Our Clinisnips channel teaches basic clinical procedures. It has seen sustained engagement since 2008. We show various approaches to data analytics, using both YouTube's analytics and activity metrics (xAPI and the LRS).
Introduction
In 2007, the eLearning group at NOSM started working on a series of procedures to illustrate how to perform basic clinical procedures. With support from the Inukshuk Fund and an unattached grant from Merck Canada, we developed these materials which were aimed to be concise with small file sizes that could be stored on a PDA.
Content was determined using a peer-review process for script and learning design. With the help of Laurentian University AV media team, and a professional narrator, we were able to craft a series which had high production values.
We first posted this onto our own web site, http://pocketsnips.org, but this had very little exposure or meaningful use. We chose to publish our videos on YouTube, which was a controversial step at the time since it was seen as a frivolous purveyor of kitten videos. While kitten videos still prevail, YouTube is now the dominant video publishing platform.
Background
With the original funding, we created a series of 20 short videos on a variety of topics, such as: nasogastric tube insertion, intravenous line insertion, shave and punch biopsies, suturing, Foley catheter insertion, various casting techniques, central line placement.
We chose to include only the dynamic or kinetic elements of the procedure in the video, reasoning that it was more efficient to insert materials like the equipment list, follow-up instructions, into ancillary text (that we called the ‘breadbasket’). This also made it easier to alter such materials without having to reshoot or edit the video itself. This also allows other organizations to apply their own contextual materials to the videos.
We added some metadata to the YouTube uploads but included links back to our original ‘breadbasket’ for those who were interested in learning more.
Methods
Right from the outset, Google provided detailed analytics about video usage. But they kept changing how data was reported. First we had Google Insight then YouTube Analytics and now we have YouTube Studio. There was a major break in data continuity in 2012 making it difficult to span these two periods.
Google provided methods that were powerful but opaque, e.g. what does “retention” mean? This was only relatively meaningful compared to other videos but not known how calculated.
In this example of the Nasogastric tube, we see it has sustained interest. Most viewers watch until end of video and the video still receives >100 views/hour, after 12 years. By comparison, the average lifespan of most YouTube videos is six weeks. We can also examine which portions of video are reviewed, which is helpful for fine-tuning the material. We were able to use the YouTube API to extract finer detail but this was difficult.
Initial Results from YouTube
When we submitted the abstract in August 2018, there were 6.0M minutes = 100k hrs, hence the title.
Now we have 7 million minutes, and that only goes back to June 2012.
Earlier results (the greyed out area on this chart) from Google Insight showed same rate of views so we can extrapolate that to now > 10M views, >10M minutes, or 19 years of attentive viewing.
We also just passed the 10k subscriber mark, which could affect revenue – more on this later.
Curation, not Creation
Although the approach we took with PocketSnips reduced the costs of video creation by a factor of 10, it was still expensive at $3000 per video.
Nowadays, there is no shortage of content on YouTube but it is rare to find a video that portrays exactly what you want. Instead, we turned to curation: extracting the creme de la creme, and embedding into additional contextual material – an extension of the original ’breadbasket’ concept. There are several video embedding tools out there but none with the features we needed so we created our own: CURIOS video mashup tool (http://curios.openlabyrinth.ca), which provides precisely timed start and end points, annotation and sound layer overlays.
We also incorporated H5P widgets (http://h5p.org) into our materials, which gave us an extended range of tools (including video embedding and annotation). Both CURIOS and H5P can be embedded into a wide range of learning platforms such as Moodle, WordPress, OpenLabyrinth, Desire2Learn and OLab4 (http://olab.ca).
Activity Metrics
With YouTube analytics, you are bound to a single source for data and cannot track usage across embedded materials.
We turned to the use of Experience API (xAPI) enabled tools as a way to tie all these together. Data from multiple sources can be sent to a Learning Record Store (LRS), using very simple programming techniques and add-on modules. We embedded xAPI into OpenLabyrinth, OLab4, WordPress and our H5P widgets, and even into a cheap Arduino nanocomputer (cost $30) – a preview of things to come from the Internet of Things (IoT).
xAPI produces very simple statements in the format of Actor Verb Object or “Bob did this”. This simplicity belies significant power and captures large amounts of detailed data on exactly how your learners interact with the materials you have curated and created.
Results
While 100,000 hours of viewer attention sounds impressive and demonstrates extensive reach for these materials, how does this compare with traditional media?
Using the methods suggested by Hovden (2013), we calculated the h-index of the Clinisnips channel to be only 5. However, this calculation is prone to some broad assumptions which may be more tuned to such popular video feeds as Justin Bieber. We hope to be able to use Altmetrics (http://www.altmetric.com) to provide a broader perspective on the impact of such materials but this is currently limited by odd factors such as departmental affiliation.
In our original article about our YouTube efforts, we noted that we had much data about our materials but little data about our users. The data available in the LRS provides this missing context, across a range of blended learning designs and platforms. LRS data can be federated thus broadening the scope of analytic tools.
Discussion
We are entering the world of Precision Education, using big data to drive personalized learning designs.
Main learning points:
Good quality material makes a good base
Embedding & increased discoverability increases impact
Activity metrics provide feedback to improve materials, learning designs and learners.
These slides are available on SlideShare.