Most graduate degrees in analytics are scams
A true American hustle. Plus, more White Lotus Power Rankings.
If data analysts had a mantra, it would be "it depends." Yes, you asked a simple question—how much revenue did we make this quarter? Do people like our new product? Was our big strategic pivot a good decision?—but I cannot give you a simple answer. Nothing is black and white; there is nuance, texture, gradient. Certainty is for brutes; absolutes are for drunks. The real world is probabilistic; Bayesian; fundamentally uncertain. To ask an analyst about concepts such as "revenue" and "people" and "good" is to ask Richard Feynman about magnets: For us to explain that, we have to start with the big bang, and work forward from there. How much money do we have? Ha. You sweet summer child, we will think. "It depends," we will say.
This is an annoying and counterproductive tic, but in some rough theoretical sense, it's usually correct. The ledger really does have things on both sides. Most decisions don't have easy answers; if a decision is worth debating at all, it probably has legitimate pros and cons. Tradeoffs are annoying, but tradeoffs are often real.1
But not always. Every once in a while, decisions are easy. They might appear hard—like they are driven by some complex calculus; that they need to be weighed, considered, measured twice, and SWOT’ed. But sometimes, they are not. Sometimes, the balance really does tilt one way, and the world really is that simple.2
For the season, here is one such decision: Should you get a master’s degree in analytics?
No. The answer is no.
On the surface, these programs serve a valuable need.3 In the late 2000s, the corporate world got enchanted with the promise of big data, data science, and the predictive power of analytical precogs. Data science is “the sexiest job of the 21st century,” said the Harvard Business Review, famously. The people who can do it are “unicorns” and “high priests,” said the Wall Street Journal. Big data will “transform how we live, work, and think,” said university professors and editors at the Economist.
But all these grand predictions included the same caveat: There is not enough analytical talent to meet the explosive demand for data scientists. A quant crunch was coming, said IBM. McKinsey issued the same warning. That was the drumbeat, at the time: Data volumes are doubling every two years; data science is sexy; there is a looming skill shortage. “Good with numbers? Fascinated by data? The sound you hear is opportunity knocking,” said the New York Times.
Reasonably, a bunch of colleges rushed to bridge the gap. In 2007, NC State launched the country’s first master’s in analytics. In 2013, NYU debuted the first master’s in data science; a year later, Berkeley rolled out the first online one. Not to be outdone, UVA created an entire institute.
Now, dozens of schools offer similar master’s programs, MOOCs, minors, concentrations, and data-related certificates. According to a 2021 study, over half of the world’s accredited business schools have business analytics or data science degrees; they accounted for 20 percent of these schools’ degree programs, and produced tens of thousands of graduates. If you want to be educated in the lucrative art of data science, there are dozens of colleges ready to help.
Or, more specifically, there are dozens of colleges ready to take your money. Because that is the first thing that has to be said about these programs: They are products, created to turn a profit.
Master’s degrees of all kinds have always been this way—“unregulated cash cows” that universities use to “enroll students at an industrial scale.” Students want to put a fancy school on their resumes, and master’s programs give colleges a way to launder them that right, without sacrificing their academic brands:
If you’re offering bachelor’s degrees, they all have to be four years long. … You have to publicly publish your acceptance rates, your average SAT scores, so to the extent that you’re selling selectivity, you actually have to back it up with data, whereas in the master’s degree market, you can call almost anything a master’s degree. Master’s degree programs do not have to publish their admission statistics, which creates, I think, an enormous temptation for institutions that have very attractive brand names, that are attractive in no insignificant part because their undergraduate programs are very selective, to open up the floodgates on the master’s side and pay no penalty in the market because people don’t know they’re doing it.
A lot of data science and analytics programs stretch this cynicism even further. One senior university administrator told me that their analytics degree was hastily thrown together in the mid-2010s, not because there was any expertise to teach it, but because there was money to be made by selling it. It was offered by the business school, as it is at most universities, and had next to no association with the math, statistics, or computer science departments. The curriculum, such as there was one, was mostly cobbled together from existing business school classes and technical buzzwords like data mining and natural language processing. It was not an academic program or even a professional degree, but a year-long job fair, full of guest lectures, networking events, and interview training. The program also explicitly targeted international students as recruits, because they were more easily exploited: The degree was sold to them as a gateway to the U.S. job market, and, more cynically, as a way to get a visa.
