How to actually read your YouTube retention graph
YouTube Studio's Audience Retention report labels four specific things on the curve, Intro, Top Moments, Dips, and Spikes, and most creators glance at the general shape of the line without ever learning what each label is actually telling them. YouTube's own documentation is specific about what each one means, and knowing the difference turns a vague "retention looks okay" into an actual diagnosis of what to fix and where.
What each label actually measures
| Label | What it shows |
|---|---|
| Intro | The percentage of viewers still watching after the first 30 seconds |
| Top Moments | Segments where almost no one dropped off |
| Spikes | Moments that were rewatched or shared |
| Dips | Segments that were skipped, or where viewers stopped watching |
Source: YouTube Studio Help, Audience Retention report.
Reading the shape, not just the labels
A flat line means viewers are watching that section completely; a gradual decline means interest is fading over time rather than dropping at one specific point; an upward movement signals rewatching or sharing; a sharp downward movement signals abandonment or skipping right at that timestamp. The graph is telling a story about a specific audience's attention moment by moment, and the shape matters as much as any single percentage.
Intro problems vs. Dip problems
A weak Intro number points at the hook itself, the same territory our hooks guide covers: the opening isn't earning the sustained attention the thumbnail and title already earned as a click. A Dip well into the video, especially one that never recovers, points somewhere different: a specific section that isn't holding up, which might mean tightening the pacing there, cutting a tangent, or reconsidering whether that part of the topic belonged in this video at all. Different diagnoses, different fixes, and the graph tells you which one you're actually looking at.
Spikes are free research
A Spike is direct evidence of what a specific audience actively rewatches or shares, more reliable than guessing from memory what "worked." Note the exact moment and what happened there, a reveal, a punchline, a specific visual, and treat it as a pattern worth deliberately repeating rather than a lucky accident.
Where Thothium fits
Thothium keeps every scene editable after the fact, so once a retention graph points at a specific weak Intro or an unrecovered Dip, fixing that exact section is a targeted edit rather than reworking the whole video. It is in free alpha, and the form below gets you a key.
Frequently asked questions
What counts as a bad Intro percentage?
YouTube doesn't publish a universal cutoff, and the right number varies by niche and format, but a sharp drop before the 30-second mark closes is the signal worth investigating first. Compare it against your own channel's other videos rather than an external benchmark; your own history is the more relevant baseline.
Are Dips always a problem?
Not automatically. A small dip during a slower explanatory section is normal and often recoverable if retention climbs back afterward. A dip that never recovers, where the line keeps sliding rather than leveling off, is the pattern that actually signals lost interest rather than a single rough patch.
What should I actually do when I see a Spike?
Note exactly what's happening in the video at that timestamp, since it's a moment your specific audience is rewatching or sharing. That's a concrete, evidence-backed data point for what to do more of, more reliable than guessing at what "worked" from memory.
Does a strong retention graph guarantee a video does well?
It's one input into the algorithm's decision, alongside click-through rate and other signals, not a guarantee on its own. Our CTR guide covers the other half of that pair, since a video can retain viewers well and still underperform if it isn't earning clicks to begin with.
Last updated September 8, 2026. Feature labels and definitions are per YouTube Studio's own Help documentation, current as of this writing; check YouTube Studio directly for your own video's data.