What's a Good Retention Rate on TikTok? Benchmarks by Length

Two creators open their analytics on the same morning. Both see 55% average retention. One of those videos is 12 seconds long and just underperformed badly. The other is 50 seconds long and is the strongest thing that account has published all quarter.

Same number. Opposite verdicts. And the difference isn’t niche, audience, or luck — it’s duration, which is the single strongest predictor of whether a short-form video gets finished at all.

Here’s the direct answer to what you searched: a good retention rate on TikTok is roughly 60–70% under 15 seconds, 50–60% from 15–30 seconds, 40–50% from 30–60 seconds, and 30–45% past the one-minute mark. A retention percentage is not a quality score. It’s a quality score divided by how long you asked strangers to stay. Compare across lengths and you will consistently reward yourself for making shorter, thinner content and punish yourself for making longer, better content.

The benchmarks, by platform and by length

These are the bands published benchmark sets converge on for 2026. Read your own number against the row that matches your video, not against the platform average.

TikTok

• Under 15s — typical 60–70%, strong above 80%, top-tier above 90%
• 15–30s — typical 50–60%, strong above 70%, top-tier above 80%
• 30–60s — typical 40–50%, strong above 60%, top-tier above 70%
• 1–3 min — typical 30–45%, strong above 45%, top-tier above 60%

Instagram Reels

• Under 15s — typical 55–65%, strong above 70%, top-tier above 85%
• 15–30s — typical 45–55%, strong above 65%, top-tier above 75%
• 30–60s — typical 40–50%, strong above 55%, top-tier above 65%
• 60–90s — typical 30–45%, strong above 50%, top-tier above 60%

YouTube Shorts

• Under 30s — typical 50–65%, strong above 65%, top-tier above 80%
• 30–60s — typical 40–50%, strong above 50%, top-tier above 65%
• Completion above 30% is solid; above 50% is top-tier

Notice what happens across a single column. On TikTok, the gap between “typical” at 12 seconds and “typical” at 90 seconds is roughly 30 percentage points. That entire gap is structural. It exists before anyone judges whether your video was any good.

So the first correction most creators need is a subtraction: stop tracking retention as one line on a chart that mixes lengths. A 12-second clip and a 45-second breakdown belong in separate rows, permanently. This is the same failure that makes global engagement medians useless — we took that apart in Engagement Rate Benchmarks for 2026, and the fix is identical: the comparison, not the number, is what carries the meaning.

Why the ceiling falls as the video grows

Research on short-form viewing behavior keeps landing on the same result: duration predicts completion more powerfully than personalization does. Not the topic, not how well the platform knows the viewer — the length.

The mechanism is unglamorous. Every additional second is another independent opportunity to leave, and the exit options are permanently one thumb-flick away. A 60-second video has to survive roughly five times as many of those moments as a 12-second one. Nothing about the content has to be worse for the percentage to fall.

Which sets up the trade every creator eventually has to make, and it is a genuine trade rather than a trick question:

• Short video, high percentage, small absolute watch time. 12 seconds at 75% is 9 seconds of attention per view.
• Long video, lower percentage, larger absolute watch time. 60 seconds at 35% is 21 seconds of attention per view.

The second video looks worse on the metric and delivers more than twice the attention. Industry analysis of over a million videos found that pieces past 60 seconds earned meaningfully more reach than the 30–60 second band — not despite the lower completion rate, but partly because total watch time is its own currency.

The honest synthesis: percentage tells you whether the video works; seconds-per-view tells you what it’s worth. Track both. A falling percentage with rising seconds-per-view is a creator successfully going longer. A falling percentage with flat seconds-per-view is a creator who padded.

The three places people actually leave

Retention isn’t lost evenly. Across short-form, exits cluster hard at the front — one large study of TikTok viewing found roughly 85% of exits happen inside the first quarter of a video’s duration. Three moments do almost all the damage.

1. The swipe decision — seconds 0 to 3

The steepest loss in the entire curve. Typical first-three-second skip rates on Reels run 20–35%, and above 40% is a flashing warning light. Between a quarter and a third of everyone the platform sent you is gone before your first sentence resolves.

The usual causes are boring and fixable: an intro card, a logo animation, a “hey guys,” a wide establishing shot that delays the subject, or a first frame that is visually indistinguishable from the four videos before it. This is a framing job before it is a writing job — the first frame and the first line do two different jobs, which we separated in Visual vs Verbal Hooks.

