Forecast Workbench
Discover how to optimize demand forecasting using Claret's Forecast Workbench, utilizing statistical models, historical data, and custom configurations."
The Forecast Workbench is a planning tool that allows planning teams to use statistical models with historical sales data to forecast demand into the future.
The workbench uses R modeling techniques and can either automatically select the model that predicts the most likely outcome or allow you to select a model yourself.
The Forecast Workbench is accessed from the main menu.
Before using the Forecast Workbench, make sure the Forecast Workbench-related Application Settings have been set.
When you first enter the Forecast Workbench page, all Items and Customer Groups are listed on the left of the page. Modeling is done at the Item@Customer Group level.
Select the Item and Customer Group you wish to work with by clicking on them. Once selected, the banner at the top will show the Item@Customer Group info.
In order to work with an Item and Customer Group combination within Forecast Workbench, you must already have defined the combination as an Item@CustomerGroup.
If you have used the Forecast Workbench previously for the selected Item@Customer Group, you will see your previous forecast data. If you have not, you will need to setup the forecast configuration.
Forecast Configuration
To access the Forecast Configuration, click on the configuration (blue) button in the top right of the screen.
On the panel that appears, you can now set the following options:
Timing and date configuration
Set the 'Timing Interval' based on what level (weeks or months) you want to forecast at. This will dictate at what level the forecast data is saved.
Next, choose the 'Forecast history start' and 'Forecast history end'. These are the start and end dates of the data set you want to use as the historical sales data.
And finally, choose the 'Forecast horizon in months'. This determines how far into the future the model forecasts.
So, for example, if we are in February, and you wish to use the last 12 months worth of data to forecast the next 12 months worth of data, but in weekly buckets, then your 'Timing and date configuration' would be as follows:
Next, we need to set the 'Model configuration'.
Model configuration
There are two key options when setting up how Claret should model the forecast data.
Auto - If all options are set to Auto - then Claret will determine the best regression model to use to forecast the data.
Manual - If you wish to set the model configuration yourself, then the auto function can be turned off by switching the 'Auto' switch across and you can then make your own selections.
The 'Model Selection' dictates which regression model you wish Claret to use when creating the future forecast based on historical sales. A number of options are available in the dropdown. Select the model you would like to use. (For further detail on how each of the available models work, see Forecasting Models)
2. The 'Missing Imputation' setting tells Claret how you want the model to handle gaps in history.
3. Setting the 'Forecast Constraint' will indicate any rules Claret should use to constrain values when forecasting. Select 'Positivity' to only allow positive forecast values.
Select 'Save' to save all your configuration settings
Running the Model
Once you have set your configuration preferences, you are now ready to run the model. To do this, simply hit the 'Play' button in the top right corner
The model will need to be rerun via this method whenever you make changes to the configuration.
Model Output
The output you see after you run the model contains 3 sections - a chart, details of a selected data point on the graph, and a table of all values where the forecast can then be overridden.
The chart
The chart provides a visual for the historical and projected forecast over time. History data is seen in red and the forecast projected by the model can be seen in blue.
If you hover on a point on the chart, you can then see the exact values for that point in the detail below the chart. History data is shown on the 'History' line and forecast values are shown on the 'Forecast' line. For example, hovering over Feb then shows the Forecast figures for that month, including any overrides.
The historical data used to generate the forecast is highlighted between the 'History start' and 'History end' bars but ALL historical data for the item is shown on the chart.
2. The data
Below the chart area is a table containing all the data for the Item@CustomerGroup. Here, you can scroll through time and see the exact values as charted above. If you wish, you can also alter the data by adding overrides. (See below).
Overriding Forecasts
One of the key features of the Forecast Workbench is the ability to override forecast data to view how this would affect the overall forecast.
There are 2 ways you can override the data for a period.
Enter the amount by which you wish to change the Hist/Forecast figure into the 'Overrides' box. So, if the Forecast is for 252.85, entering 176.60 into the Overrides box would then add 176.60 to the 252.85 to get a total of 429.55 for the month. (To decrease the forecast total, you would enter a negative value into the Overrides box).
2. Enter the total amount you want to change the forecast to into the 'Total' box, and the 'Override' value will then be automatically calculated.
The chart will then update to show the updated value as part of the main chart, but a dotted line to still show the original value.
Last updated
