Quick Summary: Forecast refinement combines the statistical forecast generated from prepared demand history with current business knowledge. The Computer Forecast provides the baseline, the Working Forecast captures ongoing refinements, and the Final Forecast becomes the approved forecast used for downstream planning throughout the app.
Understanding forecast refinement
Forecast refinement combines the statistical forecast generated from prepared demand history with current business knowledge.
The process follows three stages:
Computer Forecast
↓
Working Forecast
↓
Final Forecast
The Computer Forecast provides the statistical baseline. Forecast Promotions, Forecast Adjustments, Forecast Overrides, and changes to the forecast engine shape the Working Forecast. Once planning is complete, the Working Forecast is approved as the Final Forecast for use throughout the app.
Initially, the Computer Forecast, Working Forecast, and Final Forecast contain the same values. As refinements are made, the Working Forecast begins to differ from the original Computer Forecast. Finalizing then copies the approved Working Forecast into the Final Forecast.
The Working Forecast is represented by the dark green dotted line.
The Final Forecast is represented by the olive green measure in the Planning table and graph.
Expanding the forecast measures
The forecast row can be expanded to display the individual measures that contribute to the Working Forecast.
To display the forecast measures:
Locate the forecast row in the Planning table.
Select the forecast period or expand control.
The table displays the available forecast measures, including:
Computer Forecast
Working Forecast
Forecast Promotions
Forecast Adjustments
Forecast Overrides
Final Forecast
Changes made within the editable forecast measures update the Working Forecast.
Understanding forecast measures
Forecast measures determine how refinements interact with the Working Forecast.
Forecast Promotions
Forecast Promotions are additive.
Values entered in the Forecast Promotions measure are added to or subtracted from the existing Working Forecast. Promotions are intended for temporary demand changes associated with promotional activity.
Saved Forecast Promotions can also influence future Computer Forecast calculations.
Forecast Adjustments
Forecast Adjustments are additive.
They increase or decrease the Working Forecast for the periods in which the adjustment is entered. Forecast Adjustments are appropriate when expected demand requires correction without replacing the forecast value completely.
Forecast Overrides
Forecast Overrides replace the Working Forecast value.
When an Override is entered, demand is set explicitly for the selected period. Other forecast measures do not influence the resulting value.
Overrides are appropriate when expected demand must be defined directly.
Refining the Computer Forecast
The Computer Forecast is generated from prepared demand history and provides the starting point for planning.
When sales history changes, an update is processed to the Computer Forecast for the current hierarchy level. A top-down Computer Forecast can also be performed to carry changes to the child members beneath that level.
To open the Adjust Computer Forecast panel:
Click the Computer Forecast measure in the Planning table.
The Adjust Computer Forecast panel opens at the bottom of the screen.
This panel changes how the forecast engine calculates the Computer Forecast. It does not directly edit individual forecast values.
Forecast Method
The Forecast Method determines which algorithm the forecast engine uses to generate the Computer Forecast, allowing the method to be matched to the applicable demand pattern.
In most cases, the default Expert method is recommended. It evaluates multiple forecasting approaches and automatically selects the most appropriate method.
➜ For more on this topic, read: Demand Types & Sales Forecast Generation Explained
History Basis
The History Basis determines which historical measure the forecast engine uses when calculating the Computer Forecast.
Available options may include:
History Final
Imported History
Outliers
Lost Sales
Overrides
Promotions
Additional Reference Measures
History Final is selected by default.
Changing the History Basis does not immediately update the forecast preview. Re-forecast must be selected before the effect can be reviewed.
Reactivity
Reactivity determines how quickly the Computer Forecast responds to recent changes in demand.
A higher Reactivity setting places greater emphasis on recent demand. The forecast, therefore, responds more quickly to recent increases or decreases.
A lower Reactivity setting places greater emphasis on longer-term historical patterns.
➜ For more on this topic, read: Forecast - Forecast - Reactivity
Maximum Annual Increase and Decrease
Maximum Annual Increase limits forecast growth relative to historical demand.
Maximum Annual Decrease limits forecast decline relative to historical demand.
These controls work with Reactivity. A highly reactive forecast responds more strongly to recent demand, while the annual limits restrict how far the forecast may increase or decrease.
Seasonality
Seasonality determines the threshold used to identify seasonal demand.
A higher setting results in more forecasts being identified as seasonal. Increased sensitivity to seasonal patterns may, however, reduce forecast accuracy in some circumstances.
