Boot Camp: A Business Saga
The signaling model of education is not a model of hope. It tells a tale of stable waste, not an “education bubble.” That’s why I’ve repeatedly bet against any major fall in the share of youth enrolled in traditional four-year colleges.
Still, the signaling model does tantalize us with the potential for a better world. If you could only somehow break free of education’s “locked-in syndrome” and convince even one employer of your pre-existing employability, you could skip college and save hundreds of thousands of dollars in time and tuition. When readers share their rare success stories with me, I share them with you.
Recently my friend, educational entrepreneur Edward Nevraumont, offered to write a guest post on his experience in actually creating and running a coding boot camp to help clients fast forward to career success. Read to the end, because he thoughtfully reflects on the scalability of the model. Here’s Ed!
General Assembly is the largest “coding school” in the US. In 2011, they launched their first location in New York; today, they have a physical campus in most of the biggest American metros and an online presence in many additional cities. Plus, they’ve established programs in five other English-speaking countries around the world.
I joined GA in 2016 and oversaw sales and marketing until we were acquired by Adecco in 2018. Since that time, I have spoken to numerous private equity firms who wanted more insight into the training space before they made investments. For these calls, I generally charged $1000 per hour. These consultations have been easy, because most firms asked the same questions and remained ignorant about the same obstacles.
At the risk of lowering the likelihood that future PE firms will want to pay for an hour of my time, I thought I would share with you what I usually say to them.
General Assembly and the appeal of bootcamps
GA was originally founded as a co-working space (they were a WeWork competitor). As a perk for customers, GA began offering evening courses on various tech and tech-adjacent subjects. Before long, the courses were more popular than the co-working plans, so the company pivoted.
The first course was a full-time “web development” program. Students spent an intense three months learning front-end and back-end web development. The “brand promise” — not a guarantee — was that three months with GA would be enough to land a full-time job in the field. And it worked.
Bryan Caplan argues that education is mostly signalling, rather than training. Is a Harvard graduate great because of what they learned at Harvard or because Harvard is extremely selective on who they admit to begin with? The question practically answers itself: would you rather have a Harvard degree and learn nothing, or acquire what a Harvard grad learns but not receive any certification? In GA’s case, there may have been some selection effects in the type of people who chose to take a three-month coding course, but the school itself did very little selection. Any reasonable candidate was admitted.
And yet, after only a few cohorts, GA was able to confidently say they could take just about anyone and spend three months training them to become job-ready for the tech industry — AND find them a job.
As one strategy for attracting candidates, GA created white papers with titles like “How to get a job at a start-up.” We advertised the papers on Facebook and offered them for free — with the submission of an email address. Once we collected their email address, we would push messages to the individual about our paid courses. As it turned out, the overlap between “people looking for a job at a start-up” and “those interested in paying $15,000 for a three-month coding bootcamp” was high enough that the economics worked.
But here is the dirty secret…
Most graduates of coding bootcamps, GA or otherwise, do NOT get jobs at start-up companies. Many start-ups, at least the promising ones, are competing for talent with FANG companies (and other trendy tech sectors). Most bootcamp graduates are not landing jobs at FANG companies. (It did happen occasionally, and you can guess who ended up on the cover of promotional materials…).
Instead, most GA graduates got jobs at one of the thousands of companies in America that are not the next hot thing. Famous tech companies aren’t the only ones that need web development and UX design and data science. Insurance firms, database administrators, manufacturing companies, and lots of other “normal” businesses also need employees with technical skills. And, unlike FANG companies, these traditional companies do not have a lineup of MIT grads knocking on their doors.
To be clear, most GA graduates end up making a LOT more money as a web developer for a “boring company” than they were earning before they completed a bootcamp program. And even if the graduates’ new industry is less exciting than social platforms or electric cars, their day-to-day work is usually FAR more interesting than their previous job.
