Why data analytics tools aren’t enough?
Business intelligence and data analytics tools promise to turn data into insights. These tool features varies from integrate data from various data sources, data exploration, cleaning and analysis as well as a vast set of visualization capabilities, including the creation of dashboards. However, true data-driven insights go beyond the purely presentation of data –descriptive and inferential analytics. They still requires the expert evaluation, especially at the beginning before any automation is implemented, and this is as valid for simple reporting as for building machine learning models. But particular for the later, as analysis skills and experience in dealing with data are crucial components of predictive modeling.
A simple example is here presented to illustrate the above, in particular, how even everyday popular website statistics might lead us to misleading conclusions.
The example goes like this…
According to Wikipedia, and as of the 23rd of August 2017, Despacito was the most viewed YouTube video. Despacito is moreover the first video to reach three billion views, followed by See You Again. The following most viewed videos on YouTube are Gangnam Style, Sorry and Uptown Funk.
Comparing apples and oranges…
This rank of the top-5 most viewed videos is based on total views since date of publication to YouTube. But because each of those videos was upload in different dates and so each of them is at a different maturity level, the total of views cannot be practically compared.
Comparing apples and apples….
By adjusting the data, the rank changes. Thus, by considering the views for equivalent periods of time and with equivalent distribution of views over the same period of time, as well as eliminating the bias associated to the Internet penetration rate, the rank changes.
The rank of the most YouTube viewed vides is, of course, just a trivial example with temporal validity. However, as irrelevant as this example is, it is a common mistake resulted from poorly analyzed problems. More importantly, these mistakes might lead us to the wrong decision –and the power of data analytics is in fact to lead us to better informed decisions.
www.visual-assessment.com / firstname.lastname@example.org