What are Late Arriving Facts?
In data warehouses, dimensions are processed first and the facts are loaded later, with the assumption that all required dimension data is already available in the system and a Parent fact is already available before a child fact. But, this scenario may not always be the case because the nature of business processes differ for each system. Fact data can be sent from the source system to the warehouse way later than the actual fact data is created. So, the actual facts may not get loaded into the target warehouse system.
An Example in HCPDM – Member Contact:
Each Member in a HCPDM has a set of target tables holding contact details like Address, Phone and Email. Usually Member facts are created first and contacts are created later. Since the data is processed for a time interval incrementally, Members may arrive into the warehouse source system first and a Member may not be processed for an interval. But contact facts are available in the source system with no member parent fact. Contacts will not get loaded into the target as the member’s parent facts are not available for a contact.
Resolution:
The above scenario can be resolved by using a secondary mapping. During the Child Contact load into the target system, a secondary mapping is introduced into the contact loading job to process the parent member facts which were left out of the data processing during the primary member load using the timestamp. This helps in synchronizing the database.
