Dimension Files
A dimension is a collection of descriptive elements, attributes, or hierarchical structures that provide context to your business data. Dimensions tell the platform what your business looks like and how it operates. They describe the factual data (such as sales transactions) and provide means for aggregation and summarization throughout the platform. Dimensions follow a strict set of business rules and formatting requirements that must be followed when generating the files.
There are certain common rules that apply across all of the dimension files and must be followed without exception. Failure to adhere to these rules may result in failed data loads or incorrectly structured datasets in the platform.
-
All dimension files must be provided as full snapshots of the source data at all times, unless you change the configuration of a specific dimension to be
IS_INCREMENTAL=Y
where incremental loads are supported. Incremental dimension loading should only be done once nightly/weekly batch processing has started. Initial/history dimension loads should always be full snapshots. -
Hierarchy levels must follow a strict tree structure, where each parent has a 1-to-N relationship with the children elements below them. You cannot have the same child level identifier repeat across more than one parent level, with the exception of Class/Subclass levels (which may repeat on the ID columns but must be unique on the UID columns). For example, Department 12 can only exist under Division 1, it cannot also exist under Division 2.
-
Hierarchy files (product, organization, calendar) must have a value in all non-null fields for all rows and must fill in all the required hierarchy levels without exception. For example, even if your non-Oracle product data only has 4 hierarchy levels, you must provide the complete 7-level product hierarchy to the platform. Fill in the upper levels of the hierarchy with values to make up for the differences, such as having the division and group levels both be a single, hard-coded value.
-
Any time you are providing a key identifier of an entity (such as a supplier ID, channel ID, brand ID, and so on) you should fill in the values on all rows of the data file, using a dummy value for rows that don’t have that entity. For example, for items that don’t have a brand, you can assign them to a generic “No Brand” value to support filtering and reporting on these records throughout the platform. You may find it easier to identify the “No Brand” group of products when working with CDTs in the AI Foundation Cloud Services or when creating dashboards in RI, compared to leaving the values empty in the file.
-
Any hierarchy-level ID (department ID, region ID, and so on) or dimensional ID value (brand name, supplier ID, channel ID, store format ID, and so on) intended for Planning applications must not have spaces or special characters on any field, or it will be rejected by the PDS load. ID columns to be used in planning should use a combination of numbers, letters, and underscores only.
-
Any change to hierarchy levels after the first dimension is loaded will be treated as a reclassification and will have certain internal processes and data changes triggered as a result. If possible, avoid loading hierarchy changes to levels above Item/Location during the historical load process. If you need to load new hierarchies during the history loads, make sure to advance the business date in the data warehouse using the specified jobs and date parameters, do NOT load altered hierarchies on top of the same business date as previous loads.
-
All fields designated as flags (having FLG or FLAG in the field name) must have a
Y
orN
value. Filters and analytics within the system will generally assumeY
/N
is used and not function properly if other values (like0
/1
) are provided. -
Retail Insights requires that all hierarchy identifiers above item/location level MUST be numerical. The reporting layer is designed around having numerical identifiers in hierarchies and no data will show in reports if that is not followed. If you are not implementing Retail Insights, then alphanumeric hierarchy IDs could be used, though it is not preferred.