Just in case you haven’t noticed we are in the midst of a full blown tech feeding frenzy. Most forms of data analysis are being gobbled up by businesses around the world as if there was no tomorrow. To illustrate the point, Tableau generated as much income last year as it did in the whole of its previous history. This is the way it is with technology – adolescents raving about the latest cool musical acts have nothing on the waves of fashion driven technology spending that can sweep the globe.
Now every party needs a skunk – a role I am willing to adopt. The business analytics party is in full swing and people are doing things they will regret when a new day dawns, and the hangovers weigh heavily. So lets get to some specifics.
We’ll start with data visualisation and business intelligence. The primary focus for this activity is to look in the rear view mirror and establish how various aspects of a business have performed. Sales of a certain product in a given region may have been poor last quarter. Why was that, and most importantly – was it significant? Was it more than some variation that was within the deviations that might be expected. Most people creating graphs and charts will not know the answer to this question, and the tools they are using will not provide it either. While we have become addicted to animation, 3D, designer colour schemes and any other feature that will enhance the eye-candy factor, the tools we use will not tell us whether the data is exceptional or within range. It is quite possible that six months of increasing sales is nothing other than a random fluke. Business life is driven by decisions that are based on nothing other than noise within the data – read The Drunkard’s Walk by Mlodinow if you want an ocean of examples. Dumb data visualisation and business intelligence tools tell you almost nothing if there is no indication of what is exceptional, and what is not. We shouldn’t have to be statisticians to establish the likelihood that that nicely rising trend could easily be nothing more than random noise. We wait for the obsession with eye candy to fade and for the BI suppliers to add some intelligence to their offerings.
The next topic for the skunk treatment is big data, and specifically big data analytics. There is an unspoken assumption that more data is better data. This simply is not true in some cases. The machine learning techniques that are applied to big data to find predictive patterns will always do what they set out to do – find patterns. Whether these can be trusted and prove useful is a complex issue, and even the experts in this field (Michael Jordan for example) state emphatically that many of the problems have not been addressed yet – and may not be for some time. And so we have businesses around the world gobbling up big data technologies (which are extraordinarily complex and immature), applying machine learning algorithms, and implementing the predictive models they create without a full understanding of the risks. Big data that has many attributes is particularly prone to throw out patterns that are nothing more than accidents of the data. As the number of attributes grows so the combinations of those attributes grows exponentially, and the machine learning algorithms have a great time throwing out patterns that have no meaning whatsoever.
The whole domain of business analytics is terribly immature and in many instances will be doing more harm than good. There are some simple rules of thumb however that may help.
- People who really know the business (domain experts) will generally know whether analysis makes sense.
- Keep it simple, and don’t thrown every bit of data, and every conceivable analytic method into your analysis. Understand which variables might be important, and do not allow your analytical tools to create a pot-pourri of analytical models that simply do not make sense.
- There is no substitute for experience. Without it you may just end up with the dumbest smart analytics in your industry.
Oh – as for the other fairy tales. How about the one where the fox tells the crow perched in a tree, with a delicious piece of cheese in his beak, how lovely his singing voice is. The crow immediately starts to sing (craw, craw craw) – the cheese dropping from its beak, and the fox runs off with lunch. Caveat emptor.