Every couple of months or so, they run an article on Business Intelligence in one of the Sunday business papers here in Dublin. The tenor of these articles goes like this: Just pick the right BI tool, sit back, and relax. It will all sort itself out. This reminds me of the Fast = On database switch a lot of people are looking for when they are doing performance tuning. I am always amused by the sales tone of these articles (In fairness there are also some valid points in the article, but a lot of it is just boring sales pitch). In reality, the right tool only contributes about 5-10% to a successful DW/BI project. At the end of the day, the big three BI tool vendors (Oracle/Siebel, SAP/Business Objects, IBM/Cognos) offer pretty much the same functionality. Their offerings only differ in nuances.
Don’t get me wrong. Of course I appreciate the presence of Business Intelligence in the mainstream media. However, this just gets across the wrong message, sets the wrong expectations, and eventually is damaging to the BI industry.
So then, what is actually important for a successful BI project?
From a technical point of view, this is without a shadow of a doubt the Data Warehouse. You may get away without a Data Warehouse in small pilot type BI projects that involve low volumes of data and just run against one data source. Howver, once it gets just a little bit more complex the BI tools run into all sorts of issues. Think of performance, data integration, data quality, concurrency, complexity (did you ever have to deal with the spider-web like chaos of a Business Objects Universe run against an OLTP system?), no or limited historical data, non-replicable data queries, performance (did we have that before?), and performance (again). Also tell me, how are you going to report against a recursive hierarchical relationship in a transactional system? Not easy, not easy.
In the words of Ralph Kimball (one of the founders of modern data warehouses):
“Periodically, there’s distracting noise in the industry about the feasibility of bypassing dimensional modeling and data warehouse databases to simply query and analyze operational data directly. Vendors promise magical middleware that hides the complexity of the source system so that business users can theoretically query the system […] eliminating the costly and time consuming extract, transformation, and load processing. Though middleware may be able to mask the underlying data structures, it doesn’t address the inherent problems surrounding the performance […].You may find middleware solutions are only capable of relatively light-on-thefly data transformations, leaving your data integration requirements dangling.”
Ralph Kimball, The Data Warehouse Lifecycle Toolkit, p.238.
From a business point of view, the single most important criteria for a successful BI project is to align the BI initiative with the overall business strategy. BI has to support the core business processes, resulting either in cost reductions or increased profits.
“For any given company in any given industry, we should systematically evaluate its industry, strategy, and business design as a means of identifying potential BI opportunities”.
Nancy & Steve Williams, The profict impact of Business Intelligence.
Of course, there are a lot of other factors that determine the success of a BI project (skill sets, implementation methodology, training etc.). Get the above two wrong though, and you are out.