It’s one of the most searched questions about analytics: what’s a good retention rate? People want a number. 50%? 70%? Something to aim for.
The honest answer is that there isn’t one number, and anyone who gives you a universal benchmark is guessing. But there is a reliable way to tell whether your retention is good, and it’s more useful than any benchmark. Here’s how.
What “retention” actually means
Apps use a few related numbers, and it helps to know which is which:
| Number | What it means | The catch |
|---|---|---|
| Average percentage viewed | What share of the video the average viewer saw | Short videos score higher naturally; loops can push it over 100% |
| Average watch time (average view duration) | How long the average viewer watched, in seconds or minutes | Longer videos naturally score higher |
| Retention graph | The share still watching at each second | Shows where people left, not why |
| Viewed vs. swiped away | How many stopped to watch in a short-video feed | Only measures the opening |
| Completion rate | How many reached the end | Short videos finish more easily |
When people say “retention rate,” they usually mean average percentage viewed. But on its own, that number can mislead you badly.
Why there’s no universal “good” number
Length changes everything. A 10-second clip that 95% of people finished didn’t necessarily beat a 90-second tutorial people watched halfway through. The longer video might have held far more total attention. Percentages always flatter short videos; totals always flatter long ones.
The kind of video changes everything. A joke, a tutorial, a story and an eight-minute review all hold people differently.
Where viewers came from changes everything. People from search often skip straight to the answer and leave satisfied. People from a recommendation feed came to be entertained. The same video can show very different retention for each group.
Apps count differently, and change. YouTube, for example, now counts a Shorts view whenever a Short starts or replays, which shifts every percentage built on views.
So a number that’s excellent for one video can be weak for another. That’s why the useful question isn’t “is this good?” but “is this good compared with what?”
The comparison that actually works
Judge each video against your own typical video of the same kind:
- Same format. Tutorials against tutorials, skits against skits.
- Same length band. Under 30 seconds, 30 to 90 seconds, and longer, compared separately.
- Same traffic source, where your app lets you filter.
Then use your median, not your average or your best. Line up your last 10 to 20 videos of that kind from worst to best and take the middle one. It ignores one lucky hit and one disaster, so it tells you what’s normal for you.
A video that beats your median for its kind is a good result. A run of videos that beats last month’s median means you’re genuinely improving, whatever the internet says the benchmark is.
The numbers that matter most, by kind of video
Short videos (TikTok, Reels, Shorts)
- How many stay past the first two or three seconds (or viewed vs. swiped away on Shorts). This is the score for your opening.
- Average percentage viewed, compared within the same length band.
- Rewatches. Above 100% average viewed on a short means people looped it, which is a strong sign.
Long videos (YouTube)
- The share still watching after 30 seconds. YouTube shows this as part of its retention analytics, and suggests checking whether your opening meets the expectations your title and thumbnail set.
- Average view duration in minutes, compared with videos of a similar length.
- The share still there just before the payoff. A longer video that holds more people to its payoff can be the stronger one, even with a lower overall percentage.
Read the shape, not just the number
Two videos can have the same average and completely different stories. Open the retention graph and look at the shape:
- A steep drop in the first seconds: the opening didn’t grab, or didn’t match the cover.
- A gentle, steady slope: normal.
- A sudden dip in the middle: a slow section, a repeat or a confusing step.
- A bump: people rewatching a moment, or replaying something confusing.
- A big drop near the end: people left once they got the payoff, which usually means the ending ran long.
There’s a full guide to reading a retention graph, and the Analytics Decoder shows where each app keeps these numbers and what they really measure.
What to aim for instead of a benchmark
Pick one number that matches what you’re working on, and track it across your next ten videos:
| If you’re working on | Track |
|---|---|
| Better openings | The share still watching after 3 seconds (or viewed vs. swiped away) |
| Better videos overall | Average percentage viewed, within one length band |
| Long-form structure | The share still watching after 30 seconds and before the payoff |
| Growing your audience | Follows per view, alongside retention |
Then change one thing at a time. If you change the hook, the length and the topic at once, you’ll never know which helped.
Be careful with small numbers
Retention on a video that 150 people saw jumps around a lot. A few viewers either way can swing it by several points. Before you decide a change worked, look at several videos, not one. And if a report shows “not enough data,” the app is waiting for more views; it isn’t a verdict on the video.
Try it
Pick your most common kind of video. Write down the average percentage viewed and the share still watching after three seconds for your last ten of them, and find the median of each. That’s your benchmark. Make your next video aim to beat it on one of the two, and only change the thing that affects that number.
If you’re not sure which part of your videos is holding them back, the Weakest Step Finder points to the one step to fix first.
