Update: The video was briefly broken as Instructure inexplicably chose the day after their big announcement to change their YouTube account. I have updated the post with the new URL and it should be working now. Note that anybody else who linked to Instructure’s videos before today is going to have the same problem. The new channel for Instructure’s videos is http://www.youtube.com/canvaslms.
Instructure has just announced that they will be releasing an open source version of their Canvas LMS product. Between this announcement, the winning of the Utah Education Network contract (109,000 college students and 40,000 K12 students), and the oh-so-ever-brief lawsuit by Desire2Learn about that win, Instructure has been making quite a splash lately. In this post, I’m going to look a little more closely at the company, the software, and the business model.
When Instructure alerted me to the fact that they were about to make an announcement about releasing their software as open source (something that I’ll get to in more detail later in this post), I decided to spend a little more time getting to know their product. The company makes it possible for anyone to create a course in Canvas for free, so I created one to play around with. Probably the most astonishing aspect to me is how feature-complete the product appears to be for such a new entrant. I’ve seen a handful of version 1 LMSs in my day, and no matter how promising they were, they have always had obvious gaping holes in their functionality. I found no such holes in Canvas. Granted, playing around with a course site is not like teaching with the product; I could easily be missing problems. (In fact, I’m sure that I am. No software is perfect.) But the point is that I couldn’t find any, and I really looked. That hasn’t happened before.
Let’s take the example of the grade book, which Instructure rightly touts as one of the highlights of their system. Grade books are one of the first places to look if you want to understand the pathologies underlying the design of a particular LMS. They are fiendishly hard to do well. For every ten professors, you have at least 20 different ways to set up a grade book. They use points, or letter grades, or percentages, or a mix; they want to drop the lowest grade; they want to grade on a curve; they want to have some ungraded assignments; they want to use rubrics; and so on. Before you know it, the LMS developers have had to re-implement Excel on the web, and made it hideously hard to use in the process. Plus, electronic grade books are intimately tied into the way students submit assignments and the ways instructors give feedback, so they really are at the heart of academic workflow. They are really important and really hard.
Here’s a video demonstration of the Canvas grade book:
First of all, the design is really clean. There are no unnecessary clicks. If you teach four classes a semester with 25 students per class, small increments of time saved in the grading process add up very quickly. Add to that the fact that the Canvas grade book will render your students’ documents in-browser so you can look at them right in the grading interface, and you have some very significant time savings for faculty.
But it’s not just about speed and convenience. Canvas can take a snapshot of anything on the web and pull it into the grading interface. Want to grade a blog post on the student’s own WordPress installation? You can do it. Want to grade a video posted to YouTube? You can do it. Plus, if your computer has a webcam and built-in mic (and whose doesn’t these days?), you can record video feedback right from the grading interface, giving that increased sense of presence that is so important in successful teaching.
My first impression was that this ease-of-use and innovation was going to come at the cost of feature-richness. But as I spent more time digging into the grade book last night, I was forced to reassess that position. Some of the advanced grading features that I thought weren’t there are, in fact, there. I could think of a single thing that I would want to do with the grade book that I wasn’t able to do. Again, I’m not actually teaching with the software, so I’m sure that I’m missing things. But still, the level of finish is remarkable.
This brings us to the technical aspects. Some of the quality of the application just comes down to good design and engineering talent, but some of it is from the architecture. Instructure CEO Josh Coates makes much of the fact that the system is built on Ruby on Rails and how much more agile that lets the company be. I’m no expert on programming languages, so I have no useful comment to make on that topic. But another differentiator that I’m more confident about is the ability to design the architecture from the ground up using 2011 concepts. The current crop of LMSs in the market are mostly built on a conception of the LMS from the late 1990’s, with a bunch of stuff bolted on as new ideas came a long. But some things are hard to bolt on. Take messaging, for example. Today, we take it for granted that students and teachers are going to use multiple modes for communication—email, Facebook, Twitter, text messaging, Skype, and so on. Since the name of the game in teaching—particularly in a distance learning course—is contact, it just makes sense that you’d want to have a rich set of personalized communication channel preferences across the range of class communications. Instructure has this:
But if you designed your LMS even five or six years ago, this world of multiple channels just wasn’t there when you were thinking about your design. Most of the current-generation LMSs weren’t built with unified messaging systems. They bolted on email alerts one application at a time as it occurred to them that it would be necessary. Retrofitting a unified messaging system to an architecture that doesn’t have one is a major effort.
This difference also shows up in how they integrate third-party applications. Take the example of video recording and video conferencing. It’s not like it never occurred to any other LMS provider that it would be a good idea to integrate such capabilities. But usually it’s done through a plugin architecture that hangs it somewhere non-obvious off the main menu with fifty-seven other undifferentiated third-party capabilities. In contrast, Instructure chose to make videoconferencing and video recording a core part of the platform, giving it a privileged place on the main menu and even integrating it into assignment feedback.
