Self-service analytics: data-driven challenges and solutions
December 09, 2016 | Bill Saltmarsh
Self-service analytics is an ideal that is widely shared and promoted in the modern workplace, especially for those organizations that pride themselves on being data-driven. Self-service analytics can be described as an environment where business users are enabled and encouraged to directly access data, in order to derive insight from business information as quickly and efficiently as possible.
However, the process to achieve true self-service is fraught with obstacles and unforeseen complexities. If I were to characterize the majority of these problems, then I’d say that we, as data professionals, are underestimating the level of difficulty in transforming a data-curious business user into someone who can competently access, analyze and consume that data.
Overcoming the challenges
From where I sit, there are two primary options in overcoming these difficulties.The first is to spend an extraordinary amount of time studying the behavior of your business users, understanding their needs and then training them on both an abstract level as well as a contextual level, so that their skill set is improved enough to become truly self-sufficient in their role.
The other primary alternative is to acquire and/or build tools that drastically simplify the process of acquiring, cleaning and analyzing data. As someone who continually works to develop data products that are increasingly accessible, I constantly ask myself the question: Is this easy to use? Unfortunately, tools that are easy to use are few and far between.
For example, Tableau prides itself on its ability to convert users into analysts. However, once you get past the initial inspiration provoked by learning the potential of Tableau (or any other visualization app), you’re left with the difficult and technical work of creating actual meaning from disparate data. This is not easy work, even if it is easy to click and drag a field to make a bar chart.
Speaking specifically about Tableau, one common question from new users is: How do I compare the performance of a singular dimension against the performance of the larger category that dimension falls within? This is such a common and logical question in data analysis, yet getting to the answer is not easy. The construction of a level of detail (LOD) expression, which is one solution to this question, is not easy to accomplish. And more than that, the concept is even more difficult to grasp than the mechanics, especially for someone with minimal data skills and experience.
But on the whole, Tableau does make things easier for business users. The complexities that are overcome by using these tools like it cannot be overstated. Tableau and other software like it open up possibilities to non-technical business users in ways that were previously unattainable.
That being said, there is still a vast amount of potential for improvement. In tan ideal world, data import is clean and minimal. However, possible pitfalls abound in throughout the process. The challenge for data engineers, analysts and stewards is to continue to perfect the tools that enable data democratization. Never stop asking: Is this easy to use?