We've all read the articles and blogs. Many of us have experienced the issues directly -- the demand for deep analytical skills is outpacing the supply.
As evidence of this, in a period of economic slowdown, where we read that 50 percent of college graduates can't get a job, college graduates with degrees remotely aligned with applied analytics have multiple offers in advance of graduation. Academic training in applied (versus theoretical) statistics is helpful -- and mitigates some of this talent gap at the entry level. Nonetheless, we all know it's insufficient to meet the growing demand for what we now know as the data scientist.
A scan of offerings at universities across the country shows that none will result in a single degree program in data science. As a professor, I can see at least three reasons why this is so:
Universities historically are not massive ivory towers, but groups of towers called colleges or departments. Cross-discipline degrees aren't a strength of university curricula. Pity the engineering student who wants to minor in history or the student who wants to cobble together a degree in public policy of architecture and ecological science. Ask any student who has tried to cross colleges within a university to create a targeted degree -- it is the worst of bureaucracy, outdated registration systems, and academic-elite egos all wrapped in a Gordian knot. And yet this is exactly what data science is -- the intersection of mathematics, statistics, and computer science, combined with some potential area of content application like finance, biology, or sociology.
The data tsunami washing over all companies, not just data-driven ones, is a fairly recent phenomenon. Professors who teach statistics and computer science, in particular, are recognizing that the traditional skills we have comfortably taught for years or even decades don't work in this environment. Concepts like p-values to derive significance are meaningless when you have a billion rows of data. Professors are being challenged to teach skills that many of them don't have. Some are rising to the challenge, and some (think tweed jackets with leather elbow patches) are just hoping it all goes away, which, of course, it won't.
We need your data. Remember the datasets you saw in the classroom? They had 100 observations, three variables, and no missing values. Everything was significant in its raw form. Welcome to textbook data. We do our students an immense disservice by using this kind of dataset to teach analytics. But, believe it or not, in a sea of data, we are dying of thirst. Universities need massive, complex, unstructured, messy data with missing and (mis)coded values for use in the classroom. Ultimately, we can't teach data science skills without big-data.
I encourage people within the public and private sectors to partner with universities and in particular with professors who have recognized these issues and are trying to pivot their curricula to meet the needs of the marketplace. Sit on advisory boards. Provide real datasets (scrubbed as needed). Offer to speak in the classroom of your experiences with big-data -- everyone's story is the same, but different. Partnerships with universities in this area are particularly important and mutually beneficial. You can help us train your future data scientists.
Do you agree that change is needed if universities are to educate the data scientists that businesses increasingly need? Share your thoughts below.
Very useful suggestion about companies offering partnerships with universities. This will be a win-win situation for both. Companies will get future data scientists from the universities which will help them reduce data scientist shortages and universities will get both structured and unstructured big data to play with and executives who could come and share their experiences with data in practical life. Whats needed is a forum where universities can meet such companies so that both can discuss their needs.
I wont speak about US but certainly in many countries, Asian countries esp, we dont have universities offering such courses which help us become data scientists even if we want to. A tech geek college student who is doing majors in applied statistics may go towards that path however it is not yet close enough. May be the universities need to not only introduce the course but also increase the awareness that such a course exists. It wont be easy.
" The fact is that analytics, like technology, does not exist in a vaccuum. It is a powerful tool for all disciplines."
@mnorth Excellent point and I couldn't agree more ! I sometimes feel the focus is too narrowly focused on business needs, but as you mention analytics is used in every area of society.
And thank you Jennifer for exposing some issues that stand in the way of effective training of future Data Scientist, I am not sure where to start - I am sure there will be much fine tuning of curriculum and approach for years to come.
@Cordell, somebody sure pulled the wool over your eyes!
But seriously, you raise an interesting point about developing something that was practically meaningful but not statistically meaningful. If something is not statistically meaningful is it OK to put it to practical use?
@bulk: Unfortunately, I tend to spread myself too thin a lot, I think it's in my nature. You really can't do too much tower crossing without compromising quality, so I try to pick one or two interdisciplinay projects to participate in each year, depending on the amount of work expected for each project. It has to be an intentional and planned approach or you can find yourself with way too many irons in the fire.
As an academician, I completely agree that crossing towers is needed. Who needs analytics more: a biologist or a sociologist? A psychologist or an economist? The fact is that analytics, like technology, does not exist in a vaccuum. It is a powerful tool for all disciplines. We ought to be stretching out across fields of study, across the boundaries of colleges or departments, and helping one another accomplish real, valuable work using the tools at our disposal. Where I teach, the only way that's happened has been for me to take the initiative to work one-on-one with colleagues in other departments. When they have projects on a health epidemic, urban sprawl, poverty, teen pregnancy, etc., their projects almost always generate data, both structured and unstructured. If I'm willing, there is no end to the opportunities to offer my analytics expertise to their work, but I must be willing to embrace interdisciplinarianism!
It really can be a pain to cross between colleges, my first attempt at college left me feeling locked out when I tried to grab a minor to go with my computer science major. There was just no way I could make it work and no one in either college was very helpful. 15 years removed from that situation I can say it worked out for the best, but at the time it did seem so.
While 97% of insurers say that insurance fraud has increased or remained the same in the past two years, most of those companies report benefits from anti-fraud technology in limiting the impact of fraud, including higher quality referrals, the ability to uncover organized fraud, and improve efficiency for investigators.