Summary
The Seasonality Settings tab in Forecast Manager allows you to modify the seasonality profile being used by StockIQ for forecasting your item.
Seasonality Detection
Whenever an item or group of items has more than 18 months of demand history, StockIQ will attempt to determine seasonality for the item by looking at the degree of correlation year-over-year to the peaks and valleys in demand. By default, if the correlation is more than 15%, a mild seasonality model is used, and if there is more than 40% correlation, the special highly seasonal mode of the StockIQ forecast algorithm is used. These settings are configurable. Below the minimum seasonality threshold, a completely flat seasonality weighting is used, e.g. no changes at all due to any seasonality.
Index Type
Auto: When your Seasonality Index Type is set to "Auto", StockIQ will try to do this auto-determination if the item has any seasonality on its own.
No Seasonality: This disables any seasonality behavior for this item or hierarchy level.
Use Hierarchy Level: This option allows you to instruct your item to use the seasonality profile from a different level of your hierarchy, usually at a product group or product family type level. Sometimes you can have overall seasonality at an aggregate level that is not evident on individual parts, so using the seasonality profile at a higher level of aggregation can improve forecast accuracy by allowing StockIQ to "take a step back" and see overall patterns for your group rather than try to find seasonality for a specific item.
When you select this option, an additional "Use Level" select box appears, where you can select the level from which you want StockIQ to use the seasonality curve for this item.
A very powerful option is the "Auto" selection, in which StockIQ will run the seasonality algorithm at each level of the hierarchy AT and ABOVE your selection, and will auto-select which level provides the lowest forecast error. This gives the best of both worlds of being responsive to individual item characteristics, while allowing higher-level trends to be detected.
TIP: Very often you will want to make seasonality setting changes for all items within a product family. To do so, load that product family in Forecast Manager using the tree view, set the seasonality to use the hierarchy level of that product family, and then this setting will be inherited down the hierarchy in the same way that normal forecast settings are:
Use Hierarchy Node: This allows you to have your item use the seasonality of an exact specific other hierarchy node within your forecast hierarchy. This is not very common.
Custom: This allows you to use a previously defined Custom Seasonality Index that you configured in the Custom Seasonality Indexes screen. These are user-defined table-based or filter-based seasonality indexes that can be useful for when your seasonality does not line up neatly with your forecasting hierarchy.
Additional Settings
High Seasonality Max Yrs History - Sets how many years of history to consider when evaluating high seasonality slopes and curves. Mostly this is set to attempt to use up to 3 years of history when available. You may wish to use more (or less) data if some previous years are exceptional, such as in COVID-19 years.
High Seasonality Previous Yr. Weight - Sets how much weight to give to those previous years. As in the COVID-19 example above, unusual/outstanding years can emphasize an item's growth pattern over time, so tweaking this can provide a more reasonable seasonality year-over-year slope.
There are some additional thresholds and parameters that control when items are detected as highly seasonal in the Global Demand Forecast Settings, so if your items are / are not being detected as seasonal, and you think they should / should not, check those as well.