Why the Analytics Maturity Model is Quite Honestly.......Garbage

Undoubtedly every eighteen months or so an interviewee, consultant, lecturer or vendor trots the analytics maturity model out to some unsuspecting member of leadership in hopes of providing a shocking glimpse into how immature analytics are at their organization. This tact is usually followed quickly by a sales pitch of how their product and/or services can push that leader’s enterprise to the panacea of predictive analytics.

Which is Bull$hit.

The fact is even the most rudimentary life form on the planet is using the full spectrum of analytics all of the time. As a human being, when you drive a bicycle, motorcycle or automobile you’re utilizing the entire scale of analytics at your disposal to make decisions. You predict that a car leaning into your lane may be turning without a signal; you evaluate and remember the most effective ways to and from work. The analytical evaluation that’s used every day in driving is not the practice of building out one mode of the model before moving onto the next; it is the maturation of a model through increasing experience.

Imagine if you needed to mature your descriptive and diagnostic analysis before ever taking the wheel. The cognitive evaluation would take decades. The analytical mindset grows by living the entire experience of a situation and most importantly evaluating the predictive model’s accuracy. If the only red light that you cross driving lasts for 4 minutes, your predictive model will assume all red lights last for that long. Only through the experience of more driving would you learn whether that is representative or an anomaly of the length of a red light. 

Anyone who’s driven with a 16-year-old (as I have with two of my own) realizes their ability to predict a situation lacks NOT because they didn’t study every potential situation while climbing their way from predictive to prescriptive analytics. They simply haven’t had enough experience to effectively work through the model and make good predictive decisions. They could study drivers ed for years and it would have little effect on the learning that is attained through real-world driving.

The analytical model isn’t linear by any means. That’s a sales pitch anyone hearing should be wary of. Those selling it know part of the security in the sale is that no one ever gets to the end and it’s an easy out to say that data is bad, the tools too immature or the business isn’t ready for it. 

The model is circular and matures as a whole. The best F1 drivers in the world are able to take their world of experience to predict and act on what they strongly assume will happen next. Tesla and others have broadened that circle to expedite and improve self-learning; quickening the expansion of the circle by testing model after model based on thousands of experiences. The best in analytics are those that listen to and work with those involved with the entire processes that they are studying. Using test cases and taking the time to laboriously run things from reflective to predictive isn’t sexy. But it is evolutionary and the way the model grows.

The next time someone comes selling you that map with a destination of AI and machine learning ask for just one use-case they’ve proven out. If they can’t come up with that they’re just selling you a guide to somewhere they’ve never been.