GA found product-market fit very quickly. There was pent-up demand for an opportunity that allowed an individual to quit their dead-end or uninteresting job, and then jump into the world of tech — even if it wasn’t at a high-status “tech company.”
Initially, GA’s most powerful marketing channel was evening workshops and events. Every GA “campus” (usually part of a floor in a downtown high-rise), featured an open space that students would use during the day for completing assignments and working on group projects. In the evenings, these open spaces could be easily repurposed to host events and workshops. Everyone who attended a GA event ended up on an email list. In turn, we targeted our email lists with robust marketing campaigns.
The email list did not need a high conversion rate to make every address we obtained incredibly valuable. For instance, if just 1% of the attendees from a GA event eventually took a course, each email address from that event would be worth $15,000/100 = $150. Email addresses from events were so valuable that it rarely made sense to charge for attendance. Far better to have 100 free attendees (worth $150 x 100 = $15,000) than 50 attendees paying $10 each (~$8000 in “value”). And if free beer enticed an extra 20 people to show up for an event, then the cost for drinks was worth it.
GA’s New York campus blew up, and the young, talented team built out operations capabilities that scaled the company very quickly. GA opened a second campus in London, which made them an “international organization” almost from the beginning. Other US cities soon followed, one after the other. Growth was outstanding in any given city, and the ability to launch in new cities suggested the total addressable market was enormous. GA raised multiple rounds of capital at escalating valuations.
The General Assembly Moat (or lack thereof)
Just as GA continued raising capital, two things happened simultaneously that created headwinds.
First: The lake dried up.
Imagine a lake of fish (young people) who are looking for the GA solution. No one has ever fished in that lake before. As soon as the first coding school opens, the fish start jumping into the GA boat. Once all the fish have been caught, the fishers have a few options:
They can move to a new lake (i.e., launch in a new city), and/or,
They can fish at the mouths of the rivers feeding the lake. There will always be new fish who might be interested in the product.
We were already pushing forward on option 1. The growth potential of option 2 was limited; after the initial spike in interest for a new GA market, demand naturally slowed down.
Second: We had no moat against local competition.
Building a national rival to GA would have been very difficult, but launching a single coding bootcamp in any given city was ridiculously easy. On the smallest scale, an individual instructor could just throw up his or her shingle on some low-cost office space and start offering courses. The only costs were the rent, the individual’s time, and marketing expenses to attract students. In some ways, coding bootcamps are like restaurants. Anyone can open a restaurant, and so competition is intense. There are many obstacles to building a national chain of restaurants, but opening a “family-run” local restaurant is accessible enough that it’s a default choice for many new entrepreneurs.
In the same way that most restaurants close after a short period of time, most coding schools also fail. And when one bootcamp shuts down, another might open up — the competition will always remain fierce, even if the next round of coding camps is equally likely to fail.
General Assembly was first, but we had no moat against local competitors. Why pay $15,000 for three months with GA when you could pay $14,000 (or less) for a course taught by a former GA instructor down the street?
The value of market-tested programs
Most corporate training programs are built for the buyers, not the users. The buyers are usually HR organizations that have been tasked with developing internal training programs within a specified budget. There are usually general goals to provide a variety of options (with “good-sounding” titles). But whether the attendees of the classes actually learn something that improves their ability to earn a profit for the company is rarely considered and never measured.
Because the early General Assembly courses were paid for by the actual consumers, ensuring they provided the skills needed to get a job was extremely important. The most critical internal metric was not how many students we got to pay for a course, but rather how many students did we get to pay for a course AND go on to graduate and land a job. The feedback cycle on the latter metric was slower than most e-commerce businesses, but it was not THAT slow. Most programs were only three months long, so we only had to wait around four months to have a pretty good idea if the marketing channel and product were working. If a cohort of students attained below-average job placement rates within a month or two of graduating, then GA could quickly pivot to ensure that did not happen again for the following cohort (or at least the cohort after that).
