A Percentage Is Half a Sentence
Retention is reported as a percentage, and a percentage on its own is an incomplete statement. Fifty percent of what, exactly? The number describes a proportion of a runtime, and until you know the runtime you know nothing about how long anybody watched.
This is not a pedantic point. It is the single most common reason creators reach wrong conclusions from their analytics. Two videos, one at 65% and one at 40%, look like a clear winner and a clear loser. If the first was 12 seconds and the second was 75, the first held the average viewer for under 8 seconds and the second held them for 30. The ranking inverts the moment you do the arithmetic.
So the first habit worth building is converting before comparing. It takes a second and it changes what your analytics are telling you more often than you would expect.
The Conversion, for the Lengths You Actually Post
The arithmetic is retention multiplied by runtime. The table below does it for the retention figures that come up most and the lengths short-form video actually lives at, so you can read across rather than reach for a calculator.
The pattern in that table is worth sitting with. Moving from 30% to 40% on a 15 second video buys you a second and a half. The same ten points on a 60 second video buys you six seconds. Percentage improvements are not equally valuable, and they are worth the most exactly where they are hardest to achieve.
| Video length | 30% retention | 50% retention | 70% retention |
|---|---|---|---|
| 10 seconds | 3 seconds | 5 seconds | 7 seconds |
| 15 seconds | 4.5 seconds | 7.5 seconds | 10.5 seconds |
| 30 seconds | 9 seconds | 15 seconds | 21 seconds |
| 45 seconds | 13.5 seconds | 22.5 seconds | 31.5 seconds |
| 60 seconds | 18 seconds | 30 seconds | 42 seconds |
| 90 seconds | 27 seconds | 45 seconds | 63 seconds |
| 3 minutes | 54 seconds | 90 seconds | 126 seconds |
Why the Same Percentage Is a Different Achievement
Holding 70% of an audience through 10 seconds asks one thing of a video: do not waste the opening. There is barely time for a middle, so the structure that wins is a single clear idea delivered immediately.
Holding 70% through 90 seconds asks something else entirely. It requires a reason to keep watching at the 20 second mark, at 45, and at 70, each of them arriving before the previous one has worn off. That is a much harder problem, and it is why long-form short video rewards structure in a way a 10 second clip does not.
Judging both against the same target treats those as equivalent achievements, which is why retention targets worth using are always banded by length. We publish ours that way for TikTok and for Shorts, and the bands fall as runtime rises for exactly this reason.
The Number the Conversion Cannot Give You
Converting to seconds tells you how much attention you held. It still will not tell you where you lost the rest, and that is the part that changes what you do next.
An average is a single number standing in for thousands of different journeys through your video. Two videos with identical retention percentages can have completely different shapes: one bleeding viewers steadily throughout, the other holding almost everyone until a specific moment where they all leave at once. Those need opposite fixes. The first is a pacing problem across the whole edit. The second is one bad ten-second stretch you could cut this afternoon.
Reading the curve rather than the average is what separates the two, and it is why we walk through how to read a retention curve as its own subject. Retensis goes one step further and lines the curve up against what was actually on screen at each moment, so a drop at seven seconds comes with the reason it happened rather than only the timestamp.
Frequently asked questions
Fifteen seconds. The arithmetic is simply the percentage multiplied by the runtime, which is the whole point: retention is a ratio, and a ratio tells you nothing about duration until you supply the length it was measured against. The same 50% is 7.5 seconds on a 15 second video and 30 seconds on a minute-long one, and those are three very different outcomes wearing the same label.
Because the platforms care about attention held, not about the ratio in isolation. A 90 second video at 35% holds the average viewer for around 31 seconds. A 20 second video at 70% holds them for 14. The second number looks better and the first video kept people roughly twice as long. This is also why comparing your retention percentage against a creator who posts at a different length tells you almost nothing.
Use both, for different questions. The percentage tells you whether the video earned its own runtime, which is a question about editing and structure. The seconds tell you how much attention you actually captured, which is the quantity the platform is weighing against other videos. A video can earn its length and still not hold anyone for very long, and that is worth knowing rather than hiding behind a healthy-looking ratio.
It usually improves the percentage and often reduces the seconds, which is why it is such a tempting and such a misleading fix. Trimming a video that dragged is genuinely good. Trimming a video that was working, purely to flatter the ratio, trades away the attention you were successfully holding. Decide by looking at where the curve actually falls rather than by shortening on principle.
Find out exactly where viewers drop off in your videos
Upload any video and see your predicted retention curve in 90 seconds. Spot drop-off points before they cost you views. Free to start.
See your retention curve free →