Overview of Forecast Trees

A forecast tree is a structure of hierarchy levels at which data can be aggregated during the demand forecasting process.

The forecast tree enables Oracle Demand Management to aggregate data while performing analytical forecasts. In other words, the forecast tree for a forecasting profile organizes the data for the forecasting process.

The forecasting process begins at the most granular (lowest) level of the forecast tree. If the forecast can’t be made at this level because of inadequate or sparse data, the process moves up the forecast tree and attempts to forecast at higher levels of data aggregation. In general, forecasts are more accurate when they’re made at lower levels of the forecast tree.

The forecast tree supports up to two dimensions. The available dimensions for the forecast tree are Product, Organization, Customer, and Demand Class. Your configuration of the forecast tree depends on your business. You typically use the Product dimension along with another dimension (for example, Product and Organization, Product and Customer, or Product and Demand Class). With respect to the Product dimension, you can use levels from any hierarchy except for levels from the configure-to-order (CTO) hierarchy (Model BOM).

The runtime is short if the forecasting process has to do less data aggregation while traversing the forecast tree. If the data quality allows, use the lowest levels for two dimensions (for example, Item and Customer Site) as the first forecastable level in the forecast tree. The runtime is long when the lowest forecastable levels are aggregated, or only one dimension is used.

The quality of the forecast is influenced by the forecast tree’s construction. Thus, you must carefully include and arrange levels for the forecast tree.