Private beta30 seats for the $9 test film.Check a niche

How to Read a YouTube Audience Retention Graph: Dips, Spikes and What to Fix

The retention graph is the closest thing YouTube gives you to watching over a viewer's shoulder. This guide shows how to read each shape on it and turn it into concrete changes for your next video.

Blog, 3 November 2026, 8 min read

Where to find the report

In YouTube Studio, open Content, select a video, then open Analytics and look at the Overview or Engagement tab. The audience retention chart shows how much of the audience was still watching at each moment of the video.

According to YouTube's help pages, a few things are worth knowing before you start:

  • The report is only available per video, not for the whole channel.
  • Retention data typically takes one to two days to process.
  • Key moments are only highlighted for videos that are at least 60 seconds long and have at least 100 views.
  • You can compare a video against typical retention, which YouTube calculates from your 10 most recent videos of similar length. Click "See more" to open that comparison.
  • A Segments filter lets you split the data, for example new viewers versus returning viewers.

That last point is underrated. Returning viewers know your style and forgive a slow start. New viewers do not. If a video is meant to grow the channel, the new-viewer line is the one to study.

The four key moments

YouTube highlights up to four types of moments on the graph. Its help page describes them like this.

Intro. The percentage of viewers still watching after the first 30 seconds. YouTube says a high figure suggests the opening matched the title and thumbnail and kept people interested.

Top moments. Stretches where almost no one dropped off. If they fall late in the video, YouTube suggests moving compelling content earlier, because audiences typically shrink as a video goes on.

Spikes. Parts that more viewers watched, rewatched or shared. A spike can mean the moment was great, or that it was confusing and people replayed it to understand.

Dips. Parts where viewers skipped ahead or stopped watching. These show where interest was lost.

Reading the shape of the line

The help page also describes what the basic shapes mean.

ShapeWhat YouTube says it meansTypical cause in a faceless video
Flat lineViewers watch that section from start to finishA strong story beat, a clear payoff, good pacing
Gradual declineViewers gradually lose interest; some tapering is normalLong sections without a new question or visual change
SpikeMore viewers watching, rewatching or sharingA reveal, a striking visual, or something hard to follow
DipViewers leaving or skipping that partA slow tangent, a repeated point, a sponsor read, a weak transition

A small detail: the detailed report can show segment view counts above the video's total views, because one viewer can rewatch a section within a single view.

Diagnosing the intro

The first 30 seconds deserve their own analysis because they set up everything else.

If the intro figure is weak compared with your typical retention, check in this order:

  1. 1.Does the opening confirm the title and thumbnail? If the thumbnail shows a sinking ship and the video opens with the history of shipbuilding, viewers feel misled.
  2. 2.How soon does the first interesting thing happen? Count the seconds until the first fact, image or question a viewer would care about.
  3. 3.Is there anything to cut? Greetings, logo animations, "before we start" asides and subscription requests all push the substance back.
  4. 4.Is the voice and audio right from the first second? Low volume, music that drowns the narration or a flat read are noticed immediately.

YouTube's help page suggests experimenting with the opening when the intro figure is low. In practice that means writing several hook variants for every script and treating the hook as the most edited paragraph.

Diagnosing dips

A dip is the most actionable thing on the graph, because it points to a specific timestamp.

How to work through one:

  1. 1.Open the video at the dip and watch the 20 seconds before it.
  2. 2.Find the matching paragraph in the script. Most dips line up with a specific sentence or transition.
  3. 3.Name the cause in one phrase. Common ones:
    • The section answered the main question too early, so viewers got what they came for.
    • A transition with no re-hook ("Now let's talk about...").
    • A list of dates, names or numbers read aloud.
    • The same visual held on screen for too long.
    • A tangent that does not move the story forward.
    • A sponsor segment or call to action.
  4. 4.Write the fix as a rule for the next script, not just for this video.

Not every dip is a problem. A drop right after you deliver the answer a searcher came for is natural. The question is whether you can open a new reason to stay just before that moment.

Diagnosing spikes

Spikes need interpretation because they can be good or bad.

  • Good spike: viewers rewatched a reveal, a striking map animation, a surprising number. Make more moments like that, and consider moving one earlier.
  • Confusion spike: viewers rewound because something was unclear. Look for fast narration over a complex graphic, a name that was not introduced, or on-screen text that disappeared too quickly.

If the spike sits on a graphic, pause on it and ask whether a first-time viewer could read it in the time it was on screen.

Diagnosing a gradual slide

A steady downward slope with no clear dips usually means nothing is wrong in particular, but nothing is pulling viewers forward either. Some tapering is normal, as YouTube's help page notes. But if your slope is steeper than your typical retention, the usual fixes are structural:

  • Open a smaller question at the start of each section and close it within the section.
  • Add re-hooks at section breaks.
  • Change visuals more often.
  • Cut 10 to 20 percent of the script and see whether the slope flattens on the next video.

Diagnosing the ending

Watch for a steep drop in the last stretch. Some of it is natural, but scripts often invite it with phrases like "that's all for today". End on a strong final line and move directly into a reason to watch the next video, then keep the end screen short.

Read retention together with CTR

Retention never tells the full story alone. A video can hold its audience well and still reach few people because the title and thumbnail do not earn clicks. A video can also earn many clicks and then lose viewers quickly because the package promised something the opening did not deliver. YouTube's help page on click-through rate names high CTR combined with low average view duration as a sign of clickbait.

So when you review a video, look at three numbers side by side: impressions click-through rate, the intro figure and average view duration. The combination tells you whether to fix the package, the opening or the body of the video.

Comparing videos fairly

Retention is easy to misread if you compare the wrong things:

  • Compare similar lengths. A 20-minute documentary and an 8-minute explainer will never have the same percentages. YouTube's typical retention comparison already uses videos of similar length.
  • Compare similar formats. A ranking video and a story video lose viewers in different places.
  • Check the traffic mix. A video mostly watched from search behaves differently from one pushed on the homepage.
  • Look at new and returning viewers separately.

Keep a simple log for every video: intro figure, the timestamps of the biggest dips with their causes, and one rule for the next script. Over a dozen videos, the log shows which problems repeat.

Turning the graph into the next script

The value of the retention report is not in the number. It is in the rules you write down afterwards. Some examples of rules that come out of this process:

  • "Never start with background. Open on the event."
  • "Every section starts with a question."
  • "No graphic with more than eight words on screen for less than four seconds."
  • "Deliver the main answer in the last 15 percent of the video, not the middle."
  • "No reading lists of dates aloud. Put them on screen."

Apply them to the next script, then check whether the same dips come back.

Checklist

  • I wait one to two days after upload before reading retention.
  • I compare against typical retention, not against other channels.
  • I check new viewers separately from returning viewers.
  • I review the intro figure against the title and thumbnail promise.
  • For each major dip, I found the script paragraph and named the cause.
  • For each spike, I decided whether it was interest or confusion.
  • I wrote at least one rule for the next script.
  • I keep a running log of dips and causes across videos.

VidSwarm is built to connect to YouTube Analytics with read-only access and carry what your retention and CTR data show into the next hook, title and thumbnail for that channel, so these lessons do not stay in a notebook.

Try it on your own niche.

One real film up to 5 minutes, in your channel’s look.

Get a $9 test-film seat