➜ For more on this topic, read: Forecast - Forecast - Seasonality
Use Period-to-Date
Use Period to Date determines whether demand from the current incomplete period is included in the forecast calculation.
This option is deselected by default.
Period to Date may be useful for new items with very limited history where current sales need to influence the forecast as soon as possible.
Auto-Age
Auto-Age controls how much historical demand is included in the forecast calculation.
When Auto Age is selected, all available history is used and the Periods field is unavailable.
When Auto Age is deselected, the number of historical periods used by the forecast engine can be entered manually.
A restricted historical window may be appropriate when older demand no longer reflects current demand behavior.
Level and Slope
The Level and Slope controls visually adjust the calculated forecast.
Level: Raises or lowers the forecast across the selected horizon.
Slope: Changes the direction and rate at which the forecast increases or decreases.
The preview updates immediately when either control is changed. The corresponding numerical changes are also displayed in the panel for reference.
Level and Slope behave differently from the other Computer Forecast settings:
Forecast Method, History Basis, Reactivity, Seasonality, annual limits, Auto Age, and Period to Date: Require Re-forecast.
Level and Slope: Update the preview immediately.
Selecting Re-forecast again recalculates the forecast from the underlying settings and resets any Level and Slope changes.
Re-forecasting and updating the Working Forecast
To recalculate the Computer Forecast:
Configure the required forecast settings.
Select Re-forecast.
Review the updated preview.
Adjust Level or Slope where required.
Select Apply.
Re-forecast updates only the preview.
Apply commits the recalculated Computer Forecast and updates the Working Forecast. This allows different forecast settings to be evaluated before the result is committed.
Forecast Profiles
Forecast Profiles allow a set of Computer Forecast parameters to be saved and reused.
Profiles are useful when similar forecast behavior applies to more than one product, location, or hierarchy member.
Common uses include:
Using the demand profile of an established item for a new item with little or no sales history
Using a demand profile from one store for another store
Reusing a profile for similar promotional activity across locations.
To save a Forecast Profile:
Configure the required Computer Forecast settings.
Select Profiles.
Enter a meaningful profile name.
Select Save.
Existing profiles can be selected from the Forecast Profiles panel and applied to another calculated forecast. Profiles that are no longer required can also be deleted.
Selecting a profile updates the forecast parameters. Re-forecast and Apply must then be selected to update the Working Forecast.
Building the Working Forecast
Forecast values can be adjusted individually or across multiple selected cells using the Forecast Adjustment panel.
The effect of the change depends on the selected measure:
Forecast Promotions: Positive values increase demand, while negative values reduce it
Forecast Adjustments: Positive values increase demand, while negative values reduce it
Forecast Overrides: Replace the Working Forecast with the entered value
➜ For detailed instructions, read: How To: Adjust Values in Demand Planning
💡 Pro Tip: In some cases, a saved Forecast Override may be recorded correctly while the Working Forecast temporarily appears to display the Computer Forecast value. Refresh the browser to update the display and show the saved Override.
Shaping changes over time
Forecast changes can be distributed across time using a curve instead of assigning the same pattern to every selected period.
Available curve types include:
Growth
Decline
Available curve behaviors include:
Linear
Start Fast
Start Slow
The Curve Visualizer previews how demand will change across the selected periods, allowing the planned pattern to be reviewed before it is adjusted.
After selecting the required curve type and behavior, select Adjust to update the Working Forecast.
Working across hierarchy levels
Forecast refinement can take place at any hierarchy level.
Higher levels support broader planning decisions, while lower levels support more targeted changes.
When a forecast is adjusted at a higher level, Demand Planning distributes the change to the more detailed hierarchy members beneath it. The hierarchy structure and proportional distribution remain intact.
Forecasts can therefore be refined where business insight is strongest without editing every item individually.
The Multi view can then be used to review and refine how the higher-level change has been distributed across the next hierarchy level.
Spreadsheet Collaborator
Spreadsheet Collaborator supports larger forecast updates by allowing planning data to be downloaded, edited in a spreadsheet application, and uploaded back into Demand Planning.
The feature is accessed from the Multi view.
Selecting and downloading data
Before downloading:
Select the required hierarchy level.
Select the measures to include. Available measures may include:
Historical measures
Forecast measures
Promotional measures
Reference Measures
Other available planning measures
Select the required worksheet format. The spreadsheet can be downloaded with:
All selected measures in one worksheet.
