The Mysteries of Predicative Analytics

The two things that are most surprising about the burgeoning field of predicative analytics are: (1) the approaches people are taking to it and their varied efficacy; and, (2) the degree to which people appear to recognize its conceptual value; however, appear unsure how to use it or proceed.

Predicative analytics is the promise to do nothing less than predict the future.  And, the source of motivation in corporate executive suites appears to be largely based on envy.  In other words, executive read about technology juggernauts like Google or Facebook exploiting data and want what appears to be magical insights too for their own benefits.  This digital version of “keeping up with the Jones’” appears to be accelerated, like fuel on a fire, by management consultants from prestigious institutions, like Andrew McAfee and Erik Brynjolfsson of MIT reporting corollaries of 5-6% higher profits for those companies that have any “Big Data” and advanced analytics initiative.  (Dominic Barton, 2012)

Yet most companies, according to industry experts, remain either lacking in confidence, unsure how to move forward, or both.  While some of these experts opine the source of delay is buyers’ remorse from historically expensive and underperforming “latest and greatest” technologies like enterprise resource planning (ERP), data warehousing, or customer relationship management (CRM), my impression is that the answer is found in a 13^th century English Franciscan monk known today as Occam, or more specifically Occam’s Razor.  Simply put, it states that the simplest explanation has the highest probability of being correct.  In this case, people understand the idea of predicative analytics conceptually, but know neither the details of what it entails nor the understanding as to how to employ it.

Predicative analytics is an inter-disciplinary field combining mathematics, computer science, and subject-matter expertise.  The good news is that means it attracts scholars and researchers from three different fields; the bad news is that unless someone has knowledge in all three, they can only design things they can’t build, build things they can’t design, not design and build things that are useless or don’t provide a solution because they lacked the subject-matter expertise to know the problems or right questions to ask.

The second surprise can be seen by how hard it is, even at distinguished and premier universities, to find a concise summary of what predicative analytics entails.  As such, one can learn more from an hour on Wikipedia than conceptual lectures about the future potential of the field.  For example, one could say there are three major models, x number of software providers and what each does, and 18 major algorithms based on statistics, inference, linear regression and machine learning. (Finlay, 2014)

The second surprising problem feeds the first.  That is, because of a scarcity of primers explaining predicative analytics in a way that learners know about architectural, product, and algorithm choices, they are unable to easily expand on their source field knowledge to the other elements of this interdisciplinary field.  Finally, because there is such potential value in employing the knowledge competitively, there is little motive to share and make this specialized knowledge available, and instead, great motives to create an intellectual oligopoly to prevent competitors from prevailing.

Works Cited

Dominic Barton, D. C. (2012, October). Making Advanced Analytics Work for You. Harvard Business Review, pp. https://hbr.org/2012/10/making-advanced-analytics-work-for-you. Retrieved from Slideshare.com: http://www.slideshare.net/mitchki/predictive-analytics-context-and-use-cases

Finlay, S. (2014). Predictive Analytics, Data Mining and Big Data. Myths, Misconceptions and Methods. Palgrave Macmillan.