Are Shorter Videos Really Better? Harvesting YouTube Analytics Data for an Apples-to-Apples Comparison

Project Description Background: Lots of studies have concluded that viewer retention decreases as video length increases, but most of the time people are comparing apples and oranges in terms of content. I’m fortunate to be in the unique position to have two separate MOOCs that teach essentially the same content using two different platforms (the LEGO Mindstorms NXT and EV3 robots) but whose video lengths are quite different. So I can pair up one long NXT video with a set of 3 shorter EV3 videos that cover the same content and do an apples to apples comparison of long and short videos. All of the videos are hosted on YouTube, and I’ve already done an analysis based on aggregate analytics data that I can easily get out of YouTube (see https://doi.org/10.1145/3330430.3333617 for more info).

In order to take the next step and get what I expect will be much more meaningful results, I need to have more disaggregated data. The goal of this project is to attempt to find a way to use the YouTube Analytics API to extract data that would facilitate a more detailed comparison of the long and short videos. Ideally, if it were possible to extract data on minutes watched by each individual user, I could do a t-test to see if there's a significant difference between the %watched of video X in the NXT MOOC vs. %watched of (videoX.1 + Video X.2 + Video X.3) in the EV3 one. But it's not entirely clear that YouTube keeps track of data in that way, and even if they do, whether they allow users to extract the data. So I need help to get a better idea of what data is available, how finely it could be extracted, and whether there are any clever approaches that could be used to get the data even if YouTube doesn't explicitly provide it. (e.g., if viewers never overlapped (unlikely, but ignore that for now) then it might be possible to figure out watch time by manipulating the time-periods that we are querying about.