Last year Blackboard celebrated their 20th anniversary as a company, and this year Instructure celebrated their 10th. Yes, the company with Canvas, the “new” LMS solution, is a decade old. To gain historical perspective, it is interesting to compare each company’s first ten years in business in terms of primary financials – revenue and income – and performance in their core North American Higher Education market.
We can only go so far in comparing the companies on this basis, however, due to different circumstances.
Most importantly, Blackboard helped create the LMS market and therefore had significant portion of its expansion in a greenfield situation, picking up clients who had not previously used an institution-wide LMS. Canvas entered the market well after saturation, with greater than 90% of institutions already having a standard LMS.
Blackboard developed its product line pre-cloud, with on-premise hosting being the primary deployment model. Canvas was cloud-native, developed on top of AWS infrastructure.
Blackboard’s growth was heavily based on corporate acquisition of competitors – Prometheus, WebCourse, WebCT in the first decade – while Instructure’s growth has been almost exclusively organic.
Both companies sold in more than just the North American Higher Ed (NAHE) LMS market, even if that was the core. Blackboard acquired two companies in 2000 and launched a card-reader transaction business, and they played somewhat in the K-12 and corporate learning space. Instructure developed Bridge for corporate learning, Arc for video platform, and it has a heavy presence in the K-12 LMS market.
Interestingly, both companies went public in Year 7 (Blackboard in 2004, Instructure in 2015).
With that in mind, let’s compare revenue, operating income – both derived from public SEC filings and adjusted to 2018 dollars – and market share in their core NAHE market. The financial data comes from public filings, and data for 2018 for Instructure are estimates based on their outlook presented in the most recent quarterly earnings release. Market share data is based on our LMS Market Analysis service and associated financial analysis for premium subscribers.
Some notes:
Blackboard raised ~$185 million in venture financing pre-IPO while Instructure raised ~$95 million (both in 2018 dollars). Part of this difference can be attributed to the much higher costs associated with pre-cloud enterprise software deployment models.
Blackboard’s revenue was higher than Instructure’s at similar points, and their operating income was mostly positive after Year 6. Instructure has maintained a significant loss each year from operations. One way to explain this difference is that Blackboard’s used funds for corporate acquisitions, thus fueling growth, while Instructure invested funds into sales & marketing and organic growth. This is not the whole story, but it is a big difference between the two companies.
In market share represented as a percentage, Blackboard’s growth was much faster, but this was in a period where the market was developing and not all schools had a standard LMS. When viewing this market share in raw numbers, we see that Instructure’s growth would have exceeded Blackboard’s were it not for the Year 9 WebCT acquisition.
Blackboard was already showing organic market declines (in NAHE market share %) and slowdowns in raw numbers outside of the immediate WebCT acquisition of customers. Instructure’s organic growth shows no signs yet of slowing down, is less lumpy and more predictable.
In Spring 2015 Moody’s affirmed their B2 rating for Blackboard’s nearly $1.4 billion in debt as part of the company’s acquisition of Schoolwires, with a negative outlook for the ratings (meaning there was risk of a further downgrade). Three weeks ago, that downgrade took place.
(“Moody’s”) downgraded Blackboard Inc.’s Corporate Family Rating (“CFR”) by two notches to Caa1, from B2, as well as its Probability of Default Rating, to Caa1-PD, from B2-PD. Moody’s also downgraded Blackboard’s $135 million first-lien revolving credit facility and $920 million (remaining balance) first-lien term loan to B3, from B1, and its $378 million second-lien notes to Caa3, from Caa1. Moody’s also changed Blackboard’s outlook to stable, from negative.
Put in regular terminology, the previous B2 rating indicated that Blackboard ((Disclosure: Blackboard is a subscriber to our LMS market analysis service and is a sponsor for our recent Empirical Educator Project summit.)) had “the capacity to meet its financial commitments. Adverse business, financial, or economic conditions will likely impair the obligor’s capacity or willingness to meet its financial commitments.” But with the new Caa1 rating, Blackboard “is currently vulnerable, and is dependent upon favorable business, financial, and economic conditions to meet its financial commitments.”
The rationale for the downgrade is that Blackboard holds a very high amount of debt (now more than $1.4 billion) relative to its earnings, and revenue growth is not coming from its core markets.
Blackboard’s core North American Higher Education (“NAHE”) and K-12 units, representing 56% of total revenue, continue to show weakening top line results, suggesting that the success of its new Ultra user interface is still uncertain. Blackboard’s international segment, also weak, has shown modest stabilization of late. Only the campus enablement segment, consisting of recently acquired educational community communications and transaction processing services and representing a quarter of Blackboard’s revenues, has shown healthy, reliable revenue growth. Some ratings support is provided by Blackboard’s high level of revenue visibility, with three quarters of 2017 revenues coming from recurring products and services, and underpinned by its 90% renewal rates in 2017. Both of these measures, however, represent declines from prior years.
The turnaround is stalling, and the credit cards are maxed out. Blackboard has continued to cut costs, including “late-year layoffs”, and Moody’s expects “overall revenues to be flat to down slightly in 2018 as competitors have stifled market share gains”.
While this is not good news for Blackboard, the ratings action does give the ed tech community additional insight into the operations and health of the company. Blackboard total revenues are between $700 – $720 million, with earnings (EBITDA, adjusted by Moody’s) between $160 – $180 million.
And in the rationale for the ratings comes specific commentary on Learn Ultra.
Software renewals have been weaker than expected in the NAHE [North American Higher Ed] segment, as the company strives to sell its latest learning management system (“LMS”) software enhancement, Ultra, into a crowded and very competitive marketplace. Given operating seasonality tied to the academic year, the behind-schedule launch of Ultra, in mid-2016, meant that measurable revenue and EBITDA [Earnings before interest, tax, depreciation and amortization] contributions from it began to be realized only in the 2017 academic year. There are indications that Ultra is gaining traction relative to Canvas and Desire2Learn, and Moody’s believes the packaging of Ultra with transaction- and payment-processing services may support its competitive positioning. But the success of Ultra is far from certain, and the threat from existing and possibly new competitors remains high as barriers to entry, specifically for web-based software, are relatively low. Meanwhile Blackboard, with a brand new CFO, is focusing its research, sales and marketing, and product development resources in an effort to ensure Ultra’s future.
