Two rules dominate every discussion of TikTok video length. Keep it under fifteen seconds. Aim for a hundred percent completion. People say them in the same breath, as though they were one idea. Both sound like something somebody must have tested.
They are not one idea. In fact they pull in opposite directions. Only one of the two numbers is one TikTok has ever said it weighs. Getting them confused is how accounts end up making videos that finish beautifully and go nowhere.
What TikTok actually says about watch signals
TikTok’s own description of the For You feed is the only first-party statement on this. It rewards a close reading, mostly for what it does not commit to.
In how it recommends videos, TikTok says a strong indicator of interest carries greater weight than a weak one. Its example of a strong signal: a user finishing a longer video from beginning to end. Its example of a weak one: whether viewer and creator sit in the same country.
Read that example closely, because the usual paraphrase drops half of it. The strong signal is not “finished the video”. It is “finished a longer video”. The length is doing work in that sentence. TikTok never equates a completed three-second clip with a completed ninety-second one, and the document does not hint at it either.
The same page says two other things worth holding on to. Follower count is not a direct factor. How your previous videos performed is not a direct factor either. TikTok assesses every video on what happens when it reaches people. That is why the watch signals carry so much weight. On a new post they are close to the only evidence the system has.
What TikTok has never published is a formula. It states no weighting for completion rate, no threshold, no curve. Anyone quoting you one invented it. The next section explains why they get away with it.
Completion rate and watch time are not the same number
Completion rate is the share of viewers who reached the end. Watch time is how long they actually watched. On a single video the two move together. Across videos of different lengths they come apart fast, and the arithmetic is not subtle.
- 10 seconds at 100% completion: 10 seconds watched per view.
- 34 seconds at 45% completion: about 15 seconds per view.
- 60 seconds at 30% completion: 18 seconds per view.
The worst completion rate on that list produces nearly twice the watch time of the perfect score. Optimise for the percentage and you would kill the best video on the board. Its only crime is an ugly number.
This is the whole problem with the hundred percent target. Completion rate is a ratio, and you chose the denominator before you filmed anything. You could hit ninety percent tomorrow by cutting every video to four seconds. You would have raised the ratio by making each view worth less. A ratio you improve by shrinking what you measure against is not a goal. It is a thermometer you are holding over a candle.
Short videos have a second effect that inflates the same number. They loop. A viewer who lets a six-second clip run round twice registers as having watched it through. On very short posts, your average watch time can exceed the length of the video itself. That is a real signal of interest. But it means completion on a short video and completion on a minute-long one measure different things, and stacking them in one column is a mistake.
Why nobody can quote you a real completion benchmark
Here is the part that explains why this advice circulates unchallenged. Completion rate is private.
Likes, comments, shares and saves sit on the side of every video where anyone can read them. So we could measure save rates across real accounts for our piece on saves versus likes. The counts were public. The work was arithmetic and patience. Watch time and completion rate reach nobody but the account owner. They appear in TikTok Studio for your own posts and nowhere else at all.
An article tells you the average completion rate is 38%, or that you need 50% to get pushed. Ask where that number came from. It cannot come from watching other people’s videos, because that observation is impossible from outside. Three places produce these numbers. One agency’s own client accounts, stretched well past what that sample supports. A survey of creators repeating figures they half-remember. Or nowhere.
We are not going to add a fourth invented number to that pile. We measure things here when the data is public and we can show the method. This metric is not public. So what follows is mechanism and arithmetic, both of which you can check yourself, rather than a benchmark we would have had to make up.
The length decision, priced honestly
Stop treating completion as the target and video length becomes what it always was. A trade-off with a cost on both sides.
Going longer raises your ceiling. Sixty seconds of material can produce sixty seconds of watch time per viewer. Fifteen seconds cannot, however perfectly you make it. Going longer also raises your risk, because every extra second gives the viewer another chance to leave. But the drop-off sits at the front. Most of the people you lose go in the first few seconds, before length matters at all.
One more weight sits on the scale, and it is money rather than distribution. The Creator Rewards Program only pays on videos longer than a minute. It also wants 10,000 followers, 100,000 views in the previous 30 days, an account in good standing, and an age of 18 or over. We go through that in how many likes it takes to get paid. If monetisation is the point, that settles the length question for you. Whatever completion rate you get at a minute is simply the rate you get.
The practical read: let the material set the length. A joke with one beat is fifteen seconds, and forcing it to sixty costs you both numbers at once. An explanation with four steps is a minute. Cut it to fifteen seconds and you do not make it perform better, you make it incomprehensible. Viewers swipe those away early, which shows up as a worse completion rate anyway.
Read your own numbers instead
Open TikTok Studio, go to Analytics and then Content, and tap a post. You get its watch time and completion rate alongside likes, saves and new followers. The retention graph on an individual post is the most useful screen in there, and almost nobody opens it.