We discovered a few key learnings during our expansion phase:
Some cities lacked enough employer demand for coding skills to ensure a GA program could meet its promise. For example, there was significant interest from prospective students in Dallas and Houston, but not enough employer demand to make us comfortable with the idea of launching campuses in those cities.
Some specific fields of study had a harder time placing students after only three months of training.
When I arrived at GA, we offered three full-time programs (i.e., programs that allowed students to go from zero-to-hero and get a job upon graduation):
UX/UI/Customer Experience Design
Data Science required some prerequisites (like knowledge of Python, which we also taught as a two-weekend course), but the other two programs required no previous knowledge at all. And yet, GA was consistently able to graduate students from these three programs and place 95%+ of the graduates (who were looking for jobs) into relevant positions within a few months of graduation.
We also offered part-time courses, which ran twice a week in the evenings for ten weeks. The part-time programs delivered enough content to help dedicated students secure a raise or promotion, but they were not intensive enough to make a novice job-ready for a new career.
GA offered more than a dozen part-time courses, most of which fell into a subset of our full-time courses (e.g., front-end web development). In addition, we offered part-time courses in two other fields:
After starting in my role, I asked the executive team why GA didn’t offer marketing and product courses as full-time programs. In response, I learned the company HAD tried offering product management programs, but they could not meet the promise of “pay $15,000 and we’ll get you a job.” Instead of GA’s standard 95%+ job placement rate, the product management programs achieved ~60% job placement rates. In context, ~60% is still better than many undergraduate programs, but GA’s bar for job placement was higher. Although the full-time product management courses were scrapped, the part-time program was not only maintained, but it became one of our most popular programs, particularly with computer engineers. Many developers possess excellent coding skills but are not comfortable with managing a team of engineers. GA’s product management course gave those individual contributors the ability to see the bigger picture — and often get promoted. Digital marketing offered a similar pitch to traditional brand marketers: “take your traditional marketing skills and learn the vocabulary and skills of digital marketing — and get promoted.”
In all of these scenarios, GA was constantly judged against the market. We received (reasonably) fast feedback on who, what, if, and how we were teaching was good enough that employers were willing to pay for it.
In other words, we were the opposite of most HR-run corporate training programs.
Sometimes, a manager would be so impressed with the skills one of their employees learned from a GA course that they asked us to train their entire team. As we took on more and more of these projects, it became obvious that GA needed a B2B product to focus on corporate training. From the beginning, though, we pursued a very different type of corporate training. Instead of checking the boxes required by HR, GA’s training appealed to the line managers who needed to upskill their teams.
We worked with a consulting firm that employed 1000 data analysts. The firm’s hourly rate for data scientists was almost double what they charged for a data analyst. If we could competently upskill their data analysts into data scientists, then there would be an easy arbitrage opportunity for the firm.
And so, just as GA’s consumer business was hitting a wall (or at least significant headwinds), we identified an enormous opportunity in this “unusual” corporate training business. Unlike the consumer bootcamp business where we were barely more than a commodity product, our corporate training products were unique and valuable. Most of our corporate training competitors were still trying to meet the needs of HR. GA was the only company offering products that demonstrated a market-tested ability to upskill individual employees in ways that increased their value.
Some rival companies started recognizing the opportunity for upskill-focused corporate training programs, and we began to see the larger ones in competitive-bidding situations. GA almost always won, and here’s why: we held a national reputation. If you were an individual looking for web development training, the fact that GA was the largest course provider — with campuses around the world — might have been interesting to note but largely irrelevant to your needs. Most prospective students (especially pre-COVID) just wanted to acquire the skills required for a job at local company. The other two dozen bootcamps in their city could usually meet that need — if not as well, then at least close to the level of GA. But for a multinational company looking to train thousands of employees across a dozen cities, GA’s scale was an important factor in de-risking the entire initiative. As a result, we ended up with a larger footprint in corporate training than we did in consumer campuses, but it was the scale we built with our consumer business that gave us the authority to win in the B2B market.