We have covered the 2014 announcement and ongoingfate of LearnUltra here at e-Literate, and claiming a “launch of Ultra, in mid-2016” is generous at best. At that point there were educator previews with no ability for institutional adoptions. Fall 2017 is the closest to what I would say is an actual launch. As of a January, 2018 meeting Michael and I had with Blackboard’s management team, they claimed dozens of schools actively piloting Learn Ultra ((I have asked several times for this list, or a subset of this list, of schools piloting Ultra to allow interviews. If and when we receive schools to interview, we will cover in another post.)), and their flagship customer University of Phoenix is beginning rollout of the LMS starting this month.
What is going well at Blackboard is the movement to a SaaS (software as a service, aka ‘the cloud’) model for the LMS – for both the traditional experience and Ultra experience. As of January there were 284 schools on Learn SaaS and 49 others in migration. In a purchase-only report, Moody’s acknowledged this strength and noted the investment that Blackboard is making in this area (more than $50 m expected in 2018).
While the company has shown good progress in migrating its legacy Learn customers onto the appropriate SaaS-based platform from which those customers may choose, in turn, to migrate to Ultra, Blackboard’s efforts to make those transitions smooth for its customers have entailed elevated capital expenditures, which will likely continue through 2019, cutting into free cash flow.
Put this all together, and 2018 is the year that Blackboard needs to transition from ‘wait until Learn Ultra is ready’ to ‘Learn Ultra had better lead to increased sales’. I have been impressed with the new management team’s transparent approach to dealing with analysts, and with their honest approach to understanding the problems they need to solve. But the company needs to deliver, and this is shaping up to be a newsworthy year for Blackboard, for good or ill. There’s a lot to watch here.
Nearly two years ago I wrote two blogpostsgiving a high-level view of the Online Program Management (OPM) market landscape. This is a growing but messy market, and the market changes since mid 2016 call for an updated view.
OPM providers are for-profit organizations that help non-profit schools develop online programs, most often for Master’s level programs. These companies provide various services for which traditional institutions historically have not had the experience or organizational capability to fully support, at least for fully-online programs and often for non-traditional student populations. Some examples of the services include marketing & recruitment, enrollment management, curriculum development, online course design, student retention support, technology infrastructure, and student & faculty call center support.
The OPM market has historically been known for a full-service, revenue-sharing model, based on the premise that most traditional institutions are not only operationally unprepared to offer online programs at scale but also are not set up to invest in online programs up front. There are extensive costs, particularly in marketing and recruitment as well as curriculum and course design, that cause most scalable online programs (that is, those designed with the intent and infrastructure to allow more than just a few dozen students) to require investment over the first several years, before tuition revenue catches up. Rather than requiring the institution to spend sizable up-front money without a guarantee of repayment, revenue-sharing OPM vendors provide this financing themselves – which is in itself an expensive proposition. It often takes three to five years for an OPM company to become profitable for any online program, which is why they often require 10-year or even longer contracts.
As the online education market has matured, however, there has been a growing pushback against revenue-sharing as the only model available. Thus there is an emerging unbundled fee-for-service OPM model, in which the companies offer the same services, or some subset, for the market price of those services. The institution pays for the services used, mostly independent of the amount of tuition revenue coming into the online program. This category leads to the program, or institution, to take the up-front financial risk but not have to sign contracts sending ~40 – 60% of the tuition revenue to the vendor. Fewer strings attached but more responsibilities and risks for the school.
With the recent success of Southern New Hampshire University and longer-term models from Penn State World Campus (as well as others), there are also questions about whether non-profit schools can create scalable programs without an OPM partner at all. This category, for obvious reasons, is not presented in our market landscape but is a real option.
In our view, there is not an either-or proposition in the market. We see strong arguments for both bundled revenue-sharing models and for unbundled fee-for-service models, and I personally do not believe that the market is moving away from revenue sharing as much as there is pressure for additional models. There are a growing number of choices available to schools, but there is also a crowded marketplace that is becoming more difficult to understand.
With that setup in mind, on to the market landscape graphic.
As before, please note that this view is intended to give a visual overview of the market landscape and is not comprehensive in terms of vendors represented. This is especially important to understand for the Fee for Service category, where there is no clear definition yet of what is an unbundled OPM provider and what is a product or service provider who happens to serve online programs. Treat this category as illustrative and not exhaustive.
There are three primary changes to note in the graphic’s design since 2016:
We have combined different categories of Certificates into one column on the right side of the graphic, since it was difficult to to differentiate between OPM target markets.
We have added a sub-category of MOOC Providers under Full Service Revenue Share category to capture the recent changes in that market.
We have combined the Fee for Service and General Contractor into one category labeled Fee for Service, again due to difficulty differentiating service offerings.
We should also note that there will likely be another sub-category of Full Service Revenue Share to capture the ongoing pivots by former for-profit institutions. Graham Holdings sold Kaplan University to Purdue University and remains a single-client OPM provider for that new brand Purdue University Global. Grand Canyon University is converting to a non-profit institution with the remaining for-profit company turning into a (single-client?) OPM provider. Bridgepoint Education plans to turn Ashford University and University of the Rockies into a combined non-profit institution, leaving the remaining for-profit company as an OPM. We are not ready to add this sub-category, and it is not clear yet if these providers plan to go beyond servicing the non-profit conversions.
There are several changes to the vendors listed:
Pearson no longer refers to the Embanet brand that they acquired and now goes by Pearson ((Disclosure: Pearson, 2U, iDesign, Unicon, and Blackboard are sponsors for our Empirical Educator Project. See this post for more details.)) Online Learning Services.
2U, after the May 2017 acquisition of GetSmarter, now serves certificate programs and not just masters degree programs.
Wiley no longer refers to the Deltak brand that they acquired and now goes by Wiley Education Services.
DeVry Integrated Education Solutions no longer appears to be active in the market, based on outdated web site (key executive listed has been gone for seven months) and no marketing or competitive presence. We have removed their listing.
