What is ‘real’ analytics?

A member of our MBA Program Task Force was talking about recent alums who had been successful on the job market, and early in her discussion, she cited the students “who had, you know, taken Vijay’s classes.” This did seem a little weird – my classes have course numbers and names, after all – but we were all in the midst of a very busy semester, and so I happily let it go.

A couple of weeks later, when an MBA staff person came by my office to propose adding a section of one of my MBA electives, she mentioned the great demand for “classes in my area.” I suggested that we simplify things by just referring to them as analytics courses (while my department’s name has changed almost annually, the word “analytics” has always been part of it). She responded equivocally, and looked terribly uncomfortable doing so.
Then, just before the holidays, I arrived a few minutes late to a meeting of the Graduate Programs Committee (I was giving a final exam that ran slightly over), expecting to present my proposal for a new MBA course in data mining. However, as I organized my handouts, a colleague seated nearby informed me with a chuckle that my new “non-analytics” course had already been approved.

I wondered: “Why all this weird verbal tap dancing?”

Well, after some digging around, I got an answer, though it was not a very satisfying one. During the last academic year, my school had launched a new Master of Science in Analytics (“MSAN”) program. The administrator who owned the program had apparently sought to differentiate the content of his program by explaining to anyone who would listen that the courses that we teach to MBAs are not “real” analytics courses, since these classes do not require any computer programming (outside of the Excel environment, which is viewed pejoratively by techies) and do not delve deeply into the algorithmic details behind techniques such as optimization, regression or cluster analysis.

This is just ridiculous.

First of all, in this kind of rhetorical response, one is required by custom to provide a definition, and mine comes from Davenport and Harris’ book, “Competing on Analytics”: Analytics, they state, is “the extensive use of data, statistical and quantitative analysis, explanatory and predictive models, and fact-based management to drive decisions and actions.” Based on this definition, it is clear that the skills needed for successful analytics professionals are both broad and deep. George Roumeliotis, an analytics leader at Intuit, believes that a good data scientist needs to be a skilled business consultant who also has a broad array of technical skills for data management, analysis and modeling [1].

What this means is that preparations for a career in analytics should be built on a three-legged stool of computing skills (including the ability to gather, merge, clean and manage data), analytic capabilities (with a special emphasis on basic probability and statistics, data mining, dimensionality reduction methods and fundamentals of optimization) and business effectiveness skills (such as leadership, problem framing, teamwork, project management, communication skills and negotiation). Any academic program that purports to be focused on preparing students for a career in analytics must strive to address each of these three competencies in some meaningful way, though there are an infinite number of ways to combine each of these somewhat orthogonal vectors.

While I was thinking about all this, I came across a blog entry on Forbes.com entitled “Business Analytics Beyond BI: Rise of the MBAs” [2]. The author, John Furrier, is a tech industry veteran and the founder of the website SiliconAngle.com, which pays an awful lot of attention to analytics and Big Data [3]. Though this relatively short article covered a lot of ground, a handful of interconnected “money quotes” caught my eye:

1. “Every department within a company today is itching to apply data-driven systems to their workloads.” What he’s saying here – and what my business school colleagues are slowly starting to understand – is we’re moving toward a time when most professionals will have to be conversant in working with data and interpreting models. We will need to start expecting more of our MBAs in these areas, and to keep innovating to deliver it.

2. “Automation will empower the data scientist to empower everyone else at the company, and they’ll need the help of software.” Automating the data scientist role has been discussed ad nauseam [4], but Furrier has a slightly different take on it: Automation is essential so that these critical human resources can be better focused on changing managerial behaviors, rather than having so much of their time consumed with managing data.

3. “The role of the data scientist plays an important part in setting the tone for collaboration within an organization, as these multidisciplinary problem solvers will need to communicate clearly with each other, as well as every other department.” That is, if the more technically trained analytics professionals can’t work well with less technically trained professionals, an organization’s analytic capabilities will fall far short of their potential.

Back here at USF, my colleagues in the MSAN program have made the choice to emphasize the computational and statistical aspects of analytics, which as expected has led to incoming students and outgoing graduates who are suited for very technical roles. My MBA students, however, do not have either the programming skills needed to implement algorithms from the ground up or the inclination to acquire them. Instead, their focus is on the business rather than technology. Instead, I do my best to train them to think critically – and wherever possible, to utilize user-friendly tools, which will be flooding the market for years to come – to address a variety of data- and model-driven business problems, while also working through data quality and management issues as needed. Given the well-documented talent shortages, it is not surprising that both groups are finding good (though very different) opportunities in today’s marketplace.

But let’s be clear: Both of these types of programs (and both of these types of students) are just as worthy of the term “analytics.” And in the future, we can expect that these folks will be working closely together on all sorts of things.

Notes and Reference

1. For more from George’s view on what makes a good data scientist, check out http://online-behavior.com/emetrics/marketing-metrics-intuit
2. www.forbes.com/sites/siliconangle/2013/12/10/big-data-beyond-bi-rise-of-the-mbas/
3. http://siliconangle.com/?s=big+data 
4. For example, see http://www.allanalytics.com/author.asp?section_id=1408&doc_id=251420, http://www.forbes.com/sites/gilpress/2012/08/31/the-data-scientist-will-be-replaced-by-tools/, and http://smartdatacollective.com/radhikaatemcien/111596/data-scientist-scarcity-automation-answer.