Audience retention tells YouTube whether people actually watched a video. It’s one of the strongest signals YouTube uses when deciding whether a video should get recommended. Audience retention is the percentage of a video that the average viewer stays for. Average view duration (AVD) is the same underlying data expressed in minutes instead of percentage. Both feed directly into how much YouTube recommends a video.
|
Metric |
What it measures |
Where in Studio |
|
Audience retention (%) |
% of the video the average viewer watched |
Analytics → Engagement → Audience retention |
|
Average view duration (AVD) |
Average minutes watched per view |
Analytics → Engagement → Key moments |
|
Average percentage viewed (APV) |
AVD ÷ video length × 100 |
Same section as AVD |
|
Retention curve |
Moment-by-moment % of the audience still watching |
Analytics → Engagement → graph view |
What Is YouTube Audience Retention?
Audience retention is the percentage of a video that the average viewer watches.
A 10-minute video with 50% audience retention has an average view duration of 5 minutes, which means that, on average, viewers made it through about half the video before dropping off.
YouTube also tracks retention moment by moment: at each timestamp, what percentage of the original audience is still watching. That moment-by-moment view is the retention curve.
What Is Average View Duration (AVD)?
AVD is the average number of minutes viewers spent watching a video, the absolute-time version of the same underlying retention data.
It works out to total watch time divided by total views: 1,000 views generating 3,000 total minutes watched gives an AVD of 3 minutes.
What Is Average Percentage Viewed (APV)?
APV expresses AVD as a percentage of the video’s runtime: AVD divided by video length, times 100.
This makes it possible to compare retention across videos of different lengths on equal footing. APV can exceed 100% when viewers replay sections, which shows up most often on tutorial or how-to content where people rewind to catch a step again.
What Is the Difference Between AVD and Audience Retention?
|
AVD |
Audience retention (%) |
|
|
Unit |
Minutes |
Percentage of video watched |
|
Useful for |
Comparing a channel's absolute watch time over time |
Comparing retention across videos of different lengths |
|
Context-dependent? |
Yes. 3 minutes of AVD means something different on a 4-minute video than on a 40-minute one |
Less so. Percentage normalizes for length |
|
Shows drop-off points? |
No |
Yes, via the retention curve |
AVD is the number to track for a channel's absolute watch-time trend over time. Retention percentage (and APV specifically) is the number to use when comparing how individual videos of different lengths performed relative to each other.
What Is a Good Audience Retention Rate on YouTube in 2026?
YouTube doesn’t publish official retention benchmarks, and video length is the single biggest driver of what a “normal” percentage looks like (shorter videos post higher retention percentages than longer ones even when both hold viewers equally well in absolute terms).
Directional figures from creator and agency data suggest APV commonly runs 50-70% on videos under 5 minutes, tapering down toward 25-35% on videos over 40 minutes. A 30% APV on a 60-minute video still represents 18 minutes of genuine viewing, which is a strong result in absolute terms, even though the percentage looks low next to a short-form benchmark.
The only reliable comparison is against videos of similar length, ideally your own.
In one of AIR’s tracked cases, a crafting channel’s AVD moved from 1 minute 56 seconds to 5 minutes 58 seconds after a format change, which is a reminder that AVD numbers only mean something next to a comparable video. A flat industry number means very little against your own channel stats.
The Retention Curve: What the Shape Means
|
Pattern |
What it signals |
|
Steep drop in the first 30 seconds |
The intro isn't delivering on what the title or thumbnail promised |
|
A sharp dip mid-video |
A pacing issue, tangent, or confusing section at that specific timestamp |
|
Gradual decline to the end |
Normal, healthy attrition as the content plays out |
|
A spike or plateau |
A moment viewers rewatched or a segment that held attention unusually well |
|
A drop just before the end |
The outro runs long, or viewers already got what they came for |
How Retention Feeds the YouTube Algorithm
Retention and CTR are both among the algorithm's highest-weighted quality signals, though satisfaction signals have moved above raw watch-based metrics as the primary ranking input as of 2026. A video with fewer views but stronger retention can still end up recommended more, over time, than one with more views and weaker retention, because retention tells YouTube that the people who watch are getting real value from it.
Retention is most informative when read together with CTR:
A video that gets clicks but loses viewers quickly usually means the title or thumbnail promised something the content didn’t deliver, while strong CTR paired with strong retention is the combination that tends to expand a video’s reach the furthest.
Where to Find Retention Data in YouTube Studio
There isn't a channel-wide audience retention curve; that report only exists at the video level. What you do get at the channel level is aggregate Average View Duration and Average Percentage Viewed, under:
YouTube Studio → Analytics → Engagement
For the full retention curve and key moments on a specific video:
YouTube Studio → Analytics → Content → select a video → Engagement tab → Audience retention
The retention curve includes a "typical retention" comparison line, but it's worth knowing what it's actually comparing against: YouTube builds it from your own 10 most recent videos of similar length. That makes it a useful gut-check against your own recent output.
Below 60 seconds or 100 views, YouTube may not detect or highlight the four "key moments" patterns (intros, top moments, spikes, dips) on the graph; moments are only highlighted when detected, and the video needs to clear that length and view threshold first.
