Five YouTube Retention Stats That Won't Die (And What's Actually True)
The viral retention numbers creators repeat - 23.7% average, 55% drop in 60 seconds, 50% AVD unlocks the algorithm - traced to their sources. Most are made up.
Scroll any thread about YouTube retention and you will hit the same handful of numbers: "the average video retains 23.7%," "55% of people leave in the first 60 seconds," "get above 50% average view duration and the algorithm hands you 3x the reach." They sound authoritative because they are specific. The problem is that when you try to find where they came from, there is nothing there.
I run a channel and I built a scripting tool, so I went looking for the primary sources behind the retention advice everyone repeats. Most of the strategic stuff holds up. Almost none of the specific percentages do. They trace back to SaaS and agency marketing blogs quoting each other, never to YouTube, a study, or a dataset you can actually inspect.
This is the one post on this site where I am going to write the fake numbers down, because you cannot unlearn a myth you cannot see. Every figure in the table below is quoted so I can debunk it. None of them are endorsed. After each one, I will show you what the primary sources - YouTube's own growth team and creators like Paddy Galloway who publish their process - actually say instead.
The Stats, and the Verdict on Each
| The Claim | Verdict | What's Actually True |
|---|---|---|
| "The average video retains 23.7%" | Fabricated | No primary source - a made-up decimal |
| "55% drop off in the first 60 seconds" | Fabricated | No YouTube data behind it |
| "50% AVD unlocks 3x recommendations" | Misleading | YouTube ranks per viewer, no AVD target |
| "Pattern interrupt every 90s = +15-25% AVD" | Fabricated | Pacing helps, the number is invented |
| "Scripted 40-60% vs unscripted 25-35%" | Fabricated | No study - scripting helps other ways |
| "AVD is the only signal that matters" | Misleading | Satisfaction signals count too |
Every claim above is quoted here only to debunk it. None are verified, and none should be repeated as fact.
How to Spot a Fake Retention Stat
Before we get to the individual myths, it helps to know the shape of a fabricated stat, because once you see it you cannot unsee it. Real measurement is messy. Retention varies by video length, by niche, by whether the traffic came from search or browse, and by who your audience is. So any number that claims to be the single, universal "average retention" for all of YouTube is already a red flag before you even check the source, because YouTube's own growth lead has said on the record that there is no single benchmark number to chase.
There are three tells. First, suspicious precision. A figure like "23.7%" implies someone measured a dataset to a decimal place and then never showed it to anyone. Second, no traceable primary source. When you actually try to follow the citation, it dead-ends in another blog, which cites another blog, which cites nothing. Third, it is phrased like a law of physics ("videos over 50% average view duration get 3x the reach") rather than the honest version ("on our channel, in our tests, we saw...").
Todd Beaupré, YouTube's Senior Director of Growth and Discovery, described the recommendation system on Creator Insider in January 2025 as pulling content toward each individual viewer rather than pushing your video out to a fixed audience. He was clear that click-through rate and average view duration are inputs to a per-viewer prediction, not targets you optimise to a magic number. That one idea quietly kills most of the stats below.
So here is the test. Could this number survive being true for a 45-second cooking Short and a 45-minute video essay at the same time? If not, it is not a universal stat, and you should benchmark against your own channel's past videos instead of a stranger's decimal.
Myth 1: "The average video retains 23.7%"
The claim, as creators repeat it: the "average" YouTube video retains something like 23.7% of its viewers, and more than half your audience - usually quoted as 55% - is gone within the first 60 seconds. You will see both figures in video intros, on thumbnails, and on course sales pages, presented as if YouTube published them.
Why it does not hold up: there is no primary source for either number. YouTube has never released a platform-wide average retention figure, and if it did it would be close to meaningless, because a 60-second video and a 30-minute video cannot share a single "average," and search traffic behaves nothing like browse traffic. The 23.7% figure and the "55% in the first 60 seconds" figure both dead-end in SaaS and agency marketing blogs quoting each other. Neither links to a dataset, a YouTube post, or a study you can inspect. They are false as stated, and the decimal is the giveaway.
What is actually true: your retention baseline is your own back catalogue, not a universal constant. Beaupré's framing is that there is no absolute good number - you benchmark against your channel. The same average can be strong for a long video essay and mediocre for a two-minute explainer, which is exactly why a single platform-wide figure is meaningless. There is a real, normal drop in the opening seconds as people who misread the title bounce, and it is worth watching, but "55% in 60 seconds" is invented precision dressed up as data.
The practical version is boring and useful: open YouTube Studio, look at your last ten videos, and note your own typical first-30-seconds drop and your own average retention. That number is the one to beat. It is the only "average" that actually applies to you.
Myth 2: "50% AVD unlocks the algorithm"
The claim: there is a magic average-view-duration threshold, usually stated as 50%, and the moment a video crosses it YouTube flips a switch and pushes it into the Suggested feed. It is often quoted with a multiplier attached - "videos over 50% AVD are about 3x more likely to be recommended."
Why it does not hold up: this one does not just lack a source, it contradicts what YouTube's own team has said out loud. Beaupré described the system as pulling per individual viewer, ranking each recommendation on a personal prediction of whether that specific person will want to watch. Aggregate AVD is one input into that prediction, not a gate with a fixed cutoff. There is no "50% unlocks the algorithm" switch, because there is no single audience the switch would open the door to - the system is deciding viewer by viewer.
The "3x" multiplier has the same disease as 23.7%: it is a clean, quotable number with no measurement behind it. It sounds like it came from a YouTube engineer and actually came from a blog call-to-action. Two videos with an identical 50% AVD can perform completely differently, because one satisfies its audience and the other does not.