This approach has its tradeoffs, though. For example, Instructure’s videoconferencing platform of choice is the open source DimDim application. Unfortunately, DimDim was acquired by Salesforce.com recently and, while the source is still available, it’s not clear who, if anybody, will continue developing it. But that’s not an insurmountable problem. Instructure could, for example, swap in BigBlueButton or, for that matter Wimba or WebEx. The integration points are there.
The Open Source Model
Speaking of open source, did I mention that Instructure is releasing a version of Canvas under an open source license? Specifically, they are using the Affero GPL (AGPL) license. This is a relatively new variant of the GPL that is intended to close what was perceived to be a loophole for application service providers. The GPL requires that anyone who distributes the licensed code to submit changes they make back under an open source license. To distribute means to physically transmit a copy of the program for somebody else to run it on their own server. But if you’re running a SaaS company, you never distribute the code. Customers never download Salesforce.com or Google Docs and install them on their own machines. Thus, these companies can modify GPLed code and, in the case of Salesforce.com, make money selling the functionality it provides without ever having to contribute that code back to the community. AGPL says source code changes must be made available to all users of the code. Thus, if Salesforce.com modified AGPL code was used in the creation of functionality available to customers, they would have to make their modifications available under an AGPL license to anyone with a Salesforce.com account. What this does for a company like Instructure is it prevents somebody from opening up an LMS SaaS shop across the street, improving Canvas, and not giving those improvements back to Instructure.
It’s important to understand that Instructure’s approach to open source is very different than that of either Sakai or Moodle. Sakai has a very traditional academic consortium model at its heart. The foundation is the center of the action, it’s a non-profit, and it’s largely run by the schools who have a stake in Sakai’s success. Moodle is a company-run open source project, but it’s a highly unusual one that’s very much adapted to fit higher education. In fact, it’s so well adapted that most Moodle customers probably aren’t even aware that Martin Dougiamas’ company, Moodle Pty Ltd, is at the heart of it. Moodle has an extremely strong user community that has input into product development as well as a partner network that also has some input into governance. Instructure, in contrast, appears to be more of a traditional company-run open source project in the style of Alfresco, for example. There are no signs that the company is going to cede product development to an open source community or even cultivate a community in a serious way. On the one hand, I expect there to be some open source contributions to the project. The barriers to entry are low enough, and a Rails-based LMS is a very attractive project for computer science students and computer-savvy university faculty and staff to play with. (On a related note, I’m taking bets on how many hours it takes Chuck Severence to download Canvas and add Basic LTI support.) But I suspect that those contributions will tend to be enhancements and extensions around the edges, rather than changes to the core platform. How much that matters depends a lot on your perspective. Also, Instructure maintains a commercially licensed edition of the product that has functionality not included in the open source edition. (This is a business model that Alfresco actually moved away from.) So, for example, the open source version of the product does not include multitenant support.
Instructure’s approach to open source is indicative of the way the company is run in general. While the design of the software reflects a good understanding of the classroom, the design of the business is all Silicon Valley. That’s not an accident. The company’s new CEO and lead angel investor is Josh Coates. Ever use Mozy? Yeah, that was him. He sold that company to EMC for $76 million. The company before that he sold (or sold the intellectual property, anyway) to Intel. Instructure was born out of class he taught in entrepreneurship at Brigham Young University. Coates asked his students to list the worst software they’d ever used. Apparently, Blackboard made the top of the list. He then asked the students to think about why that software exists, what market forces drive it, and how to make it better. Two students, Brian Whitmer and Devlin Daily, came up with Canvas in response. Coates was impressed with their work and took an interest.
There are strong indications that Josh Coates is taken very seriously by both Silicon Valley and Wall Street. For one thing, I can’t remember any LMS launch being lovingly profiled in TechCrunch. I also can’t remember an LMS startup being mentioned by a journalist on CNBC, but that apparently happened in the last 24 hours. (I haven’t seen it myself.) In talking to Coates, I get the strong sense that he sees Instructure as his legacy. He claims he is in this for the long haul and not interested in selling off this company. (Ironically, he reminds me a bit of Desire2Learn CEO John Baker in that regard.) But don’t expect Instructure to behave like all the other LMS companies. Coates thinks like somebody from Silicon Valley, and he definitely has a bit of a Silicon Valley swagger. He isn’t shy about talking a little trash about his competitors and, for goodness sake, the man has an M18 Hellcat tank destroyer in his garage.
