Why video "benchmarks" mislead — and what to measure instead
Published video benchmarks compare your video to an unknown mix of content lengths, audiences and traffic sources, so they cannot tell you whether your video is good. Build an internal baseline instead: play rate, retention at 50%, completion and CTA click-through per video, compared against your own median and against the previous cut of the same video.
The problem with the published number
You have seen the format: "the average marketing video retains X% of viewers". Ask three questions of any such figure and it usually collapses.
- Which videos? A ten-second loop and a forty-minute webinar in one average is not a benchmark, it is a coin flip.
- Which traffic? Retention on an autoplaying hero and on a video someone deliberately clicked from a newsletter are different phenomena.
- Which definition? A "view" can be three seconds, or an impression, or a de-duplicated play. Vendors do not agree, and the difference can be several fold.
Even when the number is honest, it is an average of strangers. It cannot tell you whether to re-cut your demo.
The baseline worth having
For each video, four numbers, tracked over time:
| Number | Question it answers |
|---|---|
| Play rate | Is the thumbnail and placement earning a start? |
| Retention at 50% | Does the middle hold? |
| Completion | Is the length honest? |
| CTA click-through | Did the watch turn into intent? |
Then two comparisons, and only these two:
- This video against your own median for the same kind of video.
- This cut against the previous cut of the same video, on comparable traffic.
Why the second comparison is the valuable one
It controls for everything a benchmark cannot: your audience, your traffic source, your product's complexity. Cut the intro, publish, wait for a similar number of plays, compare the curve. That is a real experiment, and it takes a week.
Sample size, briefly
Do not re-cut a video on thirty plays. Wait until the two curves you are comparing are built from a similar and non-trivial number of plays — a few hundred is enough to see a cliff move; a dozen is noise. And compare like traffic with like: a spike from one newsletter is a different audience.
The number nobody publishes
The one that decides budgets: leads per hundred plays. It combines everything above and is specific to you. With per-video plays, retention and lead attribution it takes one glance, and it is the only "benchmark" that should change what you do next.
Frequently asked questions
Is there a good average completion rate for video?
Not one that transfers. Completion depends mostly on length and traffic source; compare a video against your own others and against its previous cut.