Cambridge Education Group’s CEG Digital unit has emerged as an OPM provider serving UK-based institutions.
We (re)added Udacity based on the success of the Georgia Tech Masters in Online Computer Science program, which has helped spawn the new sub-category of MOOC Providers in the market, even though Udacity appears to have no interest in pursuing other higher education degree offerings.
We added Coursera as an OPM Provider, as this market now appears to be their core market moving forward.
We added FutureLearn based on their continuing expansion as a MOOC-based OPM provider.
We added ExtensionEngine as a Fee for Service provider based on their work with Notre Dame and related program creation.
We changed Greenwood Hall to AnswerNet. After a spectacular implosion from Greenwood Hall a few months ago, AnswerNet acquired the remaining assets. It is not yet fully clear if AnswerNet will not provide more than Call Center services, so we’ll keep an eye on this one.
Excelsior College’s Educators Serving Educators (ESE) initiative, launched in 2012 to provide a non-profit OPM offering, is no longer active.
We’ll keep updating our landscape diagram over time and look for other methods to help make sense of this market.
Two weeks into March, this has already been a busy month already for the transformation of for-profits and MOOCs. For-profit universities are in a race to become nonprofit by separating academic programs from behind-the-scenes services, and MOOCs are focused primarily on monetization and moving beyond free and open courses. The common thread tying these messy transitions together is the move to become new forms of Online Program Management (OPM) providers.
Best Way to Make Money? Go Nonprofit
Arguably the biggest news was March 5th when the Higher Learning Commission (HLC), the regional accreditor, approved the Kaplan University / Purdue University deal to create Purdue Global. This was the final approval step as Purdue acquires Kaplan, leaving Kaplan University, leaving Graham Holdings (Kaplan’s parent company) to serve as a single-client OPM provider.
The following day Grand Canyon University announced that it had received approval from HLC to convert into a nonprofit institution. As described in their press release, the remaining for-profit company will become an OPM, even if they choose not to use that name [emphasis added]:
As part of the transition, GCE will sell certain academic-related assets to a non-profit entity that will carry the Grand Canyon University name. Following the sale, GCE will operate as a third-party provider of educational and related services to GCU and potentially, in the future, to other universities. The structure is similar to that at hundreds of non-profit universities in the country that outsource services to third-party providers.
And yesterday, Bridgepoint Education announced that they were formally seeking to convert Ashford University into a nonprofit in a similar deal as Grand Canyon. At least they are more direct about the OPM tie-in as described at Inside Higher Ed yesterday.
Bridgepoint will continue on as an online program management (OPM) provider — a booming space in higher education. The company will negotiate with Ashford to enter into a shared services agreement, with Bridgepoint likely handling data management, course management software and services, technology, and financial aid processing for the nonprofit university.
“As an OPM, Bridgepoint Education will bring years of technological and academic innovation and intellectual property development to other colleges and universities that desire to serve students through online education programs,” Schray said in a written statement.
In an interview here at e-Literate when Purdue and Kaplan announced their acquisition plans last April, Trace Urdan (now at Tyton Partners) described the market forces involved in some of these moves.
Non-profit entities – both public institutions and private non-profit institutions – “wanting to get into the adult market and the online market”. This is the big push behind the Online Program Management (OPM) market, kick-starting these non-profits into online programs targeting adult education.
For-profit entities “feel like they are being burdened by being for-profit”. One part of this is the regulatory burden from the Department of Education and even accreditors. But there is also a marketplace burden as non-profits like Southern New Hampshire University keep growing enrollments while for-profits are dropping.
There is a “the investor enthusiasm for the services model” with OPMS, “and this is a model that investors love – it gives you access to the growth in online education, affiliation with strong brands, and it’s more or less free from the regulatory hostility” of the for-profit sector.
So, why did this happen? First, and most obviously, we are in a different regulatory environment – at least as far as the federal Department of Education is concerned. In late 2017, the Department of Education dropped its opposition to for-profit conversions vehemently articulated by then-Secretary John King, most recently approving the sale of South and Argosy Universities and the Art Institutes owned by Education Management Corporation to a nonprofit created by the Dream Center Foundation. This change in federal policy shifted the emphasis on approvals back to the accreditors and the states.
At the accreditor level, the politics may be less important in understanding the outcome than the process. At about the same time that Kaplan-Purdue was first announced, HLC began working on revising its policies and procedures to establish new benchmarks by which such transactions would be measured. HLC made two significant changes: it updated its procedures for review of Change of Control transactions and, in a politically astute move, also established a policy that Department of Education approval must be obtained before HLC acts on a change of control application, thus insulating itself from second-guessing in Washington. (HLC’s change was telegraphed in late 2016 when it deferred acting on the sale of the parent of the University of Phoenix to a private equity group pending prior ED approval.)
Significantly, HLC now has relatively clear guidance governing what is needed for OPM relationships and for-profit conversions.
How Do We Make Money? The Answer Is Simple – Volume
Meanwhile Coursera announced their plans to further focus on the monetization of supporting online degrees, as described at EdSurge on March 5th.
These days, though, many MOOC platforms are courting the traditional higher-ed market they once rebuked, often by hosting fully-online masters degrees for colleges and universities. And today, one of the largest MOOC providers, Coursera, announced it’s going one step further in that direction, with its first fully online bachelor’s degree.
Coursera is not alone here – most notably Georgia Tech and Udacity launched an online master’s of computer science in 2013. In a related move, edX has begun its work supporting online master’s degrees through its MicroMasters program, and FutureLearn – spun out of the Open University of the UK – supports multiple degree programs.
While the Coursera news focused on the new bachelor’s program, the bigger news was the expansion its graduate programs as described at Inside Higher Ed.
Online education platform Coursera has set a goal of offering 15 to 20 degree programs by the end of 2019. The company took another step toward that goal Wednesday, announcing new degree offerings from the University of Illinois at Urbana-Champaign and France’s HEC Paris.
“This is our coming-out party for online degrees on Coursera,” Nikhil Sinha, Coursera’s chief business officer, said in an interview.
For the MOOC providers, their move into the OPM space seems to be driven by their leverage of current registered learners as a marketing channel, as described in a separate IHE article.
