Short-Form Video Statistics 2026: What 4,404 Real Posts Show

Most short-form statistics pages are aggregations: a number from one vendor, a number from another, none of them checkable. This one is different in a specific way. Every figure below is measured from 4,404 real TikTok posts we collected ourselves, and every figure is reported with the sample size behind it so you can judge whether it is worth quoting.

The headline result is one that contradicts almost every "best clip length" guide on the internet, including an earlier version of this page: the optimal clip length is not a constant. It inverts with audience size.

The Five Findings

1. Short clips beat long clips by 7.23x below 1,000 followers, and lose to them by 2.23x above 100,000. The crossover is real and it is monotonic across all four bands.
2. 31 to 60 seconds is the worst-performing length in every audience band we measured, reaching just 0.14x to 0.55x of the best length for that band.
3. Posts with spoken audio out-reach silent posts in all four bands, but the advantage collapses from 3.66x to 1.07x as the account grows.
4. On-screen text looks like a 1.50x advantage until you control for follower count, and then it disappears. It is 5.89x below 1,000 followers and 0.94x to 1.01x above. The pooled effect was account size, not text.
5. Save rate peaks in the middle of the reach distribution and falls 61% at the top, while like rate keeps climbing. Reach dilutes intent.

1. The Best Clip Length Inverts With Audience Size

This is the finding we would most like other people to check. We split every post by the creator's follower count at capture, then by duration, and took the median play count in each cell. Cells under 40 posts are suppressed rather than published with an asterisk.

Median plays. Sample sizes per cell, left to right: under 1k = 276/153/206/133/59; 1k-10k = 394/163/291/265/158; 10k-100k = 268/146/309/384/243; 100k+ = 105/84/159/294/175.

Stated as a single ratio, short (0-15s) against long (61-120s):

The practical reading: advice to "keep it under 15 seconds" is correct for a new account and actively wrong for an established one. A creator at 200,000 followers who cuts everything to 15 seconds is optimising for a constraint they no longer have. We broke this out further in the full breakdown of best TikTok length by follower count, where the same crossover shows up in saves as well as plays.

How this compares to the largest published study

We are not the first to segment TikTok length by account size, and it would be wrong to imply otherwise. Metricool's 2024 benchmark study covered 1.12 million videos across 91,840 accounts — roughly 250 times our sample — and reported an ideal length that rises with audience: 2.6 minutes for accounts under 500 followers, up to 7.48 minutes at 10,001–50,000. Our result agrees with theirs at the top of the range and flatly contradicts it at the bottom.

The most likely reason is the summary statistic. The published Metricool figures are reported as average views; every figure on this page is a median. On a distribution as skewed as short-form reach, that is not a cosmetic difference. Our own corpus contains a single post with 42.2 million plays, and long videos are the rarer category, so a handful of viral long posts moves a mean far more than it moves a median. We think that is what produces a multi-minute "ideal length" for accounts with under 500 followers, which does not match what small accounts actually experience.

We are not claiming Metricool is wrong. We are claiming the two methods answer different questions, and that for a creator deciding how to cut their next clip, the typical outcome (median) is the more useful one. If you are citing either result, cite which statistic it used.

2. The 31-60 Second Dead Zone

Across all four bands, the 31 to 60 second range underperforms the best length available to that band. It is not a small gap.

This band is where a lot of repurposed content naturally lands, because a good talking point often runs about 45 seconds. If you are cutting a podcast or a stream, the choice is usually to tighten below 30 or let it breathe past 60. Our video length calculator gives the per-platform limits if you are working to a hard ceiling.

3. Spoken Audio Beats Silence in Every Band

Posts flagged as containing speech out-reach silent posts in all four bands, which is one of the few effects in this dataset that survives stratification unchanged in direction. What does change is the size of the advantage.

A 3.66x advantage on a small account and a 1.07x advantage on a large one are different claims. If you are starting out, talking over your footage is close to the highest-leverage change available. If you already have reach, it is a rounding error.

4. On-Screen Text: The Effect That Died Under a Control

This is the finding we are most confident other people have published incorrectly, because we nearly did too.

Pooled across the whole corpus, posts with on-screen text reach 1.50x the median plays of posts without it. That is a clean, quotable, marketable number. It is also an artifact. Creators who use on-screen text in this dataset have a median 11,600 followers against 8,444 for those who do not. They were already bigger.

Once you hold audience size fixed, the advantage is 5.89x for accounts under 1,000 followers and gone above that. The honest version of the advice is narrow: on-screen text is a strong lever while you are small, and roughly neutral once you are not. Any source quoting a single pooled multiplier for on-screen text is quoting selection, not causation.

5. Reach Dilutes Intent: Save Rate Peaks and Falls

Sorting every post by play count and splitting into ten equal deciles shows two metrics moving in opposite directions.

426 posts per decile (431 in the tenth). Deciles 2, 4 and 8 omitted from the table for width; the pattern is continuous.

Like rate climbs from 3.29% to 8.14% as reach grows. Save rate does the opposite after decile 6, falling 61.5% from its 1.35% peak to 0.52%. Share rate peaks one decile later and falls 36.7%. The interpretation we would defend: a post that reaches far reaches people with weaker intent, who will tap like but not save. If your objective is saves, the highest-reach post is not the target. We took this further in the saves versus shares analysis, which found 57.6% of the most-shared posts are not among the most-saved.

What This Changes for Creators

Three of the five findings above are stated backwards in the guides currently ranking for this topic, including the version of this page we published in April. Short clips are not universally better. On-screen text is not a universal multiplier. High reach is not a proxy for high intent.

The common thread is that pooled numbers hide the variable that actually drives the result. Almost every effect we tested moved when we split it by audience size, and two of them reversed. If a statistic about short-form video does not say what it controlled for, it is probably measuring account size wearing a different hat.