Competing With Data & Analytics
What to Read Next
When Michael Rappa first suggested that North Carolina State University develop a business-oriented master’s degree program in analytics back in 1999, based on the rise of data and what he was seeing from companies like eBay and Amazon, he encountered one key obstacle: it was too early.
It was still early when he proposed it again in 2005, but this time North Carolina State asked for a full proposal, which led to a degree program that accepted its first students in 2007.
Now Rappa’s Institute for Advanced Analytics, the first business analytics program in the country, has 80 students. Nearly 60 other programs now offering advanced degrees in analytics, too. Rappa spoke with MIT Sloan Management Review’s Michael Fitzgerald about his program.
What changed between 1999 and the formal organizing of the program in 2006?
By 2005, the university was searching for forward-looking ideas, especially ones that had a strong practical orientation that would engage industry partners. Tom Davenport’s [January 2006] Harvard Business Review article [with Jeanne Harris], which led to his book Competing on Analytics, had just come out, so the term all of a sudden hit people’s radar pretty contemporaneously with the proposal.
Even so, it was a very audacious proposal. It was so outlandishly “far out there” that I think that’s why I got a chance to do it — people didn’t think it would really work. You’re proposing an entirely new degree, developing an entirely new curriculum, all new courses, and doing it within a year. If you put that in a university context, it sounds ludicrous.
Did one of the questions at the time involve asking why is this different from, say, a Master’s in statistics?
Sure. Or informatics, even other things. That’s a natural question. My argument was that we needed a university-wide initiative that wasn’t siloed in the business school, or the computer science department, or in statistics, or mathematics, but instead could draw faculty from any of those fields or specialties, so we could construct a new degree.
In my mind, it was really straightforward. I was thinking of an analogue to the MBA. The MBA is a professional degree; it’s a bundle of skills. I was thinking of Master of Science in analytics in an analogous way — a professional degree much more centered around this new world that we were entering into, that was so data-infused. We needed to create a new kind of data-savvy business professional.
Do they pat you on the back now?
It’s been a fascinating eight years. At first it was, “It really won’t work, let’s not worry about it,” then it was, “Okay, maybe it’ll work, but it will just be something small,” until we were not just growing, but doubling in size. Then it became more of a kind of anxiety, like, “What is happening here?” Now I think the university sees it as what it is — this amazingly successful program that we led the way with.
Certainly now we’re seeing plenty of analytics programs pop up. Would you say the new ones are mostly similar to yours, trying to be a new professional school?
I think so. I’ve cataloged close to 60 programs. They split out between a Master of Science in analytics or Master of Science in business analytics. If it has business, it’s almost always in the business school and looks more like a business degree. The others tend to be a little more like ours.
You commented in a recent interview that you have to turn away students who are qualified for the program, and were thinking about expanding.
We may set in place a plan to grow to at least 120 [students] by next year. We’re trying to locate new space, we’re building another building. We do think that the program will continue to grow, and I think we may land somewhere between 120 and 150 in the next two years.
Are you primarily drawing from professionals or from recent college graduates?
We originally thought it would be a five-year thing, like a Master of Accounting degree, which is like the fifth year of a four-year degree. Then in 2008, 2009, the recession kicked in, and there were these experienced people in the workforce starting to think about going back to school. We now split right down the middle: Half of the current cohort, the class of 2015, will have graduated sometime in the last 24 months, and half have three or more years on the job. That ratio has been about the same for the last six years. It bounces around a little, 55/45, but it’s been a remarkably stable split for us.
What skills did you deem important for the analytics program?
Universities normally “coursify” the curriculum — what 10 courses do I have to teach, what are the core [courses], what are the electives? We said at the start, let’s not even presume that what we think is important matters. Let’s start with the notion that the employer is the customer. We wanted to really understand what employers were looking for, and then build the entire curriculum to achieve those skill objectives.
That led us down a road of looking at five core skillsets. Employers may start with the technical skills, the programming skills and stats skills, but they also want strong teamwork skills, strong communications skills — they actually want hands-on experience with the kinds of tools [the employee] may be using. And then they want someone who is oriented around their business problems and more generic problem-solving skills. Someone who knows how to look at problem and parse it and build an analytical framework around it. So we took those five things [technical, teamwork, communications, tools experience, problem-solving] and built a curriculum to achieve them.