Roughly half of the students in Coursera’s current degree programs took one of the open online courses first, essentially enabling students to “try these degrees before they buy them,” Maggioncalda says. So not only do students have a chance to see how they like a professor, or how well they perform, before enrolling in the for-credit program, but Coursera also asserts that it can drive down the cost of acquisition of students by tapping into its 31 million users.
Coursera’s institutional partners “share a certain percentage of the learner fee with us in exchange for distribution to our world of learners, and the whole delivery of the system on our platform,” Maggioncalda says.
New OPM Models
Two years ago we described how the OPM market has evolved beyond its full-service tuition revenue-sharing origins to add unbundled service offerings – not to replace the previous model but to augment it. What we are now seeing are two new models within the OPM market becoming much more clear: the remnants of for-profit conversions into nonprofit status, and MOOCs supporting online degree programs. Both of these models are driven by markets that need to move beyond their origins as well. A lot of changes happening in the education space.
In a 2016 post on “Personalized Learning vs. Adaptive Learning”, Michael contrasted the former as a set of technology-enabled teaching practices with the latter as a project label [emphasis added].
On the other hand, adaptive learning is a label that applies to products. Further, adaptive learning products can support all of the practice areas of personalized learning. They enable teachers to move content broadcast outside of class time, they make homework time into contact time through analytics, and they provide some tutoring function. “Adaptive” tends to provoke a lot of discussion around the latter of the three practice areas. See the piece I wrote for the American Federation of Teachers if you want a primer on the strengths and limitations of adaptive learning products as tutors. But as often as not the first two capabilities, neither of which is dependent on adaptive algorithms, are the ones that enable the biggest gains in personalized learning teaching practices. “Courseware” is a set of products that, when designed well and used properly, can enable faculty to move content broadcast out of the classroom and make homework time content time. Adaptive courseware adds the tutoring element while also, if done well, increasing the value of that homework contact time by providing better feedback through more targeted analytics.
One problem, of course, is that many advocates end up presenting adaptive learning as a silver bullet, focusing on supposed magical tutoring capabilities and ignoring the messy but valuable work of enabling faculty to improve their own courses. As we will see, it is not a safe assumption that all vendors like this label or set of assumptions.
In the first episode of this series we explored the challenge of going beyond pilots and deploying systems at scale. In the second episode we explored the importance of increased visibility as an important benefit of course redesign. In this episode we explore the challenge of silver bullets – what happens when a company has no intention of being labeled as “adaptive learning” but has been described that way? Realizeit is not the only company dealing with this same issue, and it is important for educators to understand how a company defines itself and the problems their products are designed to solve. ((This post is not meant to endorse Realizeit’s platform over other companies’ platforms. We are focusing on institutional perspectives and lessons to be learned. Realizeit is one of our sponsors of the Empirical Educator Project.))
(source: https://youtu.be/xPWBoMhP_yU)
This post is part of our e-Literate TV series, which is funded in part by the Bill & Melinda Gates Foundation. The findings and conclusions (or views) contained within are those of the authors and do not necessarily reflect positions or policies of the Bill & Melinda Gates Foundation.
Doug Lederman at Inside Higher Ed wrote an excellent article describing the trajectory and demise of the “failed $75 million experiment” of the University of Texas System’s Institute for Transformational Learning (ITL).
Starting in 2011, the Texas system invested nearly $100 million in the institute ($23 million remains unspent) to try to drive digital technologies into the approaches its campuses use to reach, educate and graduate students. Over five years, the institute helped several UT campuses launch distinctive new academic programs and developed three core pieces of technology that, among other things, deliver online learning and students’ transcript information via the blockchain. But its total revenue over the five years: $1 million, The Texas Tribune reported.
Critics have derided that output as paltry given the investment and questioned whether the money could have been better spent. Advocates for the institute say it was the sort of necessary and worthwhile experiment that higher education institutions undertake all too rarely, even as they acknowledge a set of miscalculations that — along with political and cultural circumstances in part beyond its control — contributed to its undoing.
Among those missteps, according to numerous people who worked closely with the institute, were a tendency to approach campus faculty members with arrogance rather than as partners, and a lack of clarity from the start about the endeavor’s goals and business plan. Those were compounded by a problem that afflicts many educational technology efforts, whether funded by Silicon Valley investors or campus leaders: too little patience for what is almost always a long, drawn-out process of adoption and proof.
It’s worth reading the whole article, and sprinkled throughout are quotes from faculty and staff from the University of Texas system – ITL’s natural customer base – about overly optimistic claims, inability to understand or listen to its stakeholders, and real or perceived arrogance. I observed similar reactions when on the UT Austin campus for a video case study in Fall 2016 about their attempts to reinvent the large lecture class. When I asked about the overlap or collaboration with ITL on what seemed to be a perfect fit for that group’s objectives, I got eye rolls and comments about ITL being irrelevant for campus-based innovation efforts. Regardless of how that situation arose, the point is the same – there seems to have been a real breakdown of communication and trust between ITL and UT campus educators, despite the fact that “some of the work done by the institute has been divested to the various UT campuses.”
ITL was not alone as a failed, or at least abandoned, initiative to transform education based on innovative ed tech, and there is a theme in general of focusing more on the supposed tech-based innovation rather than the issue of cultural change and understanding social barriers (and this includes a culture’s view on whether the innovation is worth adopting). This is not a new subject, and similar challenges from a vendor standpoint led to the 2014 post “Pilots: Too many ed tech innovations stuck in purgatory”, an adaptation of which forms the second half of this post.
The general interest on dealing with the social change issues around real adoption of innovative education approaches – as well as the associated culture required to test theories, honestly assess results, and make continuous improvements in teaching practices – is what has led us at e-Literate to work on the Empirical Educator Project that Michael described a few weeks ago.
Despite the billions of dollars invested and provided in grants over the past several years, the vast majority of ed tech is used in only a small percentage of courses at most campuses. ((I’m not arguing against faculty prerogative in technology adoption and for a centralized, mandatory approach, but noting the disconnect.)) Most ed tech applications or devices have failed to cross the barriers into mainstream adoption within an institution. This could be due to the technology not really addressing problems that faculty or students face, a lack of awareness and support for the technology, or even faculty or student resistance to the innovation. Whatever the barrier, the situation we see far too often is a breakdown in technology helping the majority of faculty or courses.
