Hey @ncarrascosa ,

The Amplitude Data Taxonomy Playbook could be a good place to start if you haven’t already, where they have a sample taxonomy for the Media domain. 

I would also love to hear from other Community members who work in the Media space about their experience of instrumenting video interaction events :)

Like in most cases, the answer depends on what you want to do with your data. For our purposes, simple LoadVideo, PlayVideo, and related error events have been enough, but if you need to understand the usage in more detail, you may want more.

One thing we have found useful is building summary events. So during whatever on-going action you want to track (like watching a video), you store any actions you want to track as independent variables, and when the activity end, you generate one event that sums up what happened.

So using playing a video as an example, the summary event could look something like this:

• EventName: VideoPlayed
• EventProperty: VideoID
• EventProperty: EntryPoint (PromotedVideo/AnotherVideo/DirectLink/etc)
• EventProperty: ChannelFollowed (Yes/No)
• EventProperty: VideoDuration
• EventProperty: PercentageWatched
• EventProperty: ContentTags
• EventProperty: NumberOfPauses
• EventProperty: FollowUpAction (Replay/NextVideoPromoted/NextVideoManuallyChosen/Exit/etc)

That’s a random set of properties, but hopefully it conveys the idea. This approach allows you to easily analyse things like how does EntryPoint affect PercentageWatched or FollowUpAction.