Contributors Ruvisha Pillay and Bill Tonetti

By now, you should have a good understanding of the “Expert” forecasting engine, the various statistical methods that it considers and procedures for overriding the Expert to select alternative methods. You should also be familiar with the Visual Forecaster “Sliders,” which is a valuable tool for visually fine-tuning your forecasts.

In this article, we aim to describe the other forecast parameters that can be used to further refine your statistical forecasts.

To update your forecast parameters on the base level item of your model, you can use the Visual Forecasting panel. You will be able to identify a base level item by the dot found next to the selection on the Tree.

To update your forecast parameters from an aggregate node on your model, you can right-click on the selected Tree node and select “Update Forecast Parameters.” Recall that aggregate nodes are identified by having a folder next to the selection instead of a dot.

There are 5 other forecast parameters on these forms in addition to Forecast Type. They include:

Let’s explore each of these parameters.

Seasonal indexes are an extremely valuable tool for improving forecasts. While certain item groups, locations, or channels may exhibit clear seasonal patterns when forecasted at group levels, the randomness of demand at the base level can make it challenging to detect seasonal patterns at more granular levels.

Seasonal indexes can be created in two ways. The simpler way is to click on a group node that demonstrates the seasonality and then click on “Snapshot Current Index” on the Visual Forecaster panel. This user-friendly feature enables you to capture and utilize the current index that was determined by the forecast engine.

When you click on that button, you will be presented with the below form. To save an index for re-use, input a name, then click “Save.” The saved index will be stored and can be used for any items in the model.

On the Top Menu, there is an option called “Seasonal Indexes.”

You can open previously saved indexes to edit, delete or create new indexes.

To edit a previously saved index, simply type over the seasonal values to modify them, then “Save.” If you created an index previously and you want to update it, simply re-capture the index, select “Snapshot Seasonal Index” on the Visual Forecaster form, and then save it with the same name (you will be prompted to make sure you want to overwrite).

If you delete an index, it will automatically be removed from all of the items where it was being used.

When capturing an index from a forecast, it is automatically scaled to an average of 1.0. However, this scaling is not required. Let's take the example of using indexes for a "5-4-4" financial calendar. Here, you can simply enter the corresponding values in the index, assigning a value of 5 for the months that have 5 weeks instead of 4.

Furthermore, the number of index values depends on the time granularity of your model. If your model operates on a monthly basis, there will be 12 index values, one for each month. On the other hand, if your model operates on a weekly basis, there will be 52 index values. While Netstock has the capability to forecast in days as well, it is not supported in the current version of Pivot Forecasting.

Here’s how seasonal indexes work. In the example, we will use the past 4 months and the next and the next 4 months. The index values are shown in the image below, alongside the historic and forecast values.

The calculations follow these steps:

Non-seasonal indexes are the next forecast parameter we will explore. Although similar to a seasonal index, a non-seasonal index differs in that it does not repeat. With the non-seasonal index, a value is required for every historic and future period in the model. A typical model will have 3 years of history with 1-2 years of forecast, which will result in 48-60 index values. Weekly non-seasonal indexes will have even more values. To help with this, Pivot Forecasting has a menu option to upload non-seasonal indexes. This menu can be accessed from the Top Menu, along with the option for viewing and editing them manually. To make sure you have indexes for all periods, it is recommended that you upload indexes for as many years into the future as possible, and establish a process for extending them.

The final three forecast parameters are History Start Date, Forecast Start Date, and Forecast End Date.

This parameter will ignore any historic values that precede the selected date. This is very useful to avoid forecast biases due to pipeline fills and other inconsistent or erratic start-up demand patterns.

This parameter is used for new products, where the system would not prepare a forecast for any periods prior to the selected date.

This parameter is best used for end-of-life products where the system would not produce any forecasts beyond a specified date.

Have you read the related articles? Check out our Pivot Forecasting collection!

The world class forecasting in Pivot Forecasting helps companies like yours to plan better with Netstock!