A Decision Science Lens:
Descriptive Analytics
Another Bad Wrap For Education
Let’s face it. Descriptive Analytics was coined, or at least popularized, as a catch-all for those aspects of analytics that were considered beneath other concepts like Diagnostic, Predictive, and Prescriptive. That is not to say that the label is not valuable, but it is a starting point for why it is so pejorative.
While the internet overflows with diagrams and infographics describing the other areas of analytics, descriptive analytics is a desert. What is more, the two areas of descriptive analytics that do have buzz are rarely listed as components. Data Visualization and Business Intelligence (aka reporting) are not often connected with descriptive analytics. That connection just doesn’t seem to be that needed. The whole point of labeling this area of analytics, seems to be in the interest of giving it a “bad wrap”.
How Can You Say That?
You don’t have to look too hard to find one of these lovely graphs. It also doesn’t take much interpretation to see that this visual implies that descriptive analytics is simple and low value. Yes, that is relative to the other areas of analytics… but honestly, it is just flat out wrong!
Complexity here (other graphs use sophistication or difficulty) is a purely academic and algorithmic view point. There are so many other dimensions that could be weighted into that label. Dimensions where descriptive processes would create staggering complexity — especially given how many predictive processes are simple regressions and how many prescriptive processes are just decision trees.
But the Decision Scientist in me is far more disappointed by the Added-Value Contribution component of this graph. Whoever built the first of these visuals clearly had no sense of how “real” analytics creates value in an organization. Let’s use the rest of this article to better define the role of descriptive analytics in value generation.
The Leverage of Large Denominators OR True Functional Value
Related to the law of large numbers (or at least some of them, there seem to be a couple) is the leverage of a large denominator. The bigger the denominator, the more a smaller rate can produce a major outcome (numerator). It is why small optimizations in massive portfolios can produce meaningful financial impacts. It is the main reason banks, pharma, wall street, and lately social media have been able to invest heavily in data and analytics. Small lifts are very, very meaningful.
Now think about how any medium to large organization functions. Each is composed of a dozen or more departments each making any number of daily decisions. Typical analytic support influences only a small fraction of these decisions — typically the largest or most complex ones. While prioritizing those decisions for analytic investment makes sense, it overlooks two important facts.
- the sum total of the smaller decisions is likely larger than the fractional number of big ones
- descriptive analytics can and should influence all of them
Done right -
Descriptive Analytics is an educational framework for the decision-makers in your organization. It influences every decision your organization makes.
Stated differently — descriptive analytics are the processes which inform us who our customer is, what they are doing, and how that is changing. Descriptive analytics includes the mechanisms for tracking the things we do and the impacts they have. Descriptive analytics provides clarity, transparency, and measurement to everything that makes an organization function. So how valuable is that?
Isn’t that cliche?
Not really — but these are…
Descriptive analytics is not a bed of roses. It is often not done right and even when it is — it is very indirect and organic in its influence. This makes it hard to measure the value it produces, thus fueling all those bad models above.
If too many parts of your organization operate subrosa, descriptive analytics will likely not be as influential as it otherwise could. It requires transparency to produce better transparency. I can’t describe what I can’t see or don’t even know exists.
Descriptive analytics by any other name… well, lets just say that when done really well — descriptive analytics is the base that these other disciplines are built on. You can’t predict or prescribe what you can’t describe. Many organizations falsely credit “higher order” analytics, not realizing their success is far more functional until it is too late.
As for our coming AI overlords…
They have absolutely nothing to do with this particular topic. Descriptive analytics is not really in the cross hairs of machine learning or artificial intelligence right now. Perhaps when the world finally pivots to Augmented Intelligence?
No worries. AI is sure to come back in future articles and we will get to Deep Learning sometime soon. For now, recognize the real value and complexity of descriptive analytics done right. Understand how powerful it is to improving the decision-making ability of any organization. And as always — thanks for reading!