The degree’s final “capstone project,” which is a ubiquitous feature of these sorts of programs,4 was an accidental analogy for the entire industry: It has the appearance of practicality and prestige—you build a fraud model for Fidelity! A luggage fee dashboard for Southwest!—but it is in fact manufactured busywork. It is not an education or real-world experience; it is a line for a resume, a thing to talk about in a job interview, and a gold card visa program, for as “little” as $32,000—and as much as $129,000.5
Of course, maybe that’s a good deal. Analytical jobs pay well, and most people attend these programs for the job opportunities anyway. The education is incidental, to both school and student.
Conveniently, the schools have done the math for you, and will tell you it’s worth it. The University of San Diego says advanced degrees will increase your salary by $15,000 to $25,000 a year; NC State says if you attend their program, your salary will nearly double, and you’ll pay back your tuition and your lost earnings in under three years.
I mean, maybe; I’m sure there are success stories. But educators have a habit of monkeying with these numbers, and things like NC State comparing the median salary of its incoming students to the average salary of its graduates are at least a little suspicious.6
I can also say one thing more definitively: In Silicon Valley, these degrees are often disregarded. Most startups care about academic diplomas as a signaling mechanism—they like that you got into MIT and Stanford more than like that you were educated at MIT and Stanford.7 They aren’t looking for someone who went to classes for four years; they are trying to figure out your IQ.
At best, pay-to-play master’s programs are ignored, and the degrees they offer are looked at as undifferentiated paper credentials. At worst, they’re viewed as a red flag, the dark mark of someone who proudly fell for a scam.8
Even for less judgmental employers, master’s programs can frequently be seen as lessons that need to be unlearned. Most analytical jobs are messy: They are figuring out how to translate vague questions about things you can’t directly measure into simple math about things that you can measure. They are finding clever proxies for unanswerable questions, and useful facts in the incomprehensible noise of test accounts and mislabeled UTM parameters. They are politics. They are trying to read the vibes of a scatterplot of 15 customers, and squinting at wiggles.
A lot of master’s programs implicitly frame the job differently. They teach people how to do hard math on easy data. They teach the academic principles of data visualization. They teach toy problems, stripped of their authentic context.
None of which is necessarily bad; better to know the academic foundations than not. But those foundations are not the job, and a lot of startups are reluctant to hire people who think that it might be. And it is a little bit weird to get a degree that will, first and foremost, prompt a lot of employers to ask how much of that degree you’re able to forget.
So what do you do instead? I mean. First, ask someone else, who’s better at getting and keeping a job than I am.9 And second, do the capstone project yourself. Find something you’re interested in and a question you want to answer, and answer it.
A big part of the problem with capstone projects is that they’re contrived. They are sandboxes on sample data. The subject is “real,” in that they’re case studies about actual businesses, but the analytical work is not. It is a prolonged interview question that the host company probably plans on throwing away. And it’s better to answer a genuine question about anything—movies, politics, the emotional proclivities of Skynet, whatever is fun, to you—for several reasons:
If you’re actually curious about the question, you’ll dig until you uncover something interesting, rather than just trying to get to the answer in the back of the book. That digging is how you really learn; that curiosity and (ugh) passion is how you make something good that people will be impressed by.
Counterintuitively, it’s a better representation of the real job. If you analyze a hobby, you won’t know where to find the data you need; you won’t know how to interpret it; you’ll spend a lot of time chasing dead ends and finding inconclusive answers. That’s better training than some textbook simulation about a supposed grown-up business problem. Plus, if you hate these messy parts, that’s a good thing to figure out before you commit a career to it.
Agency is all the rage these days, and blogs and personal projects are high-agency endeavors. Drifting through a degree program is not.10
It makes you interesting. Hiring managers see hundreds of capstone projects; they have interviewed dozens of candidates who say they’ve built a model to optimize some marketing funnel or lead scoring algorithm. They haven’t talked to many people who built a tool to help them track their cat’s weight, or have a blog about Spelling Bee.
You always wonder what’s down the other path. What if you went to this school instead of that one? Moved here instead of there? Joined that one company, stayed instead of went, said yes instead of not now? The road less traveled is tempting; the road untraveled is haunting. What life might we have, down that road?
But sometimes, the path doubles back. You catch a glimpse of it from the one you’re on, and see where you would’ve ended up.
Back when I was looking for my first tech job, I applied to a few graduate programs that were precursors to today’s analytical master’s degrees. A few years later, I ran into several people who attended those programs at the same time I would’ve. We were all doing the same things: In one case, we were peers on the same team; in another, we had essentially identical jobs at parallel companies.
And that, I think, is the nicest thing I can say about most of these master’s programs: That, in the end, they do nothing at all.11
The White Lotus Power Rankings
It’s been a real bad week for Sritala:
Also, Rick? The most charming, by a mile? After the whole thing with the snakes? Snakes?
