Video Analytics Dashboard
An internal tool for gaining insight about how learners watch educational videos.
A key insight from the project was that the longer our videos were, average view percentage tended to trend downward. At lengths of 10 minutes or above, viewing percentages are almost always below average.
Project summary
This project attempts to address questions that our team found ourselves continually asking about our video analytics that we couldn't answer with YouTube's built in analytics tools. The main problem what that the built in analytics did not allow you to filter queries based on playlists, which was the main organizing structure for our channel.
Along with a student fellow on our team, I built a dashboard that would be playlist friendly using Google Data Studio, with data fetched via the YouTube Data API via Google App Script, along with the built in connections to YouTube data provided by Data Studio.
Role
I led this project and worked on every aspect of it from researching the questions our team wanted to gain insights on, writing the app script to query the YouTube API, to designing the interactive elements of the visualization. I also managed and mentored a student fellow who was assisting with the project.
Public Site
Querying the YouTube API
This project started from the problem of not being able to get the data we wanted from YouTube's built in analytics tools. While frustrating, it was also a great opportunity to learn to work with the YouTube API, which is something I had always wanted to learn!
While we could have tackled this in a variety of ways, we knew it would be important to quickly get data in front of the team in a format that anyone on the team could easily interact with. For us, that meant that the simplest way to get started was to use Google Sheets, App Script, and Data Studio.
The trickiest part of the project was learning the structure of the API so that our App Script would give us data in our Google Sheet in a format that Data Studio would understand.
I wrote three different scripts querying two different YouTube APIs. The main goal of the query was to get a list of all of the relevant playlists, associated videos, and video IDs and merge this data with other queries already built in to Data Studio.
After A LOT of trial and error and YouTube tutorials...we finally got the query to feed the necessary data into our sheet!
Now we could merge the data from our spreadsheet with the YouTube data already built-in to Data Studio.
Creating a dashboard that can filter by playlist
After getting all of our data into DataStudio, we now had the ability to group video analytics in terms of Playlists. We designed three pages in our dashboard that would answer three key questions our team had about how students were interactive with our videos.
Each page would allow a team member to filter both by date and by playlist, which was synonymous with a course.
In this view, the user is drilling down on the Decision Models course in order to look at the performance of specific videos within the course during a specific time period.
Student viewing data over time helped our team understand when students tended to rewatch review videos from earlier in the course. Seeing these patterns helped us inform faculty about potential guidance they could to the class about key topics to review, or content they should be reviewing but weren't.
In general, longer videos tended to have lower average view percentages, but not for all courses! Filtering by playlist helped us to understand which courses tended to be exceptions to the rule, and look more deeply into why those longer videos performed better than others.