GA’s growth trajectory was tied to the shift from “consumer pays” to “employer pays.” At first, it was much faster and easier building the consumer business before trying to expand into B2B sales. But it became clear over time that the consumer business had a much smaller total addressable market than anyone initially believed. There were just not that many people willing and able to not only drop out of the workforce for three months, but also to pay $15,000 — even if the ROI was clear. Ability to pay was a particular challenge. Unlike traditional college programs, any loans made to GA students were not backed by the government; as such, MANY students who wanted to take courses with us could not obtain any loans at all. A potential student with a low credit score and limited assets was not an attractive candidate for a traditional bank, even though GA students had a successful track record of securing high-paying jobs after graduation.
You might think this situation could provide an arbitrage opportunity for an unconventional bank, but we were never able to find those banks. In many cases, I am sure the banks were RIGHT to be conservative. While GA had a remarkably high placement rate for students who graduated and looked for a job, that eliminated from the denominator all the students that dropped out of programs before finishing, as well as all the graduates that decided NOT to look for a job. Why would someone pay $15,000 for a course that promised them gainful employment and then choose not to look for a job? Sometimes, people had clear reasons: they wanted to upgrade their skills for a promotion, or they planned to launch a start-up.
But most of the time, the lack of pursuing a job came down to laziness. GA employed managers who were tasked with placing students into jobs. I originally thought these managers would spend their time networking with local employers and teaching the community about the value of GA student skills. When I considered the calculation for fixed-vs-marginal cost, I assumed that we would only need one of these career managers per city, and that economics would get better as the number of students increased. But I was wrong. While the career managers did perform those tasks, most of their time was spent staying on top of the graduating students: teaching people to write resumes and coaching them for interviews. Plus, the managers also needed to make sure graduates met basic employer expectations, like wearing professional attire to interviews and showing up on time. A significant number of students did not like that approach, so they stopped using GA’s career support entirely — not because they got a job, but because they did not want to go through the process of looking for one.
GA was the biggest provider of bootcamps at the time. Our largest competitors were MUCH smaller and included companies like Galvanize (3 locations), Flatiron (just in New York). and Hack Reactor (The “Harvard” of bootcamps due to their incoming selection criteria). The most well-known (but not very big) coding school was “Lambda School.” Founded by Austin Allred, Lambda went through the Y Combinator program, which accelerated the growth of the company and also Allred’s personal brand (he has over 200,000 Twitter followers). His stories about students coming to Lambda — without enough money to afford basic things like shoes — and then leaving with six-figure job offers were legendary (at least in the Twitter-verse).
Lambda implemented a model where everything was free for students while they were still enrolled in courses. This model dramatically increased the number of potential students. Worries about the ability to pay for courses were eliminated. Lambda would make money by charging students AFTER graduation, using ISAs — Income Share Agreements.
In some circles, people have villainized ISAs as “indentured servitude” because they require graduates to pay Lambda a fixed percentage of their future wages. In practice, the characterization of ISAs as indentured servitude is misguided and inaccurate. The agreements are essentially a hedged financial instrument. Instead of receiving a student loan and then making payments in fixed installments after graduation — regardless of your financial success (or lack thereof) — ISAs hedge those payments and tie the amount paid to the graduate’s income. Lambda School has re-branded as Bloom Institute of Technology; their current ISA terms include 14% of a student’s income for four years after graduation, up to a cap of ~$42,000. Whether arrangements like that are a good deal or not depends on how much the individual expects to earn (versus how much they actually earn). In other words, the model is less like servitude and more like insurance.
