Skip to main content

Retention report

The Retention report answers the question that decides every VSL: where do viewers leave? It plots an aggregated, second-by-second retention curve across the videos in scope and pinpoints the exact moments people drop off.

Where to find it​

Open the Analytics workspace and select Retention in the sub-nav. The report inherits the toolbar's scope, date range, and compare mode.

KPI cards​

Four cards summarize retention health:

  • Avg retention — Average completion across all videos in scope.
  • Drop-off zones — Number of major cliffs detected.
  • Rewatch hotspots — Moments viewers scrubbed back to.
  • Pitch retention — Share of viewers still watching when the offer reveals. This is the signal that most closely tracks revenue.

The aggregated retention curve​

The curve is normalized to 100% playback length and weighted by views, so longer and more-watched videos count proportionally. Three view modes:

  • Curve — A single blended retention line.
  • Overlay — Compare segments on the same axes. Toggle Mobile, Desktop, Paid, Organic, and Converted overlays.
  • By cohort — Compare retention across acquisition cohorts at a glance. For a full cohort matrix, open the Cohorts report.

A dashed Industry line marks the network median so you can see at a glance whether you sit above or below it. Cliff markers label the Hook, the biggest mid-video Cliff (with its drop %), and the Pitch moment.

Engagement heatmap​

Below the curve, the Engagement heatmap shades each moment by interaction: bright = viewers replayed this moment, dark = they skipped it. Hover for second-level data. Filter the heatmap by All, Converters, or Bouncers to see how your best and worst viewers behave differently. A typical insight: viewers scrub back just before a cliff because the prior beat confused them — tighten that transition.

Major drop-off cliffs​

The drop-off table ranks the steepest cliffs with:

  • Time the cliff occurs and which video it belongs to.
  • Drop % and viewers lost, color-coded critical / moderate / low.
  • An AI cause (e.g. "Founder backstory · too long", "Slow B-roll · low-energy VO").

The visible Fix → and Ask AI why controls are presentational in the current implementation and do not open another workflow.

Best practices​

  • Watch the pitch line, not just average retention. A video can hold attention early and still bleed viewers right before the offer.
  • Compare segments. Mobile usually drops faster than desktop; use the overlay to confirm before you re-edit.
  • Turn cliffs into tests. When you find a critical drop, spin up a hook or pitch-timing A/B test on that video.

Frequently asked questions​

Why is the curve normalized to 100% length?

So videos of different durations can be aggregated and compared on the same axis. Each point is weighted by views, so popular videos influence the blended curve more.

What is pitch retention?

The share of viewers still watching when the offer reveals. It tracks revenue more closely than overall retention — see the Metrics glossary.

How does the AI know why viewers left?

Atomicat AI reads the drop pattern against the video content and labels a likely cause. The visible Ask AI why control is presentational; open Ask AI directly to investigate.