Recently I found myself delivering a pilot project for one of our clients that is particularly interested in learning about Power Pivot as an analyst tool. After teaching some of the fundamental data modeling and DAX concepts, we moved onto hands-on report prototyping using real data.
Having worked as a Business Intelligence professional for years, I am fully aware of the need for a robust data model to power business reports. Without a properly designed model, reporting becomes more complex than it needs to, and less flexible than the business sponsors would like.
However, data modeling work can be tedious and require a particular – almost scientific – interest on the topic of data management. Professional data modelers spend their days looking at information structures, categorizing their attributes, classifying relationships among datasets and translating logical models into physical database schemas. You can hear many of them saying, “I just love data!”
The attitude needed to enjoy data modeling work can be learned, but most of the time it requires time, experience and appreciation of abstract concepts not directly related to any particular industry or set of data. It also implies an ability to be at peace with complexity and not become frustrated with particular data annoyances: low quality (bad data), improperly defined keys (bad relationships) as well as measure context (for aggregation accuracy).
The pilot project in question confirmed to me – yet again – that data analysts thrive on the capability to slice and dice data, as well as develop compelling storylines in regard to particular trends in need of exploration. This is what BI professionals generally refer to the User Interface (UI) work, which is typically devoted to the creation of charts and grids displaying data data.
Understanding the difference between data modeling vs. UI work is key when deciding whether a Self Service BI repository is of strategic importance. Self Service BI repositories remove the need for analysts to focus on the tedious task of data modeling on their own, and avoids the issue of creating redundant models to satisfy similar reporting needs. On my next blog entry, I will go into more detail about the strategic advantage of such repositories.