Each location will attract a share of the total forecast relative to its contribution to sales over the specific time frame. A shorter timespan will result in a more reactive disaggregation while a longer one will be more stable. The time frame may be automatically extended for products with slow moving or irregular demand.

This forms part of the Advanced forecasting module. It is an add-on module and well worth investigating should a business have multiple locations and managing forecasts generated at a location level is challenging. Be sure to give the Advanced forecast module article a read if this sounds applicable.

For businesses selling items with varying sales patterns across locations, this may not be a good solution. Imagine having locations in both northern and southern parts of the hemisphere and the item in question is a snow shovel. Forecast disaggregation may result in less accurate forecasts than forecasting at a location level.

This module will also not be beneficial to businesses currently achieving accurate forecasting at a location level.

However, for the average business, this may be useful and is worth exploring.

When it comes to setting these parameters, a safe value for “Weight by the sales history of the product if it has at least X months of history at the region” is between 3 and 6. This ensures the item has enough sales history to be used in the calculation of its contribution and thus forecast percentage split. If an item only has 1 month’s sales, that month may have been way too high (the new iPhone) or way too low (introduction of the fidget spinner) and is thus not enough to base the calculation on as it may result in an inaccurate split that is not indicative of future sales.

“Otherwise weight by the combined sales history of similar products based on this group” is used to specify the fallback group. If the item in question has less than X months’ sales (specified above), then which Group’s sales should be used to calculate the contribution and thus the forecast percentage split?

Item = Pink sandals may fall part of Group 1 = apparel, Group 2 = Summer wear, Group 3 = Shoes and Group 4 - sandals.

Suppose Pink sandals don’t have enough sales history to be considered in the forecast disaggregation. Setting this parameter to “Group 1” could result in more sandals being shipped to the snowy location because they’ve been more successful in selling apparel - winter apparel that is.

However, setting this parameter to “Group 4” might be too specific and may not have enough data points, thus skewing the contribution.

It’s important to find that sweet spot of accuracy, but with enough data.

“Weight the forecast using the last X months of sales history” can be set to 12 months for a more stable split or 3 months for forecasts that need to adjust more dynamically.

Refer to this article for a detailed explanation of Forecast Disaggregation.

Question: What happens when the “Weight the forecast using the last X months of sales history” is set to 12 months, but the “Weight by the sales history of the product if it has at least X months of history at the region” is set to 3 months? Does that mean the new item with 3 months of history will have to compete with the item that has been selling for 12 months?

Answer: No, it will consider average sales over the months available.