Vote for episode four, but man, vote better than you did for episode three:
Of course, most people aren’t asking for the answer in a rough theoretical sense. When we get asked, “What will happen if I trade Luka Dončić for a 32-year-old center made of balsa wood?,” the correct answer is not, “In a rough theoretical sense, it depends.” The correct answer is, “You will get crucified.”
Ibid.
In, you know, a rough theoretical sense.
Fifteen of the 35 programs profiled by Fortune mention “capstone” projects.
In fairness, these programs do offer one genuine benefit: If you are into college sports, and your alma mater is a perpetual disappointment, getting a master’s is a socially acceptable way to buy your way into a new fanbase.
Probably an honest mistake; it’s not a degree about data or anything.
This is rough, and there are obviously exceptions, especially in highly-technical research jobs. But for the general startup population—for analysts, engineers, marketers, and so on—this broadly holds.
Also, if you have one of these degrees, that’s fine! The schools are the scammers here, not the students. What’s important, I think, is to be aware of how the degree is perceived.
No, really, ask other people. Tech jobs are extraordinarily nepotistic. The good news is that it’s light nepotism: People don't hire their cousins as much as they hire people that their friends introduced them to. Cold outreach to people can be surprisingly effective, so long as the emails are genuine, and ask real, material questions. People love feeling helpful. They do not love coming up with ways to be helpful.
This is part of makes a lot of these programs so contemptible. For some international students, master’s degrees like these really are a useful entry into the United States, at a time when those are getting increasingly rare. But because the programs are often abused by the colleges that provide them, all their students are tarred with the same stigma, even the promising ones.
I get that this is terrible analysis, and that I’m taking all sorts of artistic license to make this point in this way. This is what happens when you don’t get a real analytical education.
I have mixed feelings about these degrees.
I have a BA in economics/political science and managed to stumble into data science through some combination of self teaching and free/cheap online courses. The data science bootcamps through udemy and coursera were quite helpful for walking through python in a more specific way than a generalist programming class would. However I always viewed them more as a skill tutorial to get started than as a comprehensive career guide and I would never pay more than like $60 for them. When I see resumes with analytics masters degrees, I do have that reflexive sense that the job candidate overpaid for skills the could learn on the job or self teach.
However, I’m also a confident white dude who people frequently read as smarter than I actually am. Early in my career, people were willing to take chances on me and let me try things my resume at the time probably didn’t back up. Over time those chances and projects I wasn’t quite qualified for built up into a base of skills and experience I can now take to the job market to show my qualifications without talking about my degree at all. That’s not the case for everybody, and I think it’s harder to do in today’s job market than it was in the early days.
Not all that long ago, data teams were a rag tag bunch of nerds with a patchy old server inventing processes on the fly. There was no career path or formal academic training, and those teams were willing to take on new recruits who seemed smart and scrappy and reminded them a bit of themselves. My first mentor was a Russian Studies major who taught himself programming at the public library, my other colleagues came from across the academic spectrum and only one of us had an actual computer degree. Those early teams were fairly white, male, and nerdy and that shaped who reminded them of themselves and who might get that stretch opportunity.
I do think analytics degrees serve a purpose in allowing people who might not get those same early career opportunities to break into the field in a way they might not otherwise have access to. Several of my best colleagues have talked about the importance of that masters or pricey boot camp in allowing a career transition, often women or immigrants who just needed something to get through that initial resume screen. I think that’s especially important now as we’ve codified a bit more of the formal job skills and experiences we expect in data careers since those early days.
The moral of this story is twofold: First, these degrees do help people who might have a harder time breaking into a data careers, and I try to see their pursuit on a resume as a sign of ambition as opposed to a sign of falling for a scam (I do think they’re overpriced though). Second, even as our job responsibilities become more codified, we should try to take chances on people who are bright or scrappy and don’t have the exact technical skills yet, and we should take those chances on a broader swath of people. My best hire was a recent liberal arts major who taught herself a bit of data analysis at her last job but clearly didn’t know SQL all that well. I hired her anyways and she’s now 3 promotions into her data career and running circles around me from a technical standpoint.
I think this is overstating the case by a lot. I work in an analytics department at a large company and most people (not me) have a masters in business or marketing analytics. I learned the same skills on the job but: a. I was opportunistic and lucky b. I still needed a data science bootcamp to learn SQL and Python. (Please don’t tell me I can learn on my own. I know my own limitations.) Most of the people I work with have foreign visas, can’t be out of work for significant periods of time, and need a dependable path.