It was Albert Einstein who said things should be as simple as possible, but no simpler. If simple analytics do everything you need, then the so called advanced analytics are clearly unnecessary. By simple analytics we mean paginated reports, charts, graphs, and in fact anything that relies wholly on human interpretation. Right now we are in the middle of a fascination with all things visual – charts, graphs, maps, dials, gauges – etc etc. This is not advanced analytics, it is simply the presentation of data in forms that are more easily digested.
So what is advanced analytics? In a nutshell we can say that advanced analytics consists of methods and technologies that employ some form of intelligence. The interpretation is not wholly down to humans in other words. Maybe the simplest example is clustering. The basic idea is simple. We look for clusters of data which are similar in some way. A graph can only clearly display two dimensions, or three at a push. In the diagram above the data have been grouped into three dominant clusters, with some points being considered as not belonging to any cluster. In a business scenario this might represent customer spend against income – or some such thing. The clustering algorithm (usually an algorithm called k-means) has calculated the best clustering, and as such has displayed a certain level of intelligence. The alternative would be for someone to display a scatterplot and decide their own clusters. In two dimensions both approaches can be used. But what if you want to find clusters in four dimensions – or ten! So you might be interested in age of customer relationship, versus total spend, versus products purchased, versus promotion responses. You could do this manually if you enjoyed psychological torture, but it would be much easier just to throw the data at a clustering algorithm. And by-the-way, clustering is making its way into some of the popular visual analytics platforms – Tableau and Spotfire for example.
There are many, many forms of advanced analytics, but they will only be used when they become easy to use, and this is the challenge presented to most of the business analytics tools vendors. No one really wants to have to learn about logistic regression, Bayesian networks, k-means, support vector machines (sounds like a piece of construction equipment) – and so on. But as businesses ramp up their analytical efforts so they will need to adopt more advanced forms of analytics. This will not be the end of the story, because AI will make its appearance in our analytics much sooner than we think (look at PurePredictive for example) – and who (or what) will be doing the analysis then?