Abstraction in Visual Analytics
Often Computer Science students are given the example of a car when they learn object oriented programming. This example is based on the premise that you don't necessarily need to know the internals of a combustion engine to drive a car, the complex details here are abstracted from the driver. With the popularity and adoption of data science and analytics in every business vertical today and data storage becoming cheaper, the need for abstracted visual analytics tools [1] is on the rise.
Recently, I got the chance to play with Highcharts, an elegant framework for building charts through code. Frameworks as developed as these are feature packed, highly abstracted and beautifully documented. But this framework made me more aware of the features I had taken for granted while working with a visual analytics software built for "ease of use". Spotfire, my tool of choice, abstracted a lot more complexity than I had initially given it credit for . This realization would resonate with companies and individuals who want analytical insight as an essential value add to their core business.
The one skill crucial to navigating through the tools in this space is to have a data visualization vocabulary. Such knowledge would help you get through any product specific learning curve with relative ease. Things like markers, labels, trellis options, visual types, data types, aggregations, asynchronous drill downs etc., map to the same meaning universally. The major difference that stuck with me though was that of abstraction of complexity. For Example, while Highcharts enabled me to code to align multiple axes in a simple line chart or think about the way I wanted to name and label data points. I realized that a user might get overwhelmed with the nuances that are often taken care of as out of the box capabilities in enterprise software offerings. But then again the comparison was between tools that approach the same problem in two different ways 1)A code based developer friendly focus 2) A GUI based focus on rapid insights and ease of use.
While the first approach offers many options for customization, it needs experts well versed in both scripting/ visual frameworks as well as the business problem at hand. The second option enables a business user to help with a problem without expertise in visual analytics.
As a biased spectator, I just thought of the results analytics has bought to several businesses out there. Customers who don't necessarily have an army of data scientists and visualization experts but handle their core business with added insight and actionable information. Here are some of my favorites at TIBCO:
https://www.tibco.com/customers/mercedes-amg-petronas-motorsport
https://www.tibco.com/customers/broadreach-healthcare
https://www.tibco.com/customers/vestas
[1] In Addition to Visual Analytics, Abstraction in this field comes in many
forms. One good example could be the modular plug & play algorithms
by the company Algorithmia.
