Post Load Processing(PLP) in ETL

         The information loaded into the Business Analytics Data Warehouse is further summarized and enriched to facilitate quick retrieval for analysis by pre-aggregating voluminous information and performing complex calculations. This task is performed by the post-load processing layer packaged in Oracle BI Applications. Since the processing involves taking information from the Business Analytics Data Warehouse base/transactional tables and loading it into other aggregate warehouse tables, this layer is source independent. Post Load Processing is designed to reduce extraction time windows by freeing up source systems as quickly as possible to support today’s complex business environments that include 24x7 usage.

          Since the processing is focused on supporting quick and complex analysis needs, the nature of processing required depends on the specific analysis. Hence, a lot of flexibility is built into this layer to configure it for specific analysis needs - eg. instead of monthly aggregates, performing weekly aggregates might be more beneficial to analysts, or rather than building periodic snapshots for products it might be better to have periodic snapshots for product families. The post load processing layer is architected to easily configure for such requirements. Additionally,other summaries can be built with the building blocks used in the post load processing layer.

The post load processing layer provides several benefits:

• Since it is a source independent layer, configuration in post load processing is shielded from operational sources regardless of the number of source systems from which data is extracted and loaded into the Business Analytics Data Warehouse.

• Complex metrics, which require data from many periods and complex statistical or data mining algorithms, are computed in the post load processing layer – for example, forecasts, profitability indexes, sensitivity indexes etc.

• Normally, transaction grain tables capture facts that actually occurred. However, for the purpose of many analyses, what did not happen may be more interesting - e.g. which products were on promotion that did not show sales improvement, or which customers that were offered incentives did not buy, etc. Such information can be derived in the post load processing layer.