How can we leverage agile business intelligence (BI) to help with our BI architecture?
You should logically split your BI architecture into two categories: data integration processes and BI applications. To be most successful, an enterprise should then manage the design, development and deployment of those two categories differently.
Data integration involves processes such as gathering business and data requirements, data modeling, data profiling, cleansing and conforming data, and creating extract, transform and load (ETL) routines. Although an enterprise should follow an incremental and iterative approach to that work, data integration involving more than one data source and business group typically doesn’t lend itself to a formal agile approach.
Agile techniques are a great way to foster development of BI applications, such as dashboards, scorecards, reports and analytics applications. This is especially true when the data is already in place and the initial set of business requirements has been defined and prioritized. At that point, quick development (or prototyping) of reports, analytics and dashboards enables IT developers to show business users what they’re building and quickly gather feedback on the designs. Agile BI techniques can significantly reduce the likelihood that IT will develop something the business cannot use. In addition, it’s very common for business users to refine requirements or identify new ones that they forgot about or assumed that the IT staff knew about. An agile approach can make it easier to accommodate such changes.
The key management issue is to timebox the development tasks for each BI application. Otherwise, the project can get into an agile paralysis loop (similar to analysis paralysis, only in the development process.)
This was first published in February 2011