We have seen two approaches in the industry in building data warehouses:
“Build it first, people will come” theory – Significant time and effort was invested in integrating large sets of information assets within the organization. Usage of third normal form (3NF) to avoid duplication of data was utilized to achieve the ultimate goal of clean data. However, factors such as lack of business buy-in, time and effort, and the performance of running analytics in this data have impacted this model.
“Build it first, fix it later” theory – Building data marts that are more department centric to achieve analytical data warehouse satisfying the business sponsor’s needs. However, denormalized modeling hindered performance on loading and duplicated data causing unclean and inaccurate reports.
So what is the best design for an enterprise data warehouse (EDW)? Although many components of the data warehouse such as data models can be generic, there is no single best approach, be it Kimball or Inmon. However, the chosen data model and data integration processes for EDW have a significant effect on the time to value and adaptability to changes in the system. Our approach in building or changing EDW can be a flexible top-down or bottom-up approach as our information architects have a thorough understanding of the EDW data model and its subject areas such as Enrollment, Facility and Professional claims, Medicare advantage, and Providers.