default data that is charted.  

Clicking the chart configuration icon open a pop-up to change to a line chart and/or add additional data elements.

The bar charts is clickable, so that upon receiving a click event, the Adjusted history grid is scrolled to the corresponding Adjusted history period, and the period is highlighted for the user to edit as needed.

Please note that when view the chart as a line graph, only clicking on the unit value will toggle the corresponding period.

The Algorithms tab includes four sections:

  1. Forecast Units Grid
  2. Forecast History Units Grid
  3. History and Forecast Chart
  4. Node forecast statistics

The Algorithm tab Forecast History displays Initial and Final rows at non-forecast levels and Initial, Adjusted, Final, and Fitted at Forecast Levels. Forecast levels are selected during the forecast process.

Forecast Units Grid

Forecast units grid is the resulting forecast values and includes the following information:

  • Forecast Range
    • Aligned with forecasting planning horizon (more content is displayed by scrolling to the right) and forecasting bucket type. Starting bucket is based on last forecast run, not rolling period (i.e. Forecast run on November '20, current period Jan 21' → first period displayed will be Nov '20). In the image below the first period is December 2020 which aligns with the month the statistical forecast was run.
  • Forecast Values
    • Statistical forecast values produced from last forecast analysis.
    • There is no special color coding of these values.
  • Grid actions
    • Export all data by "xlsx" icon.

Forecast History Units Grid

  1. History Range
    • Aligned with forecasting history horizon (more content with scroll to the right) and forecasting bucket type. 
    • Starting bucket based on historical interval used during last forecast run.
    • End bucket always last full bucket before the starting bucket of the Forecast Units Grid. In case of  Ignore last N buckets, header cells are color coded in yellow and there is a tool tip to visualize which are not used during the forecast analysis.
  2. Initial 
    • This is the latest forecast history from the customer orders upload.
    • This history can impacted by changes to the customer orders data, linking of items, system settings related to basing forecast history on the original due date and quantity or shipped date and quantity, execute or unexecuted past events. As a note, many of these changes are applied with the next data upload which triggers the historical upkeep process.
  3. Adjusted
    • Provides capability to make user-defined manual adjustments (a.k.a. history editing). Users may directly enter edits to the historical values here. To learn more, read the article Adjusting Forecast History.
  4. Final
    • Snapshot of forecasting history from last forecast analysis with applied manual and system adjustments (outliers).
    • Available at forecasting levels only.
    • Same as adjusted if no outliers. In case of Outliers correction, the buckets with outlier are color coded and the value is different than the adjusted one with a tool tip that communicates the outlier. 
    • Empty value if leading zeroes removed.
  5. Fitted 
    • Snapshot of fitted values from last forecast analysis including impact of outlier removal and history edits.
    • Available at forecasting levels only. 
    • Empty value if leading zeroes removed.

Forecast Chart

Chart Data

  • Combining data sources (and their ranges) from both grids.
  • Starting point of chart will be aligned with starting point of Forecast Units Grid.
  • Includes Initial, Adjusted, Forecast for all levels.
  • Includes Fitted and Final for forecasting levels.

Chart Options

  • Accessible by clicking the "Chart Options" button next to 'Run Forecast' button.
  • Includes "Chart Type" selection
  • Data sources available for select/deselect for cases when chart are cluttered (i.e. in weeks). Аll the data sources are listed at all levels. If any data sources are not available for currently selected level, this is reflected in chart itself.

Chart Types

  • Line chart similar to plan tab one.
  • Bar chart to visualize year over year. 

Chart Colors

In case of line chart, there is only one color per data source.

In case of bars, the same color is used with different opacity for one data source but different years.

  • Initial - Green
  • Adjusted - Pink and same as Forecast History line in other charts
  • Final - Blue
  • Fitted - Yellow
  • Forecast - Purple and same as Statistical Forecast line in other charts

Historical Interval

  • Defined by the system settings Forecast history length (months) and Ignore last N buckets.
  • Always aligned to full buckets based on the forecasting bucket type.
  • Leading zeros are removed within the interval (a leading zero is the time frame before demand starts within the historical time frame selected for forecasting within System Settings.
  • Used across all forecasting related calculations (the demand profile, ADI, COV, algorithm selection, forecasting itself and forecast statistics).

Additional Info

  • All units in grids and chart are rounded based on Units rules.
  • For newly created companies, that have never run the forecast analysis, the tab elements will be empty

PCL Ration Distribution Period

This capability enables users to specify the number of periods that should be used in calculating the weights used in disaggregating forecasts at higher levels of the planning hierarchy into the forecasts of component products. Prior to this enhancement, 12 months of history of each Product Customer Location (PCL) context was used to compute its weight, which was utilized in disaggregating the forecast from its parent.

This enhancement enables the user to globally specify the number of monthly periods that will be used for calculating the weight of all PCLs. To set the ratio to be used globally for all PCLs , go the the Systems Setting, and specify the number of months to be utilized in calculating the ration in the PCL Ratio Distribution Period (months) setting. The default value is set to 12 months.

The global PCL Ratio Distribution Period (months) setting may also be overwritten for each PCL context by modifying the PCL Ratio Distribution Period (months) setting under the Forecast Algorithm tab for each PCL.  The PCL Ratio Distribution Period (months) that is defined at the context level will take precedence over the value defined at the global level.  

Once the PCL Ratio Distribution Period (months) is defined, the last full N months in the forecast interval will be utilized for the forecasting, where N is defined as the number of months specified in the PCL Ratio Distribution Period (months).

0 Comments

Add your comment

E-Mail me when someone replies to this comment

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.