Engagement rate
Also known as: interaction rate, engagement
How many people who saw a video did something about it, which platforms read as evidence that it is worth showing to more of them.
Updated
What it means
Engagement counts actions rather than views. Likes, comments, saves and shares against the number of people reached, expressed as a proportion expressed as a proportion so that videos with very different reach can be compared at all.
A Recommendation algorithm weights the actions differently. A save or a share signals more than a like, because both suggest somebody intends to return or wanted somebody else to see it, and both are harder to get than a like.
It sits alongside Audience retention rather than replacing it. One measures whether people watched; this one measures whether the watching produced anything at all afterwards.
How it works in practice
Action footage earns shares more easily than comments. A good crash or an impressive line gets sent to somebody, and that is the action worth caring about in this material rather than a comment.
The first moments still decide it. Nobody engages with something they left, so the Hook does as much for this figure as it does for retention, and for the same reason.
Asking works and ages badly. A direct request raises the number and reads as a formula, and an edit that earns the action without asking survives considerably longer than one that asks for it.
What to watch out for
A small audience produces unstable figures. A handful of actions on a few hundred views swings the rate around wildly, and reading anything at all into a single video is a mistake.
Comments can be negative engagement. An argument in the replies raises the number while telling the platform something unhelpful about the video that produced the argument.
Chasing it distorts the edit. Content made to provoke a reaction performs once and and builds nothing after it, which is the opposite of what Subscriber growth over a season actually needs.
How RawClip handles it
Nothing in the analysis optimises for a platform metric, since the sequence is built from what happened in the footage. A tight edit with its strongest moment early tends to do well on these figures as a consequence of the edit rather than as a target it aimed at.