Please keep in mind that this list is not an exhaustive list on everything that you need to consider before you start your BI implementation.  Below are simply three lessons I feel would be helpful to think about before you begin your implementation.

1)     Rapid prototype development does not mean you don’t have to get the application production ready.  A lot of times applications are developed very quickly utilizing the new data discovery tools; a simple back and forth between developers and business users is all that is needed.  Developers will implement features, test, get feedback and repeat the process.  At some point, everyone will decide that the application is ready to release to the general audience for public consumption.   In my experience, it is important for everyone to remember just because the exchange between developer and user was rapid and informal, that does not mean the application should not use the same standards as all other applications before release to the enterprise.  Rapid development is very powerful, but when the application is ready to be consumed publicly, the proper enterprise standards should be followed.

2)     Most of the new Data Discovery tools support a very large amount of data, but that does not mean you need everything in the same Data model.  It’s great to be a one stop shop.  I’m sure a lot of organizations would love to shove everything into the same data model and have everyone use this one application for all of their business needs.  However, this causes way more problems than it’s worth.  I have encountered   performance, scalability and training problems.  The larger and more complex the data model, the more likely you are to experience performance problems.  The more data you have, the longer it will take to reload your application as well as retrieve the data you need.  The larger the app the more server and memory resources the application uses.  If the enterprise server is taking up a very large foot print of memory, then that is less memory available for users.  The less memory you have available, the more difficult it will be to scale.  Typically, in very large applications, the data model is complex, meaning it will be difficult for users to understand and require additional training.  In most cases, instead of creating one large data model, it would be best to solve for a specific business need.

3)      Training is still required.  There is a misconception that the new data discovery BI tools are so easy that the users will just hit the ground running.  In my experience, there are many levels of users.  There are uses who just want their data spoon fed to them and power users who want to get in, create their own objects, and do their own business discovery.  I think it’s very important to explain the nature of the software and the difference between business discovery and traditional reporting.  If you set the right expectations and prepare users with the right mind set, it will save you a lot of headaches down the road.  This is very difficult because it requires a culture change.  Business users are familiar with consuming data, but discovery will be new to them.  Training challenges I have experienced are with understanding the power as well as limitations of the data model and the difference between business discovery and reporting.  Data discovery requires a change in approach and it’s best to address this early and reinforce throughout the implementation.