Press "Enter" to skip to content

The Importance of Data Shaping

Paul Turley shapes the youth of data:

Power BI is a new tool and dimensional modeling is an old idea. One of the challenges is that, like other modern self-service analytics products on the market, Power BI doesn’t force self-service data jockeys to transform their data before reporting with it. If you want to import a big, wide spreadsheet full of numbers and create charts in a Power BI report, knock yourself out. But, the solution won’t scale and you will inevitably run into walls when you try to make future enhancements. Similar problems arise from importing many tables from different sources and transactional systems. Several tables all chained together with creative mashups and relationships present their own set of problems. The first iteration of such an effort is usually a valuable discovery method and learning experience. Great… treat it as such; take notes, make note of the good parts and then throw it away and start over! In Fredrick Brooks’ “The Mythical Man Month“, he cites that for most engineering projects, the first six attempts should be abandoned before the team will be prepared to start over and complete the work successfully. He was a chemical engineer before working for IBM; and hopefully, our methods in the data engineering business are more effective then his 6-to-1 rule. But, this makes the case the prototypes and proof-of-concept projects are a critical part of the learning path.

The tools don’t make the rules.

Unless you’re talking about the lambda architecture, in which case that’s kind of accurate. But we’re not talking about that here.