I recently came across an article that stated ‘No Data Warehouse Required for BI’. I immediately understood that the article is going to stress the relevance of operational Business Intelligence (BI) over constructing data warehouses in a typical BI environment. I am a big believer of the fact that building data warehouses with a strong foundation in underlying database management is critical for reporting information that helps to make good decisions. Since the article intrigued me, I started exploring both the options without prejudice and here are some of my views. I would love to hear how others think.
The crux of the article was that the scope of BI today is to deliver operational information to business users and that it has nothing to do with complex analytics or advanced computation. It is as simple as the retrieval of operational information from where it resides. This provides visibility into operational performance, performance monitoring and real-time monitoring of certain technical parameters. Agreed. But, in my opinion, we cannot generalize this. Choosing Operational BI versus building data warehouses is a function of the size of the business, the nature of the reporting, what type of information is required and the timeliness in making decisions that will benefit the business. If the business wants to measure and analyze facts with respect to multiple dimensions or perform rollups at different granularities, a data warehouse is beneficial or even necessary.
Also, sometimes I wonder, if cost is a factor that is triggering decisions to go with operational reporting versus building data warehouses from scratch. It is an investment to buy BI tools, bring in expertise to model the data warehouse, use ETL tools to load the data and then use enterprise reporting tools to report on the data. So how about just throwing together a bunch of operational tables for the reporting tool to perform the miracle of generating accurate reports in a timely manner? I am not a big fan of this idea. The reason is, an arrangement like this would always be an easy solution to begin with and cost effective for the client at that point in time but the struggle starts when that model needs to be extended to accommodate future requirements. How about adding more tables? Will ‘maintenance’ be an issue? How about performance? Also, the more you start reporting on just the information, the more you will realize that adding analytics would be valuable to make certain decisions that you did not think would be helpful in the beginning. While an operational set up is ‘write-optimized’, a data warehouse structure is ‘read-optimized’ and conducive for reporting.
Finally, even if you would decide to just report on ‘operational’ data using operational BI, dealing with data quality, inconsistent data between multiple sources or invalid data could potentially hinder the ultimate goal. These hindrances are like programming bugs: if you invest in a debugger, you are saved. At the very least, invest in data quality and data management for a better foundation.