Most videos lose between 20% and 35% of their viewers in the first 30 seconds. That pattern holds at 800 subscribers and at 8 million. What separates channels that grow from channels that stall is rarely the overall shape of the curve, because every curve looks roughly the same. It is how steep the opening drop is, and how flat the line runs after it.
Here is how to read the numbers in 2026, with benchmarks by video length and a straight answer on when a graph is actually a problem.
The anatomy every retention graph shares
Open the audience retention report on any video and you will find three regions.
The cliff (0:00 to 0:30). The steepest loss on the whole graph. Some of it is unavoidable: misclicks, autoplay viewers, people who wanted a different video. The rest is your hook failing.
The slope (0:30 to about 70% of runtime). A gradual decline, ideally 1% to 2% lost per minute on a long-form video. Sharp dips here mark specific moments where you lost people: a tangent, a sponsor read, a promise you stalled on delivering.
The tail (the last 20% or so). Retention falls faster as viewers sense the video wrapping up. A tail dive is normal. A tail dive that starts at the halfway point means your video is twice as long as its content.
Benchmarks by video length
These are working numbers, not laws. Niche matters: gaming and entertainment run higher than tutorials, and tutorials run higher than podcasts, partly because people leave a tutorial the moment their problem is solved.
Under 5 minutes
Average view percentage of 55% to 65% is solid. Under 45%, the video has a structural problem, usually a slow first 20 seconds that the short runtime cannot recover from.
8 to 12 minutes
The most common long-form length and the easiest to benchmark: 45% to 55% average view percentage is healthy, 40% is survivable, and above 60% is a video worth studying and repeating. A 10-minute video at 50% means the average viewer watched 5 minutes, which is enough for the algorithm to read sustained interest.
20 minutes and up
Percentages drop and that is fine. A 25-minute video holding 35% delivers almost 9 minutes of average watch time, which beats a 10-minute video at 55%. At this length, judge absolute watch time first and percentage second.
Shorts
Different metric entirely. Watch completion rate and average percentage viewed, which can run above 100% when people loop. A Short under 35 seconds should aim for 85% or higher average percentage viewed. Past 50 seconds, holding 70% is strong.
The 0:30 cliff, specifically
YouTube's own creator guidance has pointed at the same threshold for years: videos that hold more than 70% of viewers through the first 30 seconds tend to outperform. That single number explains most retention problems you will ever diagnose.
Run the math on what the cliff costs. If you keep 55% at 0:30 instead of 70%, you have lost 15 points of your entire potential watch time before the video really started. The algorithm then sizes its next round of impressions on what those first viewers did, so the cliff compounds into reach.
What 50% at 30 seconds means
It means half your audience decided the video was not for them in less time than it takes to read this paragraph. Three causes account for nearly all of it:
- The packaging promised something the opening does not show. The thumbnail said "I built a sauna for $300" and seconds 1 through 25 are you explaining why you wanted a sauna.
- The intro is throat-clearing: a greeting, a channel pitch, a subscribe ask, a preview of what the video will cover instead of the video itself.
- The first frames are static. One camera angle, no movement, no cut for 20 seconds reads as low effort before a single word lands.
50% at 0:30 is recoverable, because it is almost always a scripting problem, and scripting problems fix faster than content problems.
How to read your own graph
Skip the averages and read shapes.
Flat is the goal. Any stretch where the line runs nearly horizontal is a section that held everyone. Find what you did there and do it again.
A dip that recovers marks a skippable segment. Viewers scrubbed past it and came back. Usually a sponsor read or a tangent. Tighten it or move it later.
A dip that does not recover is a leak. Whatever happened at that timestamp sent people to another video. Watch the 15 seconds before the dip starts, because the cause usually sits earlier than the drop.
A spike means people rewatched a moment. That is your best material announcing itself. Spikes make excellent cold opens for the next video on the topic.
Compare against your own channel average before comparing against anyone else. Beating your own last 10 videos matters more than an abstract benchmark, because the algorithm grades you on improvement within your audience, not against a global table.
When to worry
Worry when the cliff steepens across several consecutive videos. That is packaging or hook decay, and it compounds: weaker opens mean weaker signals mean fewer impressions.
Worry when dips land at the same relative spot every video. A consistent dip at the 60% mark means your structure runs out of promised material there, and the fix belongs in the script, before you record.
Do not worry about one video with rough retention but a strong click-through rate, an experiment that flopped, or a percentage drop right after you doubled your average video length. And do not chase a flat line at 100%. It does not exist. The best video on your channel right now probably loses a quarter of its viewers in the first minute. The work is making the next one lose a fifth.