What is actually true: AVD matters as a signal, but as a per-viewer input, and there is no threshold to game. Chasing a percentage can even backfire. Padding a video to lift its average duration gives viewers more chances to get bored and leave, which is the opposite of what you were trying to do. The number you cross is not what earns the reach - the satisfied viewer is.
Myth 3: "A pattern interrupt every 90s adds 15-25% AVD"
The claim: change your camera angle, cut to b-roll, or drop a graphic on a fixed timer - usually every 90 seconds, sometimes every 30 to 45 - and you will add a specific 15 to 25 percent to your average view duration.
Here is where it gets more subtle, because the qualitative advice is fine. Varying the visuals and re-hooking attention genuinely helps hold viewers, and it is practitioner consensus that a good long-form video keeps introducing new information rather than flatlining. What is fabricated is the number. No source measured a 15-25% lift, and the idea that a fixed-interval timer produces a fixed percentage gain treats retention like a vending machine. A pattern interrupt that lands on a boring moment does nothing. One that arrives right as a curiosity loop pays off does a lot. Timing to content beats timing to a stopwatch.
What is actually true: think in pacing, not percentages. A useful heuristic from the way experienced creators structure long-form is a setup-tension-payoff beat roughly every 90 to 120 seconds, plus a re-hook - a forward reference to a payoff still coming - somewhere past the eight-minute mark on longer videos. That is a rhythm to write to, not a guaranteed AVD lift.
The goal behind all of it is simple: a viewer should never go more than a minute or two without a fresh reason to keep watching. How much that actually helps depends entirely on your content and your audience, which is exactly why nobody can honestly stamp a single percentage on it. When someone does, they made it up.
Myth 4: "Scripted videos get 40-60% vs 25-35% unscripted"
The claim: scripting your video roughly doubles your retention, from a "25-35% unscripted" band up to a "40-60% scripted" band. Sometimes it gets dressed up further as "structured scripts get 3.2x higher average view duration."
Why it does not hold up: no study produced these bands. They are two invented ranges placed side by side to sell scripting, and, ironically, scripting tools. I make one, and I am telling you the stat is fake. Retention is not decided by whether you used a script. It is decided by the idea, the packaging, and whether the content pays off the click. A tightly-improvised video from someone who knows their material can beat a wooden read of a weak script every time.
What is actually true: scripting helps, but for reasons that do not collapse into a single percentage. Writing in advance lets you engineer the hook to validate the title's promise in the first 20 to 30 seconds, cut the throat-clearing, and control your value-per-minute so you are not padding. Beaupré's point that viewers reward videos that get to the point efficiently is the real mechanism at work - a script is just the easiest place to enforce that discipline.
On word budget, a natural speaking pace is roughly 130 to 150 words per minute, so a 10-minute video is about 1,300 to 1,500 words. A script mainly helps you hit that number without rambling into dead air. The benefit of scripting is real and worth it. The "40-60% versus 25-35%" comparison is not.
Myth 5: "AVD is the one signal that matters"
The claim: retention, and specifically average view duration, is THE metric. Maximise watch time and everything else follows. This one is less a fake number than a fake priority, and it is the most quietly damaging of the lot.
Why it does not hold up: YouTube's own growth team has said the opposite. Beaupré was explicit that not all watch time is valued equally, and that YouTube feeds satisfaction signals - responses to millions of in-product surveys, plus likes and dislikes - into ranking alongside raw engagement. Adding those satisfaction signals, he said, increased long-run return visits. So a video that holds you hostage for ten minutes and leaves you feeling tricked is not the same, to YouTube, as one that respects your time, even when the raw watch time looks identical.
This is why "just maximise AVD" can hurt you. Padding, clickbait that over-promises, and slow-walking the payoff can all keep the AVD line propped up while training your audience to trust you less. That erosion shows up later as fewer return visits and weaker survey signals, which is the part the vanity metric never told you about.
What is actually true: watch time matters, but as one of several signals, weighted by whether viewers were actually satisfied. The practical takeaway is to earn the watch time honestly - deliver the exact thing the title promised, get to the point, and close every loop you open.
What Is Actually Worth Watching
So if the popular numbers are fake, what actually matters is smaller and less exciting than a magic threshold: your own retention curve, and your own view counts. That is genuinely most of it.
Paddy Galloway, who publishes his process openly, has said that view counts and retention curves are the two data points worth analysing, and that basically everything else is secondary. Notice what that is not: it is not a single AVD percentage to chase across the whole platform. It is the shape of the curve on your video, read against your own baseline.
Read the shape, not a number. The drop in the first few seconds is normal - watch how steep it is compared to your other videos, not against a made-up "55%". Look for mid-video dips, the sudden cliffs where a tangent or a slow stretch lost people, because each one is a specific moment you can cut next time. Look for spikes, where the line goes flat or ticks up because people rewatched or the section was worth staying for, and do more of whatever caused them.
Value-per-minute matters more than raw duration. Beaupré's satisfaction framing and Galloway's "the idea sets the ceiling" point the same way: the win is delivering the promised payoff efficiently, not stretching a video to inflate a stat. Benchmark every new upload against your own last ten videos, fix the worst dip, and repeat. That loop beats any universal percentage.
That is also the standard we hold ourselves to when building Scripti. It structures a script around a hook that validates the title, a pacing rhythm that keeps re-hooking attention, and value-per-minute discipline - all things that genuinely help - without ever promising you a fabricated "+15-25% AVD." If a tool or a course quotes you a retention stat with a decimal point and no source, that is a reason to trust it less, not more.
How to Sanity-Check Any Retention Stat
Continue Reading
From idea to record-ready
Write for real retention
See whether a script earns attention through its hook, pacing, and value per minute, without invented benchmarks.
No account or card required.