One caveat worth internalizing, because it saves you from a lot of bad advice: there is no literal “3-second rule” anywhere in the published research. What’s real is that drop-off is front-loaded. The opening matters because that’s where the mass of exits sits, not because a timer trips at 00:03.

2. The commitment check — seconds 3 to 10

Another 15–25% of the survivors leave here, and this one is almost always a promise problem rather than a hook problem. The opening earned a few seconds of patience by implying something is coming. Second 4 through second 10 is where the viewer decides whether it’s actually coming.

Videos that fail here usually have a hook that was written as a standalone line instead of as the first move in a sequence. The fix is to show visible evidence of the payoff early — the result before the process, the after before the before, the thing you promised partly on screen — so the next eight seconds read as progress toward something rather than setup.

3. The ending — the last 20%

The final drop is the most under-diagnosed, because it doesn’t cost you much watch time and it silently costs you everything downstream. Viewers who leave three seconds before the end never see the moment that would have earned a save, a send, or a follow.

Curves that hold above 70% end-to-end are the ones that produce outsized distribution, and the common structural reason is loop design: the last frame flows into the first, so the video reads as continuous rather than as something that stopped. A trailing “anyway, that’s it, follow for more” does the exact opposite — it announces an exit point and viewers take it.

Reading a low number correctly

A retention percentage below your band is a symptom, not a diagnosis. Three different problems produce it, and they need opposite responses.

• Front-loaded loss with a flat tail. Steep cliff in the first three seconds, then a stable line. Your hook is misfiring, but the content is fine — the people who stayed, stayed. Rewrite the opening; leave the rest alone.
• Steady bleed across the whole video. No cliff, just a constant decline. Pacing problem. Cuts too slow, information density too low, no new visual or verbal information arriving often enough.
• Healthy curve, weak everything else. Retention is fine and nothing downstream fires. That’s not a retention problem at all — that’s the pattern we broke down in Ghost Engagement: Views but No Saves, Shares, or Follows.

There’s also a fourth case that looks like failure and isn’t. If you deliberately filter the wrong audience in second one, your retention percentage should drop while saves and follows climb — you traded volume for fit on purpose. That’s the mechanic behind the disqualifying hook, and it’s the one situation where a lower number is the intended outcome.

Build the benchmark you’ll actually use

Published bands are a starting line. Yours takes about twenty minutes.

1. Bucket your last 20–30 videos by length — under 15s, 15–30s, 30–60s, 60s+. Keep the buckets separate forever.
2. Take the median retention within each bucket, never the mean. One breakout drags an average somewhere useless.
3. Add seconds-per-view to each row (length × retention). This is the column that tells you whether going longer is working.
4. Compare each bucket to its own band above, not to the platform figure and not to your best video.
5. Note the shape, not just the number. Where the line falls matters more than where it ends up.
6. Change one variable at a time. Opening frame, or length, or pacing — not all three, or you’ll learn nothing.

The result you’re looking for isn’t a higher percentage. It’s knowing which of your formats deserves more length and which one is being stretched past what it can hold. Length is downstream of format, and treating it as a platform-wide rule is how good three-beat ideas end up padded to 45 seconds.

Where Blossom fits

Everything above is diagnosable after the fact, which is the problem — analytics only tell you where a video broke once the reach is already spent.

That’s the part Blossom moves earlier. Paste a video URL and you get the hook scored 1–10 with the reasoning behind the score, the format and tactics named, the moments where attention is most likely to break called out with timestamps, and specific changes to make before you publish. Your video is read against a large library of analyzed videos in the same space rather than against a platform-wide average — which matters here more than anywhere, since the whole point is that averages across lengths and niches describe nobody.

One limit stated plainly, because the alternative is marketing: these are content-side estimates. No tool predicts views, and a retention curve you haven’t published yet is a projection, not a measurement. What analysis can tell you is whether the video you’re holding has the properties that tend to hold attention — and where, specifically, it’s likely to lose it.

Run it on your next video, or start with the FAQ if you’re still working out which of the three drop-offs is yours.

The reframe worth keeping: stop asking whether your retention is good. Ask whether it’s good for that length — and whether the seconds you’re winning are worth the seconds you asked for.