At the onset of an innovative Revenue Management project, determining where to target the analysis for quick hits and maximum returns is essential. The project team is agile, quick to shift the focus from one area to another, without adding significant overhead or setup time. This approach is crucial for an effective project.
A properly designed Analysis Dataset (AD) enables project teams to have maximum flexibility, ensures a consistent base of data for further project analysis, and gives the project team an edge when uncovering quick hits and maximum returns.
The concept of the Analysis Dataset is counterintuitive to the way most datasets are designed for enterprise data warehouses. Rather than trying to normalize tables and fields, the Analysis Dataset looks to denormalize and combine all relevant information at the most granular level of data available into a single dataset.
Denormalization serves two critical purposes:
- It combines all available data into a single dataset allowing project teams to identify significant factors which may not be readily apparent when spread across tables;
- It improves the speed and performance of analysis by reducing database overhead related to table joins, typically the most time consuming database operation.
To create the essential Analysis Dataset via denormalization, you must gather all the relevant Fact and Dimension tables from customer data warehouses at the most granular level available.
Without the most granular level of data, key data insights can be lost, especially if data has already been aggregated to a higher level.
In addition to the data warehouse, external data sources are often also necessary to create a holistic data picture.
For example, when looking at retail data, the most granular dataset is typically thought of as a Point of Sale (POS) transaction. However, for an Analysis Dataset, POS data does not give a complete picture of the actual state of a particular item at a point in time. In addition to the sales transaction, it’s important to also have data for things like inventory, promotions and competitor pricing, even on days when a particular item may not have been sold at all. Without that complete picture of the data, the downstream analysis will be missing crucial information.
With the Analysis Dataset as the combined source of data, the analysis can shift rapidly among different subject areas and levels of aggregation, while retaining the consistent underlying data and minimizing overhead. Denormalization creates the right data environment to allow the right data analysis, at the right time, producing rapid productivity, and results in a successful project.