Let me start by applauding my colleagues in the academic community for their responsiveness to the demands of the market (not something I do frequently). The academic community should be applauded for pivoting university programs in meaningful and unprecedented ways to respond to the explosive demands of the private and public sectors for deep analytical talent (1). Since 2007, the number of master’s-level programs in “data science” or “analytics” has exploded from 0 to more than 100 (2). There is also a budding collection of Ph.D. programs in analytics and data science, which will soon be producing newly minted academics to teach in these programs.
While most analytics program directors (myself included) would have liked to have had a few years to assemble and meet with an advisory board, research “best practices,” test curricular options, develop a budget, conduct a series of national searches for faculty with the unique set of skills and established research streams relevant to the envisioned program – the reality is that most program directors were probably given the equivalent of one academic semester to develop a program using existing resources and faculty. (I can hear “Scotty” from Star Trek with his Scottish accent: I can’t work miracles, Captain!).
The collective results of these academic miracle-workers has become a fascinating constellation of programs across the country, which illustrates the diverse academic ecosystems that these program directors had to operate within to get these programs up and running:
In this constellation, the larger nodes represent the colleges that house these programs within the individual universities.
There is no question that there is a rich and textured mosaic of where these data science and analytics programs are housed. However, a second look at this constellation of programs reveals an interesting organizing dynamic – some programs are housed in an “application” college like Healthcare/Medicine, Engineering or Business – while other programs are “application-neutral” because they are interdisciplinary or are located in a college like Mathematics/Statistics, Computer Science or even Data Science. The location of these programs has implications beyond just organizational convenience – most of these programs also exhibit a “vertical” or a “horizontal” orientation, which is reflective not just of their location within the university, but importantly reflective of the prioritization these programs will place on the different concepts inherent within this nascent academic discipline.
It’s important to note, at this stage, that neither orientation is “wrong” or even necessarily better nor worse – these programs were developed using the resources that were available in response to the needs of the employment ecosystem they serve. Select any of these programs at random, and you will likely see common traits at varying levels of depth – applied statistics, basic computer science and domain application.
It’s the “varying levels of depth” across these common traits that differentiate horizontal vs. vertical programs.
Although graduate students may be suspicious, the number of credit hours that program directors can assign to a program is finite – as a result, these directors had to make tradeoffs regarding courses within programs. “Vertical” programs will typically trade off heavy computational courses like mathematics, statistics and programming in favor of specialized domain knowledge while “horizontal” programs typically do the opposite.
Consider an example of a typical “vertical” two-year M.S. in Analytics program in a business school. These programs will provide students with deep learnings in concepts and analytical case applications related to marketing, customer segmentation, search engine optimization, A/B testing, etc. Importantly, they are taught to “tell the story,” create dashboards and visualizations, interpret results and output – they can explain results in the context of the original business problem. They can typically ask “what if” questions related to the domain in question. The downside to this orientation is that these programs rarely integrate any depth of mathematics, the breadth of statistical techniques taught is limited, and they typically have limited programming requirements and limited exposure to analytical practices from other disciplines.
Alternatively, consider an example of a typical “horizontal” two-year M.S. in Analytics program in a college of science or computing – or one that is interdisciplinary and housed across multiple colleges. These programs typically provide students with deep learnings in concepts related to machine learning, neural networks, natural language processing, data mining, as well as the mathematics that are foundational to these concepts. They typically will require students to write their own algorithms and develop familiarity with several different types of programming languages. They typically work with data from a wide variety of application domains and work in interdisciplinary teams to solve problems. The downside to this orientation is that students from these programs frequently have limited domain experience and have less exposure to applying and explaining results in context.
A simplistic summary of these two different orientations is that vertical programs tend to emphasize “downstream” tasks on the data science continuum, while horizontal programs are more likely emphasize “upstream” tasks on the data science continuum:
As a side note, Ph.D. programs in Analytics and Data Science – which have the benefit of time (double or triple the number of credit hours associated with a master’s program) – can emphasize both the vertical and the horizontal.
Students in “vertical” programs may want to consider taking “horizontal” electives (e.g., computational mathematics, scripting languages, algorithm development) while students in “horizontal” programs may want to consider taking “vertical” electives (e.g., health policy, macroeconomics, business strategy).
So, the question that a prospective applicant should ask when considering an analytics program is – Am I looking for a “vertical” or a “horizontal” program? The good news is that there are now strong program options, which accommodate both preferences.
- Program information from college websites collected September, 2016