How to read a retention curve
Read a retention curve in three zones: the first 30 seconds, where a drop below roughly 70% means the packaging and hook disagree; the body, where the slope should be shallow and steady; and every visible cliff, each of which marks one editing decision you can name and remove.
- Three zones: the 30-second hold, the body slope, and the cliffs.
- 70% at 30 seconds is a healthy long-form benchmark; below 60% indicates a packaging-to-hook mismatch.
- Every cliff is one nameable editing decision, not a general quality problem.
- Read the curve after 7 days, and compare it to your own channel's typical shape, never to another channel's.
Average view duration is a single number and tells you almost nothing about what to change. The curve tells you where. Because the shape is produced by specific moments in the edit, reading it is a diagnostic exercise rather than a performance review: each feature of the line maps to a timestamp you can open and watch.
Published July 29, 2026 · Updated July 29, 2026
How to run a retention review
- Wait 7 days after publishing so browse traffic is represented.
- Note the percentage still watching at 30 seconds and compare it to your channel's usual figure.
- Mark every cliff and step on the curve with its timestamp.
- Open each timestamp and watch the ten seconds before it; name the decision that caused the drop.
- Fix the single largest drop on the next upload, and change nothing else.
- Compare the new curve against the old one at the same timestamps.
Zone one: the first 30 seconds
This zone measures agreement between the package and the opening. Viewers arrived with a specific expectation; the opening either confirms it immediately or spends their patience explaining itself. Intros, channel branding and throat-clearing all cost measurable percentage points here.
If the 30-second hold is weak while click-through is strong, the fix belongs to the edit, not the title. Softening the title to match a slow opening lowers both numbers.
Zone two: the body slope
A healthy body declines gently and evenly. Visible steps mean the video is a sequence of segments rather than a continuous argument, and viewers are treating each boundary as an exit.
The remedy is structural: open loops that cross the boundaries, so that leaving at a segment end still means leaving mid-question.
- Gentle even decline: the structure is holding.
- Repeated steps: segment boundaries acting as exits.
- Single cliff: one specific moment to open and rewatch.
- Small rise: viewers rewinding, usually because a point was compressed.
What the curve cannot tell you
Retention describes the people who clicked. It says nothing about the far larger group who did not, and it cannot distinguish a boring segment from a genuinely difficult one that a committed audience needed. Use it to locate moments, then judge those moments by watching them.
Worked examples
Finding the cost of an intro animation
A channel's curves all showed the same 9-point drop between 0:06 and 0:12 — the length of its animated logo sting. Removing the sting lifted the 30-second hold from 61% to 72% across the next four uploads, with no other change to structure or topic.
Turning curves into edit rules
After five reviews the same causes repeat, and they become channel-specific editing rules: no sting, no recap before the payoff, no topic switch without a bridge sentence. Rules derived from your own curves outperform general advice because they were measured on your audience.
Pair the review with outlier scoring: a video that over-performed on views but retained poorly is a packaging debt, and the next upload should repay it before pushing the promise further.
Frequently asked questions
Related reading inside Outlieo
Why weak early retention usually starts before the click.
Structures that keep the body slope shallow.
What has to happen in the first thirty seconds.
Ship a week of content, not a to-do list.
Join the Outlieo waitlist. Track the outliers, script in your voice, and repurpose to LinkedIn.