Having data scattered about the organization in relational databases is usually not the best approach for decision analysis. For example, related data stored in different DBs can easily have differing formats and/or coding systems. Consider one manufacturer that maintained product data in one DB while maintaining order entry data in a different DB. Orders were placed using a product characterization code of, say, '01.' The problem was that the product management DB might have the same product characterization, but using the code of '010.' When reports were generated for analysis that involved both data sets, analysis anomalies were created on any comparison of that particular product characterization code. One way around this problem is to copy all transactional data to a centralized relational DB, also known as a data store or data warehouse.
What kind of difficulties could you imagine for creating such a DB?