I've written several times about two seemingly divergent approaches to business intelligence: an enterprise strategy in which all elements of an organization's BI strategy are based on the same technological platform, and a workgroup approach in which different departments and divisions are allowed to use the BI tools of their choosing. Proponents of the first approach say it offers a more cohesive view of data that facilitates better decision-making and ultimately saves money over the long haul, while backers of the second say it yields a quicker ROI and allows users to select tools that best suit their needs.
With my naturally wishy-washy nature, I could see wisdom in both approaches. If only there was a way to have it both ways, I often thought. (As a working mom, you'd think I'd never mull this question, seeing as how trying to have it both ways is often a one-way ticket to Stressville.)
Yet that's exactly the approach BI Scorecard founder Cindi Howson, the author of "Successful Business Intelligence: Secrets to Making BI a Killer App," advocated when I interviewed her recently, and it makes a lot of sense. I asked Howson about it because her recent research shows a correlation between BI standardization and successful BI deployments.
Companies surveyed by Howson were roughly split on their approaches to BI, with 51 percent using an enterprise-wide approach and 49 percent a departmental approach. Twenty-nine percent of respondents using an enterprise-wide approach consider their BI initiatives very successful, vs. 12 percent of respondents whose companies take a departmental approach to BI. Six percent of those using a departmental approach said BI initiatives were a failure, vs. 1 percent of those using an enterprise-wide approach. The numbers who considered BI moderately successful were roughly the same.
The key to using a blended approach to BI, said Howson, is taking an enterprise approach to things that are common and won't change frequently, such as hardware, software, master data, procedures and best practices, while still allowing business units to build their own applications and their own reports to access departmental-specific data. She told me:
When I talk about standardization, the first thing I emphasize to clients is they need to understand their different user requirements and user segments. It's perfectly acceptable to use software-as-a-service as a tool your statisticians might use, but maybe another tool for your business offering. That's a perfectly acceptable approach. I call (a platform) a predominant standard rather than an exclusive standard. I think you always have to go back to what are the business requirements, what is the time to value, and look at it from that perspective.