Diffusion of Innovations – Back to the basics
Everett Rogers wrote the book on the spread of innovations within an organization or cultural group in his book Diffusions of Innovations. Rogers’ work led to many concepts that we seem to take for granted, such as the S-curve of adoption:
Source: The Diffusion of Innovations, 5th ed, p. 11
leading to the categorization of adopters (innovators, early adopters, early majority, late majority, laggards), and the combined technology adoption curve.
Source: The Diffusion of Innovations, 5th ed., p. 281
But Rogers did not set out to describe the diffusion of innovations as an automatic process following a pre-defined path. The real origin of his work was trying to understand why some innovations end up spreading throughout a social group while others do not, somewhat independent of whether the innovation could be thought of as a “good idea”. From the first paragraph of the 5th edition:
Getting a new idea adopted, even when it has obvious advantages, is difficult. Many innovations require a lengthy period of many years from the time when they become available to the time when they are widely adopted. Therefore, a common problem for many individuals and organizations is how to speed up the rate of diffusion of an innovation.
Rogers defined diffusion as “a special type of communication in which the messages are about a new idea” (p. 6), and he focused much of the book on the Innovation-Decision Process. This gets to the key point that availability of a new idea is not enough; rather, diffusion is more dependent on the communication and decision-process about whether and how to adopt the new idea. This process is shown below (p. 170):
Source: The Diffusion of Innovations, 5th ed., p. 170
What we are seeing in ed tech transformation initiatives in most cases, I would argue, is that the new ideas (applications, products, services) are stuck the Persuasion stage. There is knowledge and application amongst some early adopters in small-scale pilots, and there are good motivations in general from transformation groups like ITL and vendors, but majority of educators either have no knowledge of the innovation or are not persuaded that the idea is to their advantage, and there is little support or structure to get the organization at large (i.e. the majority of faculty for a traditional institution, or perhaps for central academic technology organization) to make a considered decision. It’s important to note that in many cases, the innovation should not be spread to the majority, either due to being a poor solution or even due to organizational dynamics based on how the innovation is introduced.
The Purgatory of Transformation
This stuck process ends up as an ed tech purgatory – with promises and potential of the heaven of full institutional adoption with meaningful results to follow, but also with the peril of either never getting out of purgatory or outright rejection over time.
Ed tech innovators can often be too susceptible to being persuaded by simple adoption numbers such as 1,100 institutions or total number of end users (millions served), or by new the promise of ideas that look great on paper, but meaningful adoption within an institution – actually affecting the majority of faculty or courses – is necessary in most cases before there can be any meaningful results beyond anecdotes or marketing stories. The reason for the extended purgatory is most often related to people issues and communications, and the ed tech market (and here I’m including transformation initiatives, vendors, consultants, and campus support staff and faculty) has been very ineffective in dealing with real people at real institutions beyond the initial audience for an idea.
As Larry Cuban stated in his commentary about Larry Berger’s “confession” that the engineeriomy model of personalized learning is flawed (also covered here at e-Literate):
First, those confessing their errors about solving school problems seldom looked at previous generations of reformers seeking major changes in schools.They were ahistorical. They thought that they knew better than other very smart people who had earlier sought to solve problems in schooling.
Second, the confessions seldom go beyond blaming their own flawed thinking (or others who failed to carry out their instructions) and coming to realize the obvious: schooling is far more complex a human institution than they had ever considered.
Finally, few of these confessions take a step back to not only consider the complexity of schooling and its many moving parts but also the political, social, and economic structures that keep it in place (see Audrey Watters here). As I and many others have said often, schools are political institutions deeply entangled in American society, culture, and democracy. Keeping the macro and micro-perspectives in sight is a must for those seeking major changes in how teachers teach or how schools educate. Were that to occur the incidence of after-the-reform regret might decrease.
Cover image credit: Ludovico Carracci [Public domain], via Wikimedia Commons
Let’s be honest: as an academic term of art, personalized learning is horrible. It has almost no descriptive value. What does it mean to “personalize” learning? Isn’t learning, which is done by individual learners, inherently personal? What would it mean to personalize learning? And who would want unpersonalized learning? Because the term carries so little semantic weight, it is a natural for marketing purposes: “Our personalized learning is new, improved, and 99.44% pure!” Unfortunately, this also sets it up perfectly for the inevitable War of Definitions. Remember the Great MOOC War a few years ago? Were MOOCs the creation of the Canadian Constructivists or of the Stanford professor who invented a self-driving car? Are we talking about an xMOOC or a cMOOC? Which one is the good one, and which one is the bad one? Now that the furor has died down, there is relatively little debating over the definition of the term MOOC and much more focus on how the family of approaches that are collected under that term can best serve different educational purposes.
Let’s just skip to the end this time, shall we?
The two of us spent the past three years visiting colleges and universities that have undertaken so-called personalized learning projects, and we talked to the students, teachers, and administrators about what they are actually doing and why they are doing it. We visited a wide range of institutions and talked to a wide range of stakeholders, based on our daily work as consultants to colleges and universities and as analysts of the educational technology industry and our work on a grant funded by the Bill & Melinda Gates Foundation. Through these observations, we have been looking for the ground truth underneath the hype of personalized learning. As a result of this process, we observed a family of technology-enabled educational practices that are potentially useful for a range of educational challenges. We would like to share our framework, which we hope will be useful for thinking about (1) the circumstances under which personalized learning can help students and (2) the best way to evaluate the real educational value for products that are marketed under the personalized learning banner.