Each selected measure in a separate worksheet.
Editing the spreadsheet
The downloaded spreadsheet may be edited in the preferred spreadsheet application.
The following fields must remain unchanged:
System tab: Contains metadata required to process the upload.
ID column in the Data tab: Maps the uploaded values to the correct hierarchy members.
After the changes are complete, save or export the file before uploading it.
Uploading and validating the spreadsheet
When the completed spreadsheet is uploaded, the app validates the file by cross-checking the System tab and checking for issues such as:
A missing ID column
Incorrect planning periods
Once validation succeeds, select the measure that should receive the uploaded values.
The destination does not need to be the same measure from which the data was downloaded. Values can be mapped to another applicable forecast measure or Reference Measure.
After processing, the uploaded values are automatically distributed proportionally (prorated) to the lowest hierarchy level.
Product lifecycle planning
Historical demand is not always an appropriate basis for future demand.
A new product with no sales history will not yet have a forecast. A product approaching the end of its lifecycle may also require demand to decrease intentionally rather than follow its existing historical pattern.
Forecast Overrides and lifecycle curves allow these demand patterns to be defined directly.
Planning demand for a new product
To model demand for a new product:
Navigate to the product level.
Enter the expected steady-state demand in the Forecast Overrides measure.
Select Growth.
Select the required curve behavior.
Review the Curve Visualizer to see how demand will change across the selected periods.
Select Apply.
Available growth curve behaviors include:
Linear
Start Slow
Start Fast
The curve builds demand gradually toward the expected steady-state level rather than entering the full value immediately.
Planning demand for a product in decline
To model declining demand:
Copy the Computer Forecast into the Forecast Overrides measure.
Select Decline.
Select the required curve behavior.
Review the Curve Visualizer to see how demand will change across the selected periods.
Select Apply.
The curve reduces demand gradually rather than causing an immediate drop.
Improving forecast decisions
Forecast Risk, Forecast Bias, the Analysis view, and Comments can be used to identify where planning attention is most valuable and record the reasoning behind forecast decisions.
➜ For more on this topic, read: Demand Planning: Forecasting Effectively
Finalizing the forecast
Once all forecast changes have been saved and the Working Forecast reflects expected demand, it can be finalized.
Finalizing copies the Working Forecast into the Final Forecast. The Final Forecast is then used throughout the app for downstream planning.
To finalize the forecast:
Confirm that all required changes have been saved.
Select Finalize, represented by the green flag icon.
The Working Forecast is copied into the Final Forecast.
If additional refinements are made later, select Finalize again to replace the Final Forecast with the latest Working Forecast.
The Working Forecast remains visible as a separate measure, allowing the planning version and approved Final Forecast to be compared.
How To: Adjust Values in Demand Planning
⚠️ Watchouts
Select the appropriate forecast measure: Forecast Promotions, Forecast Adjustments, and Forecast Overrides affect the Working Forecast differently. Select the measure that best reflects the reason for the change.
Re-forecast before selecting Apply: Changes to Forecast Method, History Basis, Reactivity, Seasonality, Auto Age, Period to Date, and annual limits require Re-forecast before the updated result can be committed.
Re-forecasting resets Level and Slope changes: Level and Slope update the preview directly, but selecting Re-forecast again recalculates the forecast from the underlying settings.
Preserve Spreadsheet Collaborator system fields: Do not modify the System tab or the ID column in the Data tab, as they are required to validate and map uploaded data correctly.
Save changes before finalizing: Unsaved changes are not included in the Working Forecast and should be saved before the forecast is finalized.
Finalize after further refinements: Changes made after finalization do not automatically replace the Final Forecast. Finalize again when the latest Working Forecast is ready for downstream planning.
💡 Tips
Start with the Computer Forecast: Use the statistical forecast as the baseline before incorporating business knowledge.
Refine where insight is strongest: Forecast changes can be made at higher hierarchy levels and automatically distributed to more detailed levels.
Use Forecast Profiles for recurring demand behavior: Saving commonly used forecast settings reduces repeated configuration and supports consistency across similar products or locations.
Use the Multi view to review distribution: After making a higher-level adjustment, review the lower-level values to confirm that the distribution remains appropriate.
Use lifecycle curves for deliberate growth or decline: Growth and decline curves provide a structured way to model products that cannot be forecast reliably from historical demand alone.
Forget about these 👇 😞 😐 😃 Have your say here!