What to Measure in Video Analytics (and What to Ignore)
Most video dashboards are full of numbers nobody can act on. A short list of the metrics that actually change a decision.
Video analytics packages report a great deal and explain very little. The test for any metric is simple: if it moved 20%, would you do something differently? If not, stop looking at it.
Metrics worth watching
- Play rate per placement. Of the people who saw the unit, what share started it. This is the number your thumbnail and preview move.
- Watch time per visitor. Not per view — per visitor. It resists inflation from autoplay in a way view counts do not.
- Completion by quartile. Where people leave tells you what to fix. A cliff at 25% is an intro problem; a slow bleed is a pacing problem.
- Repeat video sessions. Whether someone who watched once returns. The closest thing to a loyalty signal video has.
- Share of library viewed. How much of the catalogue got any audience at all.
Metrics that mislead
- Total views. Rises with autoplay and with traffic. Says nothing about whether the video worked.
- Average view duration alone. Dominated by asset length; a 30-second clip and a 40-minute talk are not comparable.
- Site-wide video starts. Hides the distribution — and the distribution is the story.
- Impressions. A measure of how many placements you added.
The comparison that matters most
Track share of library viewed over time. If total views rise while that share falls, you are extracting more from fewer assets — the state that precedes a library nobody watches.
Instrument for decisions
Report per placement and per template, never site-wide. Site-wide averages are the reason so many teams believe their video is fine while most of it is invisible.