Personalized Learning as Practice
Imagine for a moment that personalized learning is not already a term in the ed tech lexicon and, further, that there is no need for any new term to be “catchy” or “sticky.” The most descriptive label we could come up with for the practices that the two of us have observed in our school visits might be undepersonalized teaching. If the ideal, most personal teaching modality is one-to-one tutoring, there are many reasons why we fall short of this ideal in real-world classrooms. The most stereotypical depersonalized teaching experience is the large lecture class, but there are many other situations in which teachers do not connect with individual students and/or meet the students’ specific needs. For example, even a small class might contain students with a wide-enough range of skills, aptitudes, and needs that the teacher cannot possibly serve them all equally well. Or a student may have needs (or aptitudes) that the teacher simply doesn’t get an opportunity to see within the amount of contact time that the class allows. The truth is that students fall through the cracks all the time, even in the best classes taught by the best teachers. Failing a course is the most visible evidence, but more often students drift through the class and earn a passing grade—maybe even a good grade—without getting any lasting educational benefit.
If we choose to think of personalized learning as a practice rather than a product, we can start by taking a hard look at course designs and identifying those areas that fail to make meaningful individual contact with students. These gaps will be different from course to course, subject to subject, student population to student population, and teacher to teacher. Although there is no generic answer to the question of where students are most likely to fall through the cracks in a course, there are some patterns to look for (as we will discuss later in this article).
Technology then becomes an enabler for increasing meaningful personal contact. In our observations, we have seen three main technology-enabled strategies for lowering classroom barriers to one-on-one teacher/student (and student/student) interactions:
Moving content broadcast out of the classroom: Even in relatively small classes, a lot of class time can be taken up with content broadcast such as lectures and announcements. Personalized learning strategies often try to move as much broadcast out of class time as possible in order to make room for more conversation. This strategy is sometimes called “flipping” because it is commonly accomplished by having the teacher record the lectures they would normally give in class and assign the lecture videos as homework, but it can be accomplished in other ways as well, for example with reading-based or problem-based course designs.
Turning homework time into contact time: In a traditional class, much of the work that the students do is invisible to the teacher. For some aspects, such as homework problems, teachers can observe the results but are often severely limited by time constraints. In other cases, such as comprehension of assigned readings, the students’ work is invisible to the teacher and can be observed only indirectly and with significant effort. Personalized learning approaches often allow the teacher to observe the students’ work in digital products, so that there is more opportunity to coach students. Further, personalized learning often identifies meaningful trends in a student’s work and calls the attention of both teacher and student to those trends through analytics.
Providing tutoring: Sometimes students get stuck in problem areas that don’t require help from a skilled human instructor. Although software isn’t good at teaching everything, it can be good at teaching some things. Personalized learning approaches can offload the tutoring for those topics to adaptive learning software that gives students interactive feedback while also turning the students’ work into contact time by making it observable to the teacher at a glance through analytics.
None of these techniques, by themselves, undepersonalize the teaching. They generally need to be designed and implemented by skilled educators as part of a larger course design that is intended to address the particular problems of particular students. In the business world, an analogous initiative might be called “business process redesign.” Emphasis is on process. The primary question being asked is, “What is the most effective way to accomplish the goal?” The redesigned process may well need software, but it is the process itself that matters. In personalized learning, the process we are redesigning is that of teaching individual students what they need to learn from a class as effectively as possible (though we can easily imagine applying the same kind of exercise to improving advising, course registration, or any other important function).
We saw a noteworthy example of this educational design work in action at Essex County College (ECC) in Newark, New Jersey. The majority of ECC students need to take developmental math in order to complete their degrees, and the majority of those do not pass developmental math. Of those who do, the majority do not pass the college-level math course that follows. ECC leadership believed that this educational failure could be attributed to two main factors. First, students came into developmental math with an enormous range of prior knowledge: some had the equivalent of a fourth-grade math education; others needed to learn only a few concepts. Students on one end of the spectrum typically got lost because they were not receiving the individual help they needed, whereas those on the other end often got bored and eventually failed or dropped out because they were being forced to spend a lot of their time on skills that they had already mastered. Second, many ECC students had never been taught good study skills, and faculty did not have the class time needed to teach those skills. So to address the two personalization gaps in this particular course, the college redesigned developmental math using personalized learning techniques.
ECC used an overall pedagogical framework called Self-Regulated Learning. Students in the course spend part of their class time in a computer lab, working at their own pace through an adaptive learning math program. Students who already know much of the content can move through it quickly, giving them more time to master the concepts that they have yet to learn. Students who have more to learn can take their time and get tutoring and reinforcement from the software. Teachers, now freed from the task of lecturing, roam the room and give individual attention to those students who need it. They can also see how students are doing, individually and as a class, through the software’s analytics. But the course has another critical component that takes place outside the computer lab, separate from the technology. Every week, the teachers meet with the students to discuss learning goals and strategies. Students review the goals they set the previous week, discuss their progress toward those goals, evaluate whether the strategies they used helped them, and develop new goals for the next week.
Note the role of the software in this design. In the lab, it primarily takes on the role of tutor, helping most of the students most of the time with routine skill coaching and practice so that the teacher is freed up to give individual attention to those students who really need it. In the goal-setting sessions, the software acts mainly as a record keeper. It helps students track their time on task, number of problems solved, and so on. The teacher then helps students figure out what to do with that information. In both cases, the software is an important enabler of the new teaching practices. But the value that it adds is quite different from the way personalized learning software products are often characterized by sales reps, marketing materials, and many news stories. It is thus worth taking a minor detour from our exploration of personalized learning as a practice to examine the significant gap between the ground truth of the practice and the popular characterizations of the products.
Why Such Hype?
So far, we have deliberately and explicitly set aside the various policy, political, and business pressures that have brought the term personalized learning into broader use so that we could focus on the educational value that lurks underneath the hype. But it is also important to understand these pressures so that we can be on guard for the ways in which they might deform the discussion and distract us from the real value that we should be talking about. None of the three approaches that we identified above are particularly new; nor do they require fancy algorithms and expensive products to achieve. There are two specific reasons why these approaches are being attached to heavily marketed products right now.
First, on the policy side, there has been a shift in emphasis from access to degree completion. President Barack Obama set the tone by announcing a goal that the United States be number one in the world in the proportion of college graduates by 2020. Since then, state and federal policy makers have followed suit. Colleges and universities now have to account for gainful employment metrics and track their institutional scorecard results. Base funding, particularly for public institutions, is increasingly tied to performance against these and similar metrics. Grant funding is also increasingly outcomes-driven. As higher education institutions have narrowed their focus on these metrics, the students who fall through the cracks and fail out of the standard educational model have come into sharper relief.