What are the company’s success prospects? It’s hard to say. They certainly are getting off to a roaring start. In the short term, they are in a race against time to get on the evaluation list of schools that are switching off of WebCT or ANGEL. They probably have about 6-12 months to make that happen. Even if they do, they still have to overcome schools’ native conservatism about going with unknown startups. On the other hand, since Canvas was built from the beginning with multitenant support, I strongly suspect that their cost structure for hosting is lower than that of their competition and that they therefore can be pretty aggressive about pricing. That matters more than ever these days. Longer term is harder to analyze. I expect the entire educational technology landscape to change considerably by 2014. It’s very difficult to know what the competitive landscape will look like by then.
But I’ll say this much. This is a company that should be taken seriously.
I can’t tell you how simple but critical this observation is, particularly after hearing substantial frustration expressed by new Sakai instructors about how difficult it is to grade student forum contributions in a course of 60 (bad enough that one instructor was contemplating giving up on the forum approach). I would go so far as to insist that any user testing on more advanced features in OAE (grading in general, but also on things like forum contributions) must be designed to account for just this: more clicks of the mouse or switching of pages = design fail.
And yeah, that rubric-based grading interface does look pretty slick. The paper grading is similar to what you see in TurnItIn’s GradeMark service, but nice to see it integrated into the LMS.
Steve Terry says
Very nice article about Canvas – really helpful observations about the product/company. Again, thanks!
Luke Fernandez says
A licensing question:
As per your post: “Instructure maintains a commercially licensed edition of the product that has functionality not included in the open source edition. (This is a business model that Alfresco actually moved away from.) So, for example, the open source version of the product does not include multitenant support.”
“AGPL says source code changes must be made available to all users of the code. Thus, if Salesforce.com modified AGPL code was used in the creation of functionality available to customers, they would have to make their modifications available under an AGPL license to anyone with a Salesforce.com account. What this does for a company like Instructure is it prevents somebody from opening up an LMS SaaS shop across the street, improving Canvas, and not giving those improvements back to Instructure.”
So does this mean that the AGPL license applies to the open-source Instructure source code but not to the code base that Instructure is using? More largely, how does this particular licensing model shape the kind of open source ecologies that could develop around the initiative? Your post mentions that in terms of development we could expect to see code contributions around the edges. But what about the development of alternative LMS SaaS shops? Naturally, Instructure can’t be faulted for creating a license that maintains some competitive advantage for them. After all they aren’t an eleemosynary organization. But do the dual licenses that have been created provide enough economic incentive for other SaaS shops to host the Canvas source code?
Ted Major says
Demo video “is no longer available because the uploader has closed their Youtube account.”
Is video available elsewhere? I’m growing increasingly frustrated with Bb9’s interface, so I’d like to see how Canvas works.
Michael Feldstein says
Ted, I have emailed Instructure and will update the video link as soon as I have a working one. Stay tuned.
Luke, the AGPL absolutely does impact the ecology around Canvas. Right now, Instructure is reserving multitenant hosting, and presumably other potent, enterprise-y features, for the commercially licensed version. So a company that wants to compete with them using their code would have to catch up right out of the box by redeveloping those enhancements. And any enhancements that the competitor developed would have to be released under an AGPL license.
Sakai effectively imposes no limitations on commercial support vendors. Moodle imposes significant limitations, enforced through their ownership of the Moodle trademark, but they still are actively attempting to cultivate a network of support vendors. Instructure’s strategy all but ensures that they are the only game in town for Canvas.
“Instructure inexplicably chose the day after their big announcement to change their YouTube account.”
Very similar to my experience with trying out the free site to use for my class. Something is always broken somewhere, someone is always fixing it upon request.
I am curious to see how going Open source will affect the day-to-day operations of the LMS.
Michael Feldstein says
Video is fixed now.
Jane, thanks for your comments. I’m interested to hear from others as well who have actual teaching experience with the platform. Details–good or bad–would be appreciated.
Doesn’t this statement need to be qualified a fair bit?
What it seems this strategy does do is establish a fairly high barrier to entry for commercial competitors who want to distinguish themselves on proprietary features. But I don’t see how it (severely) discourages a business model that focuses, let’s say, on support. For sake of argument, a small group of developers could fork the code, add those missing features, and try to build a community around that.
I go back and forth on where I stand on the licensing question (GPL-like or MIT/BSD-like), but tend in the end to believe that the benchmark for whether a particular strategy works is whether it encourages a vibrant development ecosystem around it, and discourages a lot of proprietary forking. Licensing alone clearly is only one issue that impacts this, but there are others (the architecture of the code, the openness of the community, etc.).
I have been teaching with Canvas since Jan, and have 35 university students. The students absolutely love it. The nice thing about Canvas over Bb 9.1 (which I taught with last summer) is it is less click-ish, and it is so much easier to grade with. I can say that my grading time has been drastically reduced by the grading efficiencies. There is not a big learning curve. Once you understand a couple of things about the metaphor of Canvas, you are off and running. I am sold!