Visual Analytics – Just a Dumb User Interface

2272

Is visual analytics the final victory of form over content? For sure the business world has fallen in love with pretty graphics and reassuring dashboards, although I personally tire of people posting their latest chart showing the distribution of sharks around the globe, or net migration of people, as if something significant had been achieved. And what none of these pretty visuals can do is tell you what sharks and migrants will be doing next year. If analysis provides no clue of future behaviors it is pretty much useless, particularly in a business context.

It might be useful to put visual analytics into some kind of context. The visual interface provided by the ocean of visual analytics platforms is something akin to Microsoft Windows – a pretty interface that hides a less than pretty operating system beneath it. No operating system, and the graphical user interface (GUI) is wholly meaningless. But not so in visual analytics. Somehow we have come to believe that the picture is the analysis. This is a deeply flawed approach for several reasons. Here are the two most important:

  • Since we are pattern seeking creatures we find these patterns where in reality absolutely none exist. Many graphs and charts are like a Rorschach test – random noise we interpret as meaningful. Numerous books have been written on this topic. Fooled by Randomness by Taleb and The Drunkard’s Walk by Leonard Mlodinow are two of the more readable. Many of those trends, cycles and clusters we find in our data are nothing but random noise dressed up to look respectable – and then we make decisions based on them.
  • A recent article titled The Dark Side of Business Intelligence speaks the unspeakable. It’s the one article I wish I had the courage to write. In a word it says that business managers fudge analysis to make their performance look good, and are frightened to death of machine generated impartial analysis. But I think many business managers already know this.

So we are using visual analytics, which is nothing more than a user interface, as if it was the whole story. Well it isn’t. Much more important is the analytical engines that sit under this user interface. These might be bespoke or default standards such as R and Python. Either way we need machine assistance to generate meaningful analysis – just looking at a graph doesn’t cut it, and particularly if we are making decisions based on the analysis. Below the computation layer is the data layer. As a minimum this allows us to connect to various data sources, but more importantly it should allow us to clean our data, transform it and combine multiple sources in a meaningful way. This is a whole universe of its own and is sadly ignored by many suppliers of visual analytics platforms.

As the future rushes toward us we find ourselves perhaps just four or five years away from AI driven BI. BeyondCore was a taster of this, and was quickly snapped up by Salesforce. When the machine is producing the analysis I’m afraid there will not be much room for fudging the numbers and using misleading visuals. And as mentioned earlier, analysis is useless unless it gives us some insight into the future. Machine learning does this very well, and so the visual interface will become simply a mechanism for presenting analysis to people, and not a platform in its own right.

So to answer the opening question. Visual analytics certainly has come very close to being the final victory of form over content – but expect things to reverse fairly quickly. Every extreme implies its opposite!