It shows you where people left. That is far more actionable than knowing what share of them reached the end. A cliff in the first two seconds is a hook problem, not a length problem. We wrote up that mechanism in how to write a TikTok hook. Most videos show a slow, even decline, which is normal. A cliff in the middle is an editing problem at a timestamp you can go and watch.
Compare against your own back catalogue rather than any published figure. Your last twenty videos share your audience, your niche and your format, which no benchmark does. Read the trend across ten posts rather than the spike on one. Our piece on TikTok engagement rate covers how to do that without over-reacting to a single bad week. And how to get more views covers the distribution side underneath all of this.
One comparison settles the argument for your account rather than in general, so make it deliberately. Take your ten best-performing videos by views. Write down the length of each. If they cluster at one end, the material you are good at has a natural length. You just found it without anyone’s rule.
What this means for the first three seconds
Everything above makes the opening more important, not less. The reason is worth separating from the usual advice.
The people who leave in the first two seconds do not leave because the video was too long. They never found out how long it was. They left on the frame, the first line, or the sense that this was not for them. That loss is identical on a ten-second video and a ninety-second one. It is not a length problem, and trimming cannot fix it.
Which reframes what trimming is for. Cutting a video from 40 seconds to 25 does not rescue the viewers who left at second two. It only changes what happens to the ones who stayed, and it lowers the watch time they can possibly give you. Cut for pace, cut because a section is dead, cut because the thing is finished and you are padding. Do not cut to move a percentage.
Where we stand
We sell views, including on buy TikTok views, so weigh this accordingly.
Bought views do not carry watch time in any meaningful sense. We would rather say that plainly than let you infer it. This whole article describes what real viewers do with the seconds after they arrive. Whether they stay. Where they drop. How much of the thing they consume. A delivered view is a number moving. It is not a person deciding to keep watching, and that decision is precisely the signal TikTok weighs most heavily.
Maybe a post looks unwatched and the count puts people off. That is a real cosmetic reason, and the service does that job. We describe the same effect for likes in how to get more likes. But maybe you read that views will improve your retention metrics or push the video further. That is the one thing they cannot do, and anyone telling you otherwise is selling harder than we are.
The same caution applies as everywhere else. TikTok’s Community Guidelines say it removes inauthentic metrics when it detects them. The scale is not theoretical: in 2024 it says it prevented over 54 billion fake likes and removed a further 8 billion. Nobody selling this can call it risk-free, us included. We set out what actually happens in the risks of fake engagement.
Frequently asked questions
What is a good completion rate on TikTok?
No published benchmark exists, and no credible one can. Completion rate stays private to each account and invisible from outside. Any specific figure comes from one agency’s own clients, or from nowhere. Compare a video against your own recent posts of a similar length instead.
Is watch time more important than completion rate?
For distribution, TikTok has actually described watch time as carrying weight. Its own example is a viewer finishing a longer video. Completion rate works as a diagnostic. It tells you whether the video held the people it got. But it is not a target, because shorter videos improve it on their own.
Should TikToks be under 15 seconds?
Only if the material is fifteen seconds long. The rule is a leftover from when the platform capped uploads at a minute. Longer videos have a higher watch-time ceiling and a higher risk of losing people. The Creator Rewards Program also pays nothing on anything under a minute.
Does rewatching a TikTok count as extra watch time?
Yes, and on short videos this matters. A clip that loops while the viewer stays on it accumulates watch time. So average watch time on a very short video can exceed the video’s own length. That is also why completion figures on short and long videos do not compare directly.
Why did my completion rate drop when my views went up?
This is normal and usually a good sign. When a video travels beyond your followers, it reaches people with no prior interest in you. They leave earlier than your regulars do. The rate falls because the audience widened, not because the video got worse.
Can you buy watch time on TikTok?
No, not in any way that reproduces the signal. Services sell views, and a view is a count. The mechanism that matters is a real person choosing to keep watching, second by second. Nobody can deliver that to a video from outside.
The short version
Completion rate and watch time are different numbers. TikTok’s own documentation describes only one of them as carrying weight. Its example is a viewer finishing a longer video, so the length is part of the signal rather than incidental to it.
A 60-second video at 30% completion delivers more watch time per view than a 10-second video at 100%. Chasing the percentage means optimising a ratio whose denominator you control. You can always improve it by making each view worth less.
No published completion benchmark is credible. The metric stays private to each account, and nobody outside can observe it. That is why every figure in circulation is unsourced. It is also why this article carries no measured sample: the data does not exist to measure.
Let the material decide the length. Use the retention graph to find where people actually leave. Fix that timestamp. Compare against your own back catalogue rather than anyone’s rule of thumb.