From a provider’s perspective, ISAs cause all sorts of adverse selection effects. For students with no intention of looking for a job after graduation, ISAs are essentially a ticket for free skill development — there are no upfront costs or loans to repay. The ISA model also works well for motivated students interested in lower-paying opportunities (e.g., the non-profit sector or jobs with compensation based on equity rather than salary). If you are a prospective student that is (1) very confident you will earn a lot of money upon graduation, and (2) able to secure a traditional loan, then the loan is usually a better option than an ISA. As such, companies that offer ISAs are more likely — relative to tuition-based programs — to attract unmotivated students, people unable to get a bank loan, etc. But that’s the trade-off you make for increasing your pool of potential students. And for Lambda, they were able to take on the risk of accepting people with low-credit scores — who the banks wouldn’t approve — thanks to help from substantial VC funding (at absurd valuations). It all looks good until it doesn’t…
During my time at GA, our COO spent a great deal of time trying to make ISAs available for our students. Because of the adverse selection issues described in the previous paragraphs, it took him years of negotiating to arrange something we thought made sense financially for both us and our students. That payment option launched shortly after the Adecco acquisition, and I understand that it was helpful in driving up applicant conversion rate (helpful, but maybe not a game-changer).
GA’s most impactful strategy was shifting the program costs from the backs of potential students and onto the books of interested employers. All hiring managers know that finding quality candidates consumes both time and money. One of GA’s more successful pitches to companies emphasized this cost-saving appeal: “Instead of spending all those dollars on recruiting, what if GA did the recruiting for you? Tell us what type of candidates you are looking for, and we will recruit them, train them up to a basic level of skill, and then pass them onto you.” Companies could easily pay a recruiter $10,000 or more to hire a web developer; alternatively, they could pay us $10,000 and we could find a potential student (maybe a liberal arts graduate without any tech skills) and recruit them into a three-month GA course, at no personal cost. The arrangement stipulated that graduating students would agree to work for the specific company for some period of time (without any income sharing requirements). GA could afford a smaller payment from the company ($10K) than the student ($15K) because our costs were lower; it took a lot fewer marketing resources to recruit a student for a free program with almost guaranteed employment.
When GA launched part-time courses, students (usually) paid the program fees by themselves, but that shifted over time as more and more of the payments were being made by employers. And, quite obviously, all of the training in the B2B business itself was paid for by the employer.
In 2018, General Assembly was sold to Adecco, a large, publicly traded, global corporate training company. While all of GA’s prior valuations were based on their consumer revenue, Adecco bought the company almost entirely for the B2B business. From Adecco’s point of view, the consumer business was the moat that allowed GA to have a virtual monopoly in “real” training — programs that were market-tested rather than driven by HR. Adecco had (and still has) a massive global sales organization that targets buyers of corporate training. General Assembly became a unique product that the sales team could offer.
When I speak to VCs and PEs about this space, they generally begin by asking questions about the consumer business. By the time our call finishes, the smart ones understand that the real potential is connected to employer-paid programs. While there is a sizable segment of the population interested (and able) to pay for job-related training, that segment isn’t actually THAT large. Employers, on the other hand, are always willing to pay for a given level of human capital (usually in the form of wages or recruiting fees). And so, a training company that can provide market-tested, value-generating skills should be able to convince employers of the opportunity for reducing hiring costs and developing employee talent.
Can a given training company create that value and capture their fair share from employers? That, I guess, is the remaining $1000 per hour question.
Keep it simple,
Edward Nevraumont is a Senior Advisor with Warburg Pincus. The former CMO of General Assembly and A Place for Mom, Edward previously worked at Expedia and McKinsey & Company. He writes Marketing BS, a non-conventional look at marketing tactics and strategies and co-creates SuperSerious616, a non-conventional look at 1960s Marvel comic books.
Very interesting read. Thank you!
Two things I think boot camps should try that they haven't yet (or at least I don't think they have).
1. Why not poach from high school graduates who would likely go to college? Except instead of college, you advertise to them a Bootcamp. Screen them via SAT or ACT if need be. A small market since you'd need parents that can co-sign or pay, but certainly large enough to make some money off of.
2. 6-month boot camps that teach twice as much and make people more likely to get a FANNG job or something.