Daniel Rubiolo blog about his view of Business Intelligence, with some great metrics on the components and percentage of resources dedicated to a BI solution.
The traditional definition of “data warehouse, or DW” from late ‘70s/early ‘80s involved the end-to-end solution for business users… from extracting data and aggregating it into special data models, the queries, all thru the applications/reports the users interacted with to consume that data.
According to Daniel, a BI project is 70% getting the data and 30% showing it off. I would suggest it fits with the 80/20 rule in most cases. 80% of the time and effort in a BI project would be sourcing the data and gathering requirements, designing, developing, testing, and deploying a solution. 20% of the time would be adjusting and presenting reports.
This is probably why PowerPivot is such a favourite for power users, since they don’t necessarily require outside IT resources to build their models. It becomes a 50/50 game with a faster ROI. However, scalability and maintainability of a solution that resides on a user’s desktop remains in question. The “hit by a bus” rule makes desktop-based custom models a potentially risky solution for a business