Retention curve

A retention curve plots the proportion of viewers still watching against position in the video. It starts at one hundred per cent and falls; the shape of the fall is the useful part.

Three patterns recur. A steep drop in the first ten to fifteen seconds means the opening failed to establish why the video is worth watching, and it is the most common and most fixable problem in work video. A steady, gentle decline is normal and healthy. A sudden cliff in the middle points at a specific moment — a tangent, a long silence, a section that answered a question nobody asked.

Occasionally the curve rises. A bump means viewers went back and rewatched something, which usually indicates a passage that was either especially valuable or badly explained. Both are worth investigating, and the transcript will usually tell you which.

The curve is the most directly actionable thing in video analytics because it converts into an edit. Where the cliff is, cut. Where the rewatch is, slow down or add a caption. A team that reads the curve after each recording improves faster than one that reads only the totals.

Sample size deserves a mention, because retention curves look authoritative at any volume. A curve drawn from six viewers is noise with a shape, and acting on it will send you rewriting an opening that was fine. Wait for a few dozen viewings before treating a dip as a finding, and be more patient still with videos sent to small named audiences.