Since teamwork skills are important, rather than saying, “Who’s going to teach the teamwork course?”, we’ve structured the entire curriculum around teamwork and peer evaluation in the context of teams, and coaching of individuals in the context of teams. And then we have some lectures about how to function in teams, and leadership, and followership and stuff that really helps the student develop teamwork skills. Our entire curriculum is team-based.
What skills are important now that weren’t when you started?
In 2006–2007, text mining was not as predominant as it has become. That part of the curriculum has grown and grown over time. Machine learning is becoming bigger. Early on, we did less than we do today with things like the design of experiments.
We call our degree an “artisanal degree” [laughs]. It’s handcrafted every year. We don’t have 10 courses that are kind of chiseled in stone. Instead, we’re constantly sizing and resizing to what we think is needed.
How do you keep up with trends?
We do an open solicitation of practicum proposals. This practicum is really a core element of our curriculum. We typically get 50 to 60 a year, of which we can do 17 projects where the students are teamed across eight of their 10 months. They receive data — complex, messy data — in large quantity from a sponsoring organization and go through a very structured process of trying to understand the problem, cleaning, augmenting, and building the data, wrangling and so forth, building the models called for, and producing a report and presentation for the sponsor.
We look very closely at those 60 or so proposals to see what kind of problems organizations are posing, what kind of data they’re analyzing, what kind of methodology they think is important. We’re analyzing the content of that call for proposals each year, to get a litmus test on how things are evolving in industry, and then either adding or sort of shrinking elements of the curriculum based on our perceptions.
Is the practicum the main way of keeping in touch with what companies want?
That’s an important one, because we see concretely what kinds of problems they’re grappling with. But also, the Institute manages its entire placement process on its own. We bring about 50 companies a year through the recruiting process, one time in the fall and then in the spring, when they’re actually doing the outside interviews of our students.
We sit down with them over lunch or to hear their presentations, to hear the things they’re talking about. Every time a company comes, we congregate all 80 of our students into a lecture hall, and the faculty get together and go listen to what’s happening. It doesn’t simply depend on professors’ consulting activities or how much they’re talking with folks.
What’s an example of a new skill that’s emerging?
I was hearing more and more that a really important skill for data scientists is storytelling. My immediate reaction to this notion was, back at MIT [Rappa taught at MIT Sloan School of Management for nine years], if someone said the presenter was storytelling, it wasn’t really a compliment [laughs]. And I kept hearing it. And so many companies I would talk to were saying there’s a breakdown over the last mile — connecting with the decision maker is like the “last mile” of analytics. That breaks down very frequently in organizations.
So I thought that for connecting with decision makers, when what you’re doing is very black box, you have to get the story across and keep that individual engaged, so that they can understand it and be confident in the work you’ve done. You can’t talk about it like an academic presentation.
So I asked around and someone said, “Oh yeah, we have this person doing a workshop on storytelling as part of our analytic meeting.” I called her and said, “How about you come down here and give us three days of storytelling in the spring?” She’s like, “Yeah, great!”
What kind of jobs do graduates get?
It breaks down to three main categories. First one consists of consultant [jobs], top tier management consulting. That’s about a quarter [of the graduates]. Another quarter will be analysts — put any adjective in front of the word analyst, financial analyst, risk analyst, and so on. The third main category, really something that is surging now, is data scientist. That title didn’t exist when our first class graduated in 2008. It was probably a quarter last year, and a third this year. The rest, the balance, will be manager of something; maybe they’re branded developer, systems developer, systems architect, because they have strong programming skills.
Is job satisfaction higher with one group than another?
We haven’t asked pointed questions about satisfaction in employment. But my impression is [that] there’s really no difference. I think the managerial pool is relatively small by comparison to people going into categories like consulting, analyst or data scientist. I can say there’s very low attrition in terms of people leaving consulting. The data scientist category is so new in terms of being used as a job label, but it has given them a vision of a professional future that’s been a catalyst for more students.
