Convoluted Analytics — Part 2
It Is NOT All About Outcomes
It is all about outcomes. We hear it all the time. It has to be true at this point doesn’t it? Amateur analysts certainly believe it. They are so often concerned with the whats and the whys. The problem is that analytics has always been about The How.
This often seems contrary to the wisdom of business. We are told that ability and skill, process and effort, are secondary to outcomes. We must measure ourselves by outcomes we are told.
Surely, given how much importance is placed on measurement in analytics — it too must be concerned with outcomes. Right?
Not really. Science and analytics are concerned with the How. Outcomes are only useful as a benchmark for accountability. Real learning and real science require feedback. Feedback is a process, not a destination. Perhaps someone should put that on a successories poster?
But wait, isn’t this an appeal to analyze causality?
Touche’ … well more like passe’. It is true that proving causality is often a lost cause. I tell analysts that if they want to be focused on why, they should have been philosophers. So how do I now say that outcomes are not the right focus and causality matters? Pay attention to the words… actually the verbs… or might I offer the process… more scientifically, the relationship.
The scientific method, which is at the heart of analytics, as much as any other science, specifies a relationship. Too often, amateur analysts gloss over the relationship part of this equation. It is not enough that variable A increases as variable B does. That is only correlation. There must be a causal link — NOT a causal focus. We will rarely be able to prove causality, but we can’t ignore it either.
When we analyze the process, we learn the relationship between various things. This allows us to make predictions and, better still, to be prescriptive. It is the cornerstone of modeling. Those focused on outcomes tend to find their predictions, their models, and their prescriptions falling woefully short. Oddly, this is the one time they opt to ignore outcomes.
So while focusing only on outcomes is likely to lead to bad practice. If you can’t predict an outcome, your practice is bad.
If that statement is too convoluted, you may not be ready for the world of analytics, yet. The fact is — analytics tends to be pretty damn convoluted. Understanding the nuance of outcomes is just one of many examples.
Thanks for reading. Stay safe. More importantly, stay informed. For more on Convoluted Analytics consider:
For a chance to further your experience and analytic education:
