How to increase audience retention on YouTube: reading the graph and fixing the dips
Direct answer: To raise audience retention, work on three things in YouTube Studio's retention report: the first 30 seconds (YouTube's "Intro" metric), the dips where viewers skip or leave, and a length your audience actually finishes. YouTube publishes no "good" percentage, so compare each video with your last 10 of similar length, then fix what each dip points to.
Where is the retention graph, and what does YouTube say each moment means?
In YouTube Studio: Content, then the video's Analytics, then the Overview or Engagement tab and the Audience retention report. The graph shows what share of viewers is still watching at each moment of the video. YouTube labels four kinds of key moments; here they are in its own words (YouTube Help, read October 13, 2026):
| Key moment | YouTube's definition | What it tells an editor |
|---|---|---|
| Intro | "Intro tells you what percentage of your audience still watched your video after the first 30 seconds." | Whether the opening delivered the title's promise fast enough (hooks guide). |
| Top moments | "Top moments are moments in your video where almost no one dropped off while watching." | What to make more of in the next video. |
| Spikes | "Spikes are moments in your video that were rewatched or shared." | Either the best moment or a confusing one: watch it to know which. |
| Dips | "Dips highlight moments in your video that were either skipped or moments where viewers stopped watching your video completely." | Where to cut, tighten or show more. |
Two conditions and one warning from the same page. Key moments only show up when the video meets this: "Your video should also be at least 60 seconds long and have at least 100 views." And a falling line is normal: "Videos on YouTube generally taper off during the playback period."
Absolute and relative retention
The curve is absolute: the share of viewers left at each second. To judge it, YouTube gives two comparisons. SEE MORE shows "how your video compares to all YouTube videos of similar length", and typical retention lets you "compare your 10 latest videos of similar length". The second is the one to use week after week, because it compares you with yourself in the same format.
What is a good audience retention rate? (every public benchmark, with its source)
YouTube does not publish one. Third parties do, and they disagree. Every figure below was read on its own page on October 13, 2026.
| Source | What it calls good | What it is based on | Link |
|---|---|---|---|
| YouTube Help | No number. Compare with "all YouTube videos of similar length" and your "10 latest videos of similar length". | Official | support.google.com |
| vidIQ, updated June 29, 2026 | About 30% average percentage viewed for a 6 to 9 minute talking-head or education video; about 60% still watching after 30 seconds, over 65% for the strongest hooks. | Its own channel: 30.3% average over 16.5M long-form views (2.39M subscribers); six of its videos at 38.1% to 67.9% at 30 seconds | vidiq.com |
| Subscribr, no date | 30–50% average; 50–70% at 30 seconds; 40–60% at 60 seconds. | "general benchmarks often cited by YouTube strategists", no source named | subscribr.ai |
| OpusClip, November 17, 2025 | Losing more than 40% in the first 30 seconds means the intro needs work; 50%+ average viewed is strong under 10 minutes, 40% solid for longer videos. | The author's rules of thumb, no source cited | opus.pro |
| Filmora (Spanish), September 29, 2026 | "alrededor del 50%" (about 50%). | "la comunidad de YouTube", no source named | filmora.wondershare.es |
Uppbeat's retention guide ranks for the same question but could not be read (HTTP 429 twice on October 13, 2026), so it is not in the table.
How to read it: the averages run from 30% to 50%, and the figures for the 30-second mark from 50% to 70%. The only measured numbers are vidIQ's, from one channel about YouTube itself. So the comparison that means something is with your own last 10 videos of similar length. When you compare videos of different lengths, compare minutes, not percentages: 35% of a 30-minute video is 10.5 minutes watched; 60% of a 3-minute video is 1.8.
The dips, one by one: what causes them and the edit that fixes each
In the client graphs I review, the same five shapes come back. The minute where they appear changes from channel to channel; the shape is what to look for. For each one: what you see, the usual cause, the cut that fixes it and how to check.
1. A drop in the first 30 seconds
Cause: the promise arrives late, behind a logo, a greeting, a recap or a slow first line. Fix: start on the line or the moment that matches the thumbnail, with the voice in the first second (43 of 50 recent videos from 25 big channels start that way). Check: the Intro metric against your last 10 videos of similar length. The full method, with seven real openings, is in how to hook viewers in the first 30 seconds.
2. A step down where the intro hands over to the content
Cause: a setup longer than the payoff needs. In the graphs I see, this step sits where the setup ends and the main content begins; its minute depends on the video. Fix: move the first payoff forward and cut the setup to what the payoff needs (what retention-led editing actually means). Check: the step gets smaller over the next videos in the same format.
3. A vertical drop at one second
Cause: dead air, a confusing cut, an audio drop or a long sponsor read. Fix: trim the wait, replace the cut, level the audio. Check: watch that exact second and note what changes. A pause in the voice is not automatically dead air: in those 50 big-channel videos the longest voice pause has a median of 8.6 seconds. A pause with nothing happening on screen is the problem.
4. A long, slow slope through an explanation
Cause: nothing new on screen for a stretch. Fix: one new image per idea, and text on screen where the words carry the video. In the documentaries I measured, text is on screen 19.6% of the time in one channel and 34–42% in another (pacing data). Check: the slope flattens in the next video.
5. A spike
YouTube's definition: "moments in your video that were rewatched or shared". It is either the clearest payoff or something viewers had to replay to understand. Watch it: if it is the payoff, put a taste of it earlier; if it was unclear, put it on screen.
Does faster cutting fix retention? What the measurements say
Not by itself. In the same genre, I measured one spy documentary channel at about 2 picture changes a minute (median shot 29 seconds) and another at about 14.7: same subject, roughly six times apart (pacing data). Inside one video the pace moves: Philip Thompson's documentary runs about 19 changes a minute in its 61-second cold open, 14.5 in minutes 0–10, 12.5 in minutes 10–25, 10.8 in minutes 25–30 and 8.2 at the close. And across 24 big channels, only 11 cut their first minute faster than the rest of the video, while 6 open slower (opening data). Pace is a tool for a specific dip, not a dial to turn up.
Length and watch time: how long are the videos people actually finish?
Long is not the enemy. The 50 most-viewed recent long-form videos from 25 big channels have a median length of 28.4 minutes, and 33 of them run 20 minutes or more (loudness study).
In my own work: of the 1,159 long-form client videos tracked on my YT Jobs profile (lengths read with the YouTube Data API on October 13, 2026), the median runs 12.1 minutes and 98% run 8 minutes or more. That is the length YouTube sets for mid-roll ads: "On monetized videos that are 8 minutes or longer, you can turn on ads during the middle of the video" (YouTube Help).
Does one length do better than another? Within the same channel, here is how each length performs against that channel's typical video:
| Length | Videos | Median views vs the channel's typical video | Share above the channel's median |
|---|---|---|---|
| 8–10 min | 228 | 0.78× | 40% |
| 10–20 min | 447 | 1.05× | 54% |
| 20–35 min | 68 | 0.93× | 47% |
| 35+ min | 113 | 1.13× | 54% |
Source: YouTube Data API, October 13, 2026. 859 videos from the 7 client channels with at least 20 videos older than 90 days (ashlele, Mud, Puff, Rex, DakBlox, Puff Blox, Crazz); 1.00× = as viewed as the channel's typical video. Videos under 90 days old and Shorts are left out.
What it does and does not say: within each length the spread (from about half to about double the channel's typical video) is bigger than the difference between lengths, so the topic matters more than the minutes. It is correlation on one editor's clients, not a cause, and it says nothing about retention: the public API does not give it. One more caution when you compare over time: since August 24, 2026 a public view counts "from the very first frame", while retention and average view duration are still measured on engaged views (YouTube, August 19, 2026).
A real case: a steady pace, not tricks
Mud runs a Roblox Rivals news channel. I started editing it on March 5, 2025; through October 2, 2026 that is 258 videos, about 3.1 a week. The monthly median first passed 100K views in December 2025 (136,700), about nine months after I started. The channel's quarters, read with the YouTube Data API on October 13, 2026 like that monthly figure (videos at least 90 days old):
| Quarter | Videos | Videos a week | Median views |
|---|---|---|---|
| 2025 Q2 | 39 | 3.00 | 66,082 |
| 2025 Q3 | 40 | 3.04 | 73,122 |
| 2025 Q4 | 42 | 3.20 | 86,540 |
| 2026 Q1 | 41 | 3.19 | 157,601 |
| 2026 Q2 | 41 | 3.15 | 114,077 |
Same rhythm, 3.00 to 3.20 videos a week, and the median went from 66,082 to 157,601 in three quarters. Two honest notes. It was not a retention trick: the pace held at about three videos a week and the median views rose over the same quarters (the full case). And I do not publish Mud's retention graphs, because they are Mud's; the figures above are public views.
"The difference between the people who make it and the people who don't is that they keep trying until it works, and they enjoy the process. I've seen creators work six months and not see results until month eight. Others get results in the first month, others after a year or two. Every process is different."
Kevin Tabares, from working with clients
Tools: measure before you cut
- Retention improvement estimator: turns a change in retention into minutes watched.
- Retention diagnostic checklist: ten checkpoints across the whole video.
- Loudness checker: whether your export sits where big channels do.
- youtube-openings-2026.csv: how 50 big-channel videos open (CC BY 4.0).
FAQ
What is a good audience retention rate on YouTube?
There is no official number. Published third-party ranges run from 30% to 50% average percentage viewed, and vidIQ's own channel sits at 30.3% over 16.5 million long-form views. Compare each video with your last 10 videos of similar length, as YouTube suggests.
Why does my retention drop at the start?
YouTube's Intro metric covers the first 30 seconds. The usual cause is a promise that arrives late: a logo, a greeting or a recap before the moment the title and thumbnail promised. In 43 of 50 recent videos from 25 big channels the voice starts in the first second.
What does a spike in the retention graph mean?
YouTube defines spikes as moments that were rewatched or shared. It is either the clearest payoff of the video or something viewers had to replay to understand. Watch it and decide which: move a taste of the payoff earlier, or put the unclear part on screen.
Does a longer video lower retention?
The percentage usually falls with length, but watch time can still rise. 33 of the 50 most-viewed recent videos from 25 big channels run 20 minutes or more, with a median of 28.4 minutes. Compare minutes watched, not only percentages.
Can editing alone fix low retention?
It fixes dead air, late payoffs, confusing cuts and audio problems. It cannot invent a promise the footage does not keep or change what the video is about. Check title, thumbnail and the first 30 seconds together before you decide.