Has hiring been that strong since the start?
It has been a consistent step function year after year. Our first class graduated before the recession hit, in spring 2008. In 2009, there really was no let-up in demand. What became apparent in 2009 and 2010 was that these skills were more important than ever. The only thing that’s changed was in 2009 through 2011, the demand function of different industries shifted. Banking went down. Right now, banking is one of strongest sectors for us.
Students on average last year had between 3 and 4 offers. That’s a lot of offers to have. That’s up from 3 the year before, 2 and 2.5 the year before that. It’s pretty incredible to see that level of demand. And half of those people have no prior work experience. Think about that for a second.
The other thing we’re seeing is a ratcheting up in demand. These guys are no sooner on the job than they’re getting pinged by recruiters on LinkedIn with job offers. It’s a very interesting, intense environment to be in if you’re identified as having these skills. You’ll be relentlessly pursued.
That must be gratifying.
For them! For me, it’s stupefying. As faculty, we sit around and say, “I don’t remember having more than one job offer at any one time.” It’s an amazing thing to see.
Some question whether any individual can be the master of all the skills companies want. Is that a problem?
The skills bundle companies want seems to be expanding, and for a single individual really to have that mix — they do start to sound like unicorns, relatively rare. They probably have a PhD, and are uniquely comfortable with computer programming, applied mathematics and statistics, which is very rare.
If you are trying to build that type of individual you need a six- to seven-year pipeline to produce it, and that’s not very scalable, that’s not realistic. We have focused our efforts on saying, when you look at industry, they’re producing teams of people to do this work and different people come to the table with different skill mixes. Our program is deeply vested in building a strong team player in this process.
We don’t believe there’s this single template of what a data scientist is. We believe it is this very broad mix of skills, and some of the people will be strong in certain areas and others in other areas, but what they have to have is a very strong teaming ability. Which isn’t really part of the typical skillset of the archetypal PhD.
University faculty get where they are through a thorough-bredding process. They accomplish more and bring in more money than anyone else. They don’t get there through a teaming process. When industry builds a team, they need to bring people together, and that teaming process is very different from the academic process.
Are the data scientists getting jobs that are as innovative as they hoped? We hear stories of data scientists getting bogged down in improving legacy IT systems.
I haven’t heard that. What you’re saying to me is new.
I will say that students get exposed to the really painful aspects of working with the data while they’re here. Our students are well primed for what it’s going to be like, how messy it is, how challenging it is, that it’s not the glitz and glam.
At the same time, when I do interact with students who’ve gone on to data scientist positions, I think that they are excited about roles they’ve taken on. We vet our employers pretty closely too. We’re very focused on whether an employer is for real, and if the positions they offer are really the kind of positions students can excel in.
Also, when a student graduates and has an average of three to four offers, they’re not going to stay [with an employer] in a bait-and-switch situation. They’re going to leave almost instantaneously. Those other folks who put offers on the table, I think they’re communicating and saying, “If you’re not happy, please call us.”
How intense is the competition? Is it dotcom-esque, with super perks, like cars, for top candidates?
No cars — it’s not eye-popping behavior. But competition is pretty intense. I think it’s about the speed of the process. The turnaround is very fast. They’re interviewing, and two days later, they’re making offers. You see companies using the corporate jet to bring candidates on-site for interviews. It’s very aggressive.
IT and business famously have trouble speaking the same language. Analytics is neither. Does this make for yet another communications gap companies have to manage?
Our vision right from the start was that these folks [graduates] would be the bridge between the two, IT and business, so they had to be able to communicate with both sides, and had to be able to interrelate with IT staff. That’s why the curriculum, right from start, emphasizes team-building skills.
Do the hot analytics hires, your students, spur resentment from the people who’ve been in the analytics trenches for a while?
I’ve not seen the resentment issue. What I’ve noticed, though, is that after our students start their jobs, it’s not unusual for us to see employees in those organizations inquiring about the program and even applying to the program. Maybe people in the organization are getting the sense they need to take it to another level.