Monthly Archives: October 2013

LMSs By the Numbers: Hosting and Versions

Like it or not, the Learning Management System has become the primary mechanism for facilitating hybrid and online courses at higher education institutions. In our previous post, we identified that 99+% of US institutions with enrollments of greater than 2000 students use an LMS. Even for traditional face-to-face courses, the LMS plays a key role in delivering course materials, supporting online submission of assignments, storing grades, and hosting discussions. Anecdotally, it is often the second-most-visited online service used by students (with webmail being first, and registration and billing systems receiving more traffic but only during certain time periods each semester).

It is becoming increasingly important for institutions to make smart decisions with regards to LMS implementation. One of the goals of this project is to provide more data points on which to base LMS strategy.  While our last post focused on which LMS or LMSs were in widescale use at each higher education institution, in this post we’d like to dive into more detail relating to two implementation details of the US dataset: (1) which LMS versions are in use and (2) whether the LMS is hosted or not. These two criteria directly impact how successful institutions as a whole and end-users individually can be with their LMS. As examples, decisions such as choosing to be hosted can impact whether an institution has direct access to run analytics or support queries against a production database. Similarly, upgrading to an unstable new version of an LMS can lead to user confidence and operational support challenges.

There are cascading impacts as well. LMS partners and other learning tool developers who plug into LMSs face challenges deciding which LMSs to develop for, and which versions to support. Institutions collaborating in communities of practices that focus on teaching and learning online, even when focusing on a specific type of LMS, lose effectiveness when features and tool sets differ between versions.

To provide additional clarity as to how we compile this data, let’s consider what happens when you load an institution’s LMS into a web browser. As an example, Moodle looks fundamentally different from Sakai. It runs on a different technology platform (PHP versus Java). Its HTML page structure is different. It uses different graphics, icons, CSS and JavaScript libraries. There are other differences, too, and together these form a “fingerprint” that clue us into an institution’s LMS type and version.

Each LMS also runs on a server owned by someone. In the cases of ANGEL, Blackboard, and Desire2Learn, only the vendor can host the LMS. Instructure strongly encourages its clients to host in its Amazon EC2-based environment, but because it is an open source product anyone can host it. Moodle and Sakai each have many hosting vendors from which to choose. The server on which the LMS runs is assigned an IP address issued by the American Registry for Internet Numbers which provides an API to query the owner of each address. It is with this data that we determined whether each LMS environment was hosted by the institution itself, directly by a LMS vendor, or by another third-party.


As you can see, there are some interesting differences.

Instructure’s Canvas has the least variability of all LMSs (it is also the newest entrant to the marketplace). While greater than 95% of institutions that use Canvas host with Instructure, some institutions appear to have chosen to host using their own environment. Even with this small implementation difference, greater than 99% of Canvas environments were running the latest version (as was available at the time of this posting). One institution served as an outlier, choosing to run Canvas from a virtual machine image/appliance provided by Turnkey Linux.

Institutions running Blackboard Learn have the greatest variability of major versions. Sakai has the largest breadth of commercial hosting providers including (formerly) rSmart, LongSight, UNICON, and ETudes. The largest single hosting provider for Moodle is Moodlerooms (now owned by Blackboard). Desire2Learn hosts the largest percentage of environments of any proprietary (ie: non-open-source) LMS. We’ll let you draw other conclusions for yourselves from the graphics provided above.

Data-driven Campus LMS Strategy

Two months ago, we launched MOOC Map, a chronological, interactive map showing the growth of the major xMOOC platforms. It was our first attempt at using data and visualization to better understand the adoption of educational technology. It received mild traction in the press but well over 100,000 visitors during its first month of publication.

Since then we’ve been busy working on our next project. In 1990, The Campus Computing Project began helping higher education institutions to better understand educational technology trends (and not just those relating to LMSs or MOOCs). However, most of the data collected for this project, as the name suggests, comes from a survey. We wanted to see if we could build an automated way to generate a more complete data set around just the LMS usage component of this survey.

Using open data sets from the US Department of Education National Center for Education Statistics, the UK Higher Education Statistics Agency, Statistics Canada, the Association of Universities and Colleges of Canada, and the Australian Open Data site, we created a dataset of all universities in the United States, Canada, the United Kingdom, and Australia including their enrollment numbers, geographical coordinates, and URLs. We then used a combination of Google’s search API, Microsoft’s Bing search API, and a number of open source software libraries to crawl each university’s web site to identify the primary LMS (or in more than 200 cases, multiple primary LMSs) in use. After verification and removal of LMSs that were  not in production use or used only by a single instructor or department, then categorizing those that could not be identified, we aggregated and correlated the data. The data set for institutions with greater than 2000 enrollments in the US is 99% complete (some institutions, predominantly those focusing on the arts or trades, did not appear to utilize an LMS). Similarly, the datasets for Canada, the UK, and Australia are both 100% complete. 

Our approach looked at each institution individually and searched for its LMS(s). Therefore there is a 1:n mapping of institutions to LMSs (ie: each institution is guaranteed to have at least one LMS; some institutions will belong to more than one category below). For consortia sharing a single LMS instance, each member institution was tagged as using that type of LMS. The category of “legacy” seen in the table below includes Blackboard versions prior to 9.0 including Blackboard Academic Suite and the former WebCT line of products. The “other” category includes various LMSs including Learning Studio/eCollege, Jenzabar’s e-Racer, SharePoint, WebStudy, OpenClass, Scholar360/Edvance, other commercial LMSs, other open source LMSs,  and LMSs developed in-house.

Detailed LMS usage data for higher education institutions with > 2000 enrollments (United States)

Most of our effort was spent focusing on collecting data for US institutions. For most of the major LMSs we were able to extract the LMS version, whether the LMS environment was hosted by the institution or vendor, and in some cases even which add-ons and plugins were installed in the LMS (we hope to provide these details in a subsequent post). We were also able to complete an analysis of global LMS usage for institutions in the US, Canada, UK, and Australia.

LMS usage for HE institutions in US, Canada, UK, and Australia

Comparing LMS usage by country, we began to see some interesting differences.

HE LMS usage by geography as percentage of total

We hope that this post generates some interesting conversation at the EDUCAUSE Annual Conference being held this week. In future posts, we hope to provide further interesting insights into educational technology usage and trends.