With these policy changes and the funding that follows them, being “student-centric” is no longer a nice-to-have goal. Rather, it is a critical success factor for improving measurable student outcomes and therefore getting funding and being seen as a successful institution. For example, whereas in the past a few forward-thinking community college administrators might have thought to take on a project like ECC’s personalized learning developmental math redesign out of a sense of mission, now every community college administrator must be looking for ways to improve degree completion by eliminating failure traps such as developmental math. The security of institutional funding depends on it.
The second big change has been the widespread commercialization of the adaptive learning techniques that have existed in educational research laboratories for over fifty years. As the term adaptive learning suggests, these products provide students with a certain amount of one-on-one tutoring (although the methods that these systems use for analyzing students’ progress and providing useful feedback vary widely by product and discipline). This change in the market has been enabled by technological advances that are increasingly resulting in one networked computer for every student in a class and affordable developer access to machine learning tools.
Market forces are playing a big role in publishing too. Textbook publishers have found that their traditional business model is collapsing as more students find ways to avoid buying new textbooks. Cengage, for example, was forced to go through bankruptcy. McGraw-Hill Education was sold to a private equity firm. Pearson’s stock is near historic lows. All of these companies have had multiple rounds of major layoffs. They are in desperate need of a new product, and they are increasingly latching onto the personalized learning trend as the cure for what ails them. Meanwhile, according to Ambient Insight, U.S. ed tech companies received $3.6 billion of angel and venture capital funding in 2015. (This doesn’t even include mergers and acquisitions, which have also been huge.) Startup founders find themselves in an increasingly crowded field and are under strong pressure to promise and produce big results. Every vendor of a developmental math product, whether that vendor is an established textbook publisher or a young startup, is aware that campus presidents and provosts need to solve their degree-completion problem and that developmental math is very likely to be a big part of that problem. The vendors therefore market their products as a solution to degree completion. Every textbook vendor and aspiring textbook disruptor knows that stories about improving pass rates through technology sell. But what to call these products? Personalized learning is a term that sounds good without the inconvenience of having any obviously specific pedagogical meaning, so it becomes the flag that all vendors fly, even though different products do very different things and even though undepersonalization is rarely accomplished through software alone.
Thus, through policy and commercialization, the personalized learning marketing juggernaut was born. Unfortunately, the combination of the marketing shortcuts and the funding pressures created a strong temptation for magical thinking. Campus leaders are being asked to believe that they can solve their degree-completion and other accountability metric problems by buying software that will somehow magically provide personalized learning (in a way that faculty members, by implication, do not). For obvious reasons, faculty are likely to reject this stunted conception of personalized learning. Leaders who want to see their campus communities benefit from personalized learning approaches need to guard against product-centric characterizations and should suggest that discussions of vended solutions take place in the context of course and curricular designs that undepersonalize teaching. Otherwise, the baby will probably get thrown out with the bathwater.
Good Candidate Opportunities
One of the benefits of reframing personalized learning as undepersonalized teaching and focusing on the three techniques we outlined earlier is that faculty can readily translate this framework into their own contexts and start identifying opportunities that are good candidates for undepersonalization (not all of which will require vended products). In contrast to a product-centric conception of personalized learning, a practice-centered conception is something that faculty can own. That said, we have seen areas of opportunity where personalized learning is often a good fit, and not all of those areas are always obvious.
To begin with, any course that students enter with a wide range of prior knowledge and ability is a good candidate. ECC’s developmental math is a prototypical example. Another example is Austin Community College’s ACCelerator lab, used heavily for developmental math courses. But the course doesn’t have to be remedial and the institution doesn’t have to be access-oriented in order for personalized learning to be helpful. For example, at Middlebury College, a geography professor realized that some students in his course on geographical information systems (GIS) were struggling. For context, this is a general education course that is taken by students in a wide range of majors and specialties. In an elite college like Middlebury, nobody worries too much about completion rates, particularly at the institutional level. The professor simply observed that some students were working very hard. In fact, the course had such a strong reputation for difficulty that taking and passing it was considered a badge of honor. Students were sleeping in the labs. But the professor didn’t see this as a sign that he was inspiring his students to work hard. He saw it, rather, as a course-design problem. Some students were working harder than they should because he wasn’t reaching them in the way that they needed to be reached.
As it turns out, one critical skill that the course was teaching—spatial reasoning—is rarely taught in high school. A handful of the students in the class came in with either prior training or natural talent. They did well. The others were the ones who were sleeping in the labs. They needed more time and more help. The professor decided to make videos of his lectures and assign the videos as homework. With this change, struggling students could watch the videos as often as they needed in the (relative) comfort of their own dorm rooms while the professor’s class time was also freed up for more interactive work. (After he had done this, a colleague told the professor that this technique is called “flipping the classroom”—a term he had never heard before.) He is also thinking about developing tutorial software that can help students work through the homework problems in ways that best suit their needs.
Another obvious opportunity to undepersonalize teaching is in large lecture courses. For example, administrators at the University of California, Davis, became interested in redesigning their survey biology and chemistry courses because they recognized that they were losing a high percentage of first-year students—the ones who typically take these large lecture courses. It is very easy for a student to become passive in this broadcast-heavy course design. The team involved in the course-redesign projects wanted students to both get more individual attention and take more individual responsibility for their learning. To accomplish these goals, the team employed personalized learning practices as a way of making room for more active learning in the classroom. Students used software-based homework to experience much of the content that had previously been delivered in lectures. Faculty redesigned their lecture periods to become interactive discussions. Meanwhile, the teaching assistants who ran the discussion sections used analytics from the homework software to identify the areas where students were struggling; as a result, they could better focus their class time on those areas. Importantly, teaching assistants received additional training in how to employ active learning principles in their teaching techniques. Once again, in contrast to marketing pitches and popular narratives, the software played only a supporting role, albeit an important one, in undepersonalizing the large lecture.
