...Or how to spam Facebook with requests otherwise avoidable if they provided proper data through the Graph API.
Why is it useful? Doesn't Facebook already do this? Aren't there a lot of platforms that provide this service?
Actually, no. Live Videos are still a new feature and the media has yet to explore their full potential. Except the usual celebrity and freelance reporter, Live Videos are not yet used that widely. However, this is bound to change, as it can be seen from services like Snapchat and Whatsapp, which prove the users' need for easily accessible video.
Being this new, developers haven't really looked into acommodating the evolving trends and this is why, right now, Live Insights is the best tool out there for providing feedback for your Live Videos (well, at least after it gets reviewed and approved by Facebook). Facebook shows a graph based on view counts, but that is pretty much the whole feedback you have on your video.
Isn't a live video like an ordinary video? Or like a post / photo?
No. The biggest difference between a normal video and a live one stands in how people react to them. On a normal video you get comments and reactions, but they usually refer to the whole video. People know how long it is, and usually comment / react after they watch it entirely, in order to have a full picture about the subject. Thus, the totality of comments and reactions represents feedback for the entire video as a single entity, taken at once.
However, on a live video, where people don't know when it ends, the whole idea is to react / comment exactly when something relevant happens. If you show a kitten or something "cute", you can expect people to react with "love" and post comments like "aww how cute". But in the same stream you can also show something gruesome, and people will react / comment accordingly, showing disgust and disapproval. Thus, in a live video, it is of utmost importance the exact time when comments and reactions are created, because they only make sense in that short timeframe. Indeed, the total view count is the only metric that reflects the quality of the entire video, while the others absolutely need the context in order to be correctly interpreted.
This is exactly where Live Insights comes to play. After a user starts streaming, they click a button on the dashboard and the server begins making requests to Facebook's Graph API for updates on the latest live video of that user. While the video is live, the API is polled for new reactions, storing the results in the database. This is needed because Facebook's API does not provide a way of retrieving the timestamp of the reactions.
After the video ends, Live Insights uses the reactions to plot several public engagement charts over the time of the video, providing a custom web video player that allows the chart cursors to be synchronised to the video cursor. The platform also takes the comments and analyses their sentiment value, retrieving scores for approval / rejection and sentiments (positive, negative, neutral), while also providing visualisation for the general trend of the sentiments i.e. how much does the approval vary throughout the duration of the video.
Hopefully, if it gets reviewed by Facebook, the platform will be deployed in beta under this domain.