Sound Data Science

Just like a building should be sound and not collapse, a data product needs to be sound to be able to create business value. Soundness is where the science and the data meet. The soundness of a data product is defined by the validity and reliability of the analysis, which are well-established scientific principles as shown in Figure 2.3. (Anderson, C. (2015). Creating a Data-Driven Organization: Practical Advice from the Trenches. Sebastopol, CA: O'Reilly Media Inc) Soundness of data science also requires that the results are reproducible. Lastly, data, and the process of creating data products, need to be governed to assure beneficial outcomes.

The distinguishing difference between traditional forms of business analysis and data science is the systematic approach to solving problems. The key word in the term data science is thus not data, but science. Data science is only useful when the data answers a useful question, which is the science part of the process. (Caffo...