So which came first the report or the data. This is a trick question, just like the chicken and the egg. It is obvious, the business.
In a recent presentation from a prominent reporting company, the presenter said the following (and I paraphrase):
"People build their Data Warehouses without any thoughts of the reporting tools"
Some of us were agast. We know that often, source OLTP data structures are in place and even operational data stores, but DW? As most business intelligence purchases are driven by the Business, often the BI tool is chosen first and then IT goes looking for the data to build the reports and gets frustrated on not finding it and builds a datamart specifically for this reporting application to organize the data for the new tool.
Upon reflection, realization was that it often is a tug of war between the data structures and the BI tools on who drives the process. Do you translate your data from third normal form (3NF) to Star/Snow-flake schema, build datamarts, cubes to support the best practices of the BI tool? Or do you manipulate the tool to leverage 3NF and go right off your Data Store? I do not want to fight a religious battle here so I let you answer these questions, but leave you with the following thoughts:
- the more you manipulate data the more the latency, data quality and cost
- if you cant find the data in a timely manner your BI tool is useless
- the better you make your data available to the appropriate parties, the better evolution to self serve for BI