There is a certain topic that comes up a lot. In fact, I responded to a LinkedIn discussion on the topic just a few days ago. While there are many variations, the core of the question always gets back to this concept: “Is it a good idea to let non-analysts utilize friendly user interfaces to build models or do deep analytics?”
- Advertisement -
There is a certain topic that comes up a lot. In fact, I responded to a LinkedIn discussion on the topic just a few days ago. While there are many variations, the core of the question always gets back to this concept: “Is it a good idea to let non-analysts utilize friendly user interfaces to build models or do deep analytics?”
In recent years, analytics tools have moved from a purely programmatic interface to more of a “point and click” or graphical user interface (GUI) environment. While many people can’t program, the argument is that they are now enabled to do the work through the friendly GUI.
As an analyst myself, the question makes me think of a quote from Albert Einstein, “Technology is like an ax in the hands of a pathological criminal”. Should we be excited about a person with no meaningful analytics skill, experience, or training gaining the ability to use a GUI to do advanced analytics?
The answer is “Absolutely not!” I know some will vehemently disagree, but I’ll explain why.
There is a big misconception that a tool that has an interface which is easy to navigate means that it is easy to navigate it correctly. Setting people, who do not know what they are doing, loose on any tool of any type is risky. Lets face it: there is a lot more behind an analysis than an algorithm just cranking through a bunch of data.
- Advertisement -
Have you framed the problem correctly? Are you predicting the right behavior? Do you have the best set of independent variables? Are you able to identify when things “just don’t look right” in the data?
The bottom line is that there is a lot that goes into an analytics effort beyond just pointing to data and saying, “Go!” If you could magically ensure that the novice had the right data, for the right problem, with the right algorithm, then I would grant you the damage would be minimized. However, without that perfect set up you just might be getting a great answer to the wrong problem or, even worse, the wrong answer to the right problem. In real life, the decisions and work that go into an analysis are complex and require an understanding of what you are doing and why.
How many marketing or advertising departments would turn me loose to generate Creative just because the tools they have make it easy to generate graphics?
What accounting department would let me close the books just because there are some menus to guide me through the process?
And God help anyone within a hundred feet of any neighbor that would let me cut down a huge oak tree for them because my wife bought me a professional grade STIHL chainsaw for Father’s day!
- Advertisement -
Why is analytics any different? For some reason, people often consider allowing a novice to produce analytics they don’t understand just because an interface makes it “easy” or data sets can be generated in a more automated fashion. While laughing at the idea of me generating effective creative or cutting down that oak tree, they embrace the vision of “Bob the marketing guy” generating effective and accurate analytics.
I am a huge supporter of using technology to expand individual capabilities, help drive more accurate and consistent decisions across a larger group, and drive greater productivity. However, I guess I am just one of those guys that don’t equate technology with magic. This leads me to think of another quote, and I wish I knew who said it, “Technology… is a queer thing. It brings you great gifts with one hand, but can stab you in the back with the other.”
Beware of ”Bob the marketing guy” and those who are sponsoring his forays into deep analytics!