Visiroll
Log inStart free

VIDEO ANALYTICS GUIDE

Read the drop-off.
Keep the denominator.

A player load, a qualified view and a completed video answer different questions. Start with the decision you need to make, then choose the metric.

By VisirollReviewed 30 August 20268 minute read

One playback funnel, five questions

Impressions
How often did the player load? This measures opportunity, not attention.
Plays
How often did playback start? A play can still be accidental or immediately abandoned.
Qualified views
How many sessions accumulated at least three seconds of watched time? Visiroll uses this as the first meaningful-view threshold.
Engaged views
How many sessions accumulated ten seconds or reached 25% of the video? This adapts to short clips without pretending every play is engagement.
Completions
How many sessions reached the end event? Completion is useful only beside the video length and number of qualified views.

The exact implementation and retention windows are documented in Visiroll analytics definitions.

Calculate rates with an explicit denominator

Counts describe volume. Rates make two videos easier to compare, but only when the denominator is visible and the traffic sources are comparable.

Qualified-view rate = qualified views ÷ impressions
Engagement rate = engaged views ÷ qualified views
Completion rate = completions ÷ qualified views
Average watched time = total watch time ÷ qualified views

Do not mix plays into one report and qualified views into another. Segment by website, video, date range and campaign before comparing. A homepage visitor, a paid-campaign visitor and a support-page visitor arrive with different intent.

Read milestone retention as a shape

Visiroll records Started, 25%, 50%, 75%, 90% and Completed milestones. It is a milestone view of the audience, not a second-by-second heatmap.

Large loss before 25%

The opening may not match the promise around the player, the first seconds may be slow, or the visitor may not have intended to watch.

Stable middle, weak completion

The core content may work while the ending runs long. Move the result or call to action earlier and test a tighter ending.

Good completion, low play rate

The people who start are satisfied, but placement, poster, title or surrounding copy may not earn the click.

Strong watch time, weak conversion

The video may explain well but leave the next action unclear. Playback analytics cannot prove the downstream CTA worked.

A repeatable review for one video

  1. Write down the job.Is the video meant to introduce, explain, reassure, or convert? One video should not inherit every objective.
  2. Choose one comparison.Compare the same video across two sites, or two edits in similar placements. Avoid changing traffic, placement and creative at once.
  3. Inspect the funnel.Start with impressions → plays → qualified → engaged → completed. The largest useful break points to the next hypothesis.
  4. Change one thing.Test the opening, duration, poster, placement or CTA. Keep the previous period and denominator in the report.

What retention cannot tell you

  • It does not identify why one individual stopped watching.
  • It does not prove causation between a video and a sale or form submission.
  • It is not an industry benchmark: length, placement, autoplay and traffic intent change the expected curve.
  • Daily pseudonymous unique viewers are not persistent person identities.

Use the smallest claim the data supports. “More qualified views after changing the poster” is defensible; “the poster caused more sales” needs conversion instrumentation and a controlled comparison.

Method reviewed against the Visiroll player and analytics aggregation on 30 August 2026.