A less obvious opportunity for personalized learning is in the design of problem-based learning courses. The previous two examples fit within the common understanding of personalized learning being used to help students work through traditional, didactic courses with more support. But Arizona State University incorporated the “flipping” aspect of the technology into an online STEM lab course for non–science majors. In this course students, for their final project, are asked to evaluate the likelihood that there are other intelligent civilizations in a randomly assigned field of stars. The teaching philosophy of the faculty member who was the lead designer of the course is that he should be a coach or a guide, helping students navigate difficult problems. In his view, both content delivery and assessment are activities that take time away from that core function. He and his colleagues framed the course, Habitable Worlds, as a series of challenges. Overcoming each challenge requires the students to learn new knowledge and skills. The course design is based on mastery learning: students must demonstrate that they have learned one skill before moving on to the next. The course is also difficult. Students often get stuck, which is by design. Faculty and teaching assistants, freed up from both content delivery and assignment grading, spend most of their time responding to students’ questions. And because the coursework is all software-based, they can see exactly what students are doing, how far students have progressed, and where students are struggling. Students can proceed at their own pace, moving quickly where they can and getting help where they need it. Yet despite the self-paced nature of the course, there is also a strong social component. Students can and often do seek out each other’s help. Because personalized learning practices make space for more interactivity, these practices often go hand-in-hand with active learning. And active learning is often social.
We suspect there are many opportunities in addition to the ones we have identified here. If faculty are given a commonsense framework and a chance to experiment, refine, and share, they will find novel and exciting ways to better support their students’ individual educational needs.
Doing It Right
Because personalized learning is a family of educational practices that support good course designs, implementing those practices well is not as simple as buying a product. To begin with, course design is always a time-consuming process when done correctly. Second, in many cases faculty will be trying techniques they have not used before, requiring them to teach in ways that are very different from how they have taught before, that are far removed from their experience (and therefore instincts) of what works and what doesn’t, and that may have ripple effects they don’t anticipate. On top of all this, the vast majority of faculty are neither trained in course design/research nor compensated for any time they invest in it. They will need time and support. In many cases, implementing personalized learning well can require an institutional effort analogous to the one required to implement an online learning program well.
Looking across a range of personalized learning projects that have had varying degrees of success, both at the schools we visited and elsewhere, we can identify six steps for a successful strategy. It should be noted that these themes could be applied to any number of pedagogical innovations.
Identify the student need that is to be addressed. The various personalized learning approaches are just one set of tools in the toolbox. Successful programs generally start by identifying a significant educational problem that faculty and program staff believe can be corrected with a change in course design.
Design the pedagogical structure. If the problem that is identified can be addressed through personalization, then how will the course support different students differently? The answer has to be more than just “adaptive learning.” Successful programs identify opportunities in the course design to improve individual support for students.
Pick the products or technologies. The details of different products or technological approaches are most meaningful when they impact what can be done with the course design. Successful programs pick the right tool based on the job at hand rather than on who has the best marketing pitch.
Don’t forget faculty training. Because personalized learning, done properly, generally means implementing new pedagogical approaches, faculty may need to learn to teach in ways that they haven’t taught before. Successful programs provide faculty with training and pedagogical support.
Don’t forget technology support. Software helps with learning only when it works, and Murphy’s Law can hit with a vengeance when technology is mixed with teaching. Successful programs make sure that faculty have the technology training, equipment, and support staff that they need in order to be successful.
Be prepared to measure, fail, and iterate. Because personalized learning approaches often require new software, new teaching techniques for faculty, new responsibilities for students, and in some cases new scheduling challenges, institutions will almost inevitably get some things wrong in the first couple of iterations, and those mistakes may have real impact on outcomes. Successful programs approach implementation empirically but with patience.
On the bright side, the fact that personalized learning is now being attached to funding-related metrics such as degree-completion rates means that attaching institutional support costs to a funding stream will also be easier. In many cases, schools can build personalized learning “muscle mass” by focusing on metric-relevant projects first and then expanding the initiative once the critical institutional knowledge and support mechanisms have been put in place.
Final Thoughts: Ed Tech Groundhog Day
There is a lesson to be learned here, and it is broader than personalized learning. Every popular ed tech trend, going at least as far back as the original online asynchronous distance learning courses in the 1990s, has brought with it a food fight, with proponents hyping the trend as revolutionary and opponents attacking it as harmful. And in every case, a policy or other institutional driver has resulted in a rush of companies responding to the market opportunity created by that driver. Together, these forces generate hype and magical thinking, which in turn provoke an equal and opposite reaction.
In this article, we have tried to identify specific teaching practices being used by educators, and we have tried to describe them in commonsense terms that should make intuitive sense to experienced teachers. These practices always exist at the beginning. Some teacher somewhere comes up with a specific approach to a specific problem. External forces then make that problem a more institutionally consequential one, and companies rush in to name, market, and sell solutions. In the process, we lose track of the original educational idea. It’s like playing a game of telephone in a noisy airport. Except in this case, the message in the game is an actual plan for how we are going to help our students, and when it gets to the end of the telephone line, we will act as if we received the message with perfect fidelity. And then fight over it. Endlessly. Much of the Gartner hype cycle can be attributed to this process.
We can break out of this hype cycle with a fairly simple (though not necessarily easy) approach. Whenever a new ed tech trend gets named—whether it is distance learning, adaptive learning, personalized learning, competency-based education, MOOCs, or something else—we should start trying to understand that trend by looking for the best examples of what teachers and students are doing when they are doing the thing we just named. We should ask them what they are doing, and why. We should ask how their practice is working and what they are learning and what they don’t yet know. We should attach the name of the new trend to those educational practices and those reasons—rather than to any products, technologies, or services. We should not waste time debating whether the name we came up with for those practices is the perfect name or exactly what it includes or excludes. Instead, we should spend our time trying to understand the practices themselves and their applicability to the educational problems we are trying to solve.
Yes, personalized learning is a lousy term, but it is attached to legitimate educational practices that have the potential to improve the lives of many students. It is also a term that is trapped in the early stages of its hype cycle. So let’s just skip to the end and break personalized learning out of the hype cycle by doing our best to understand—and explain—what it really is and why it really matters.
