The online program management company 2U is hoping to take some of the risk out of offering degree programs online with an algorithm it says can predict whether the program will be a success for the company and the college.
The company has been in business since 2008, but last year 2U began making business decisions based on what it calls its program selection algorithm. The algorithm combines information from a number of different data sets: degree conferral and job growth data from the federal government, online search trends, rankings, geographic information, and data from the degree programs 2U already runs.
By crunching those numbers, 2U says it can make a prediction of how many students a program will enroll, potentially alerting the company to which programs it should take online and which to avoid. There are drawbacks -- the algorithm struggles to predict enrollment in emerging disciplines and fields where 2U doesn’t have any experience -- but CEO and co-founder Chip Paucek says the insights gained from the algorithm make it “the most critical component of the company’s success.”
Most online program management companies (commonly known as OPMs) follow a high-risk, high-reward business model. In order to lower colleges and universities’ barrier to entry into for-credit online education, many of the companies in the market assume start-up costs -- paying for course development, marketing, recruitment and more -- and then recoup their losses with revenue-sharing agreements. It is not unusual to see OPM companies take more than 50 percent of tuition revenue during the first few years following a program's launch.
And 2U is no different. When the company takes a college’s first vertical -- corporate lingo for program in a particular field, like business or health care -- online, it invests up to $10 million over the first four years of the program, Paucek said.
With millions of dollars on the line before any students have set foot in their virtual classrooms, OPM companies have every reason to keep their college clients happy to ensure a healthy return on investment.
That’s not always the case. Last fall, for example, the University of Florida canceled a massive contract with Pearson, which had worked with the university to launch UF Online. By one estimate, the decision meant Pearson missed out on up to $186 million over the contract’s 11-year span (though the company has said the figure is overblown).
2U, too, has seen its fair share of stumbles, including an ambitious but ultimately unsuccessful project involving online courses for undergraduates across a consortium of member universities. Paucek said the company has had to “learn the hard way” how to allocate its marketing dollars.
“Ultimately we have to drive a successful outcome for our partners, and that’s something I do worry about in the space,” Paucek said. “More people are just guessing, and they might guess and come up with the wrong number for the school. That hurts all online programs.”
2U, he added, is now “at the point where we’re definitely not guessing.”
Blind Spots and Business Strategies
Like most OPM companies, 2U has focused on graduate education. Within that segment, the company has targeted highly ranked programs at prestigious universities. Examples include Yale University’s physician assistant program (which is still awaiting accreditor approval) and the data science program at the University of California at Berkeley, among others.
But there is a finite number of top-tier programs out there, and 2U, a publicly traded company since 2014, is looking to grow. Some investors believe the company has the potential to triple its market value over the next five years.
In 2013 2U hired Harsha Mokkarala, formerly a digital marketing specialist in Capital One’s credit card acquisitions group, to help the company with college and student recruitment. Mokkarala now holds the title chief marketing officer.
Of the various data sources that make up the algorithm, Mokkarala singled out the company’s own data as the most important. He described the data as enrollment “benchmarks” for future degree programs.
“The only reason we are able to build this model is we have experience building programs,” Mokkarala said. “The machine just collects more and more data as we grow the programs.”
The algorithm also stresses regionality, Mokkarala said. In addition to looking at a particular college’s track record of producing graduates, the algorithm considers the state of that degree in a particular region. If, hypothetically, 2U wanted to take one of the University of Pennsylvania’s degree programs online, the company might look at data from Drexel and Temple Universities, for example. Most 2U programs, Mokkarala said, draw a majority of their enrollment from students who live up to a few hundred miles from the college.
Max Woolf, a vice president at the investment banking firm Tyton Partners, pointed out that looking for students locally isn’t new -- for-profit colleges have done it for years, opening dozens of campuses across the U.S. to create a local presence and grow their enrollments. He said OPM companies are bringing some of that “for-profit mentality” to nonprofit institutions, which could be beneficial if overall higher education enrollments continue to decline and competition for students increases.
With the data sources 2U is looking at, “you should be able to say, ‘Here’s an area where we can charge X amounts of dollars to either undercut or fill a void in the market,’ and you’re either taking students from other institutions within that region, or you’re taking ‘greenfield’ adult learners who are looking to go back to school,” Woolf said. “It’s the right way to think of how to approach the market.”
John Fallon, CEO of Pearson, said he was skeptical that OPM companies can guarantee outcomes for colleges. He declined to comment specifically on his company’s work with the University of Florida, but said OPM companies will “inevitably” strike some deals that work well and some that won’t.
“This is pioneering work, and it’s a sector of higher education in its relative infancy,” Fallon said. Most analysts agree the sector is going to continue growing, he said, but “anyone who professes to know exactly how it’s going to go is almost certainly going to be wrong.”
The algorithm does have some large blind spots, Mokkarala and Paucek acknowledged.
When 2U is looking at a degree in a field in which it is already active online -- like an M.B.A., which it already runs four of -- the company can make predictions “with extraordinarily high confidence,” Paucek said. It knows less when it enters a new field, such as when it last year launched Syracuse University’s master of science in communications -- the first of its kind available through 2U.
Worse yet are disciplines such as data science that have recently become popular but lack historical data. In those cases, 2U will try to use data such as search traffic as a proxy, “but the reality is the proxy isn’t as good,” Paucek said.
Of course, there are many other factors that could make or break a relationship between a college and an OPM company that the algorithm doesn’t take into account. Chief among them is the human element. A college may look like a perfect partner on paper, but offering a program online may be a nonstarter if the faculty lacks interest.
“At the end of the day, regardless of what the algorithm says, if we don’t have the full weight of the school [behind us], we have nothing,” Paucek said.