How We Measure Short-Form Video Performance

Median, not mean

Most published benchmarks report the mean (average), which a handful of viral posts can inflate dramatically — making the "typical" result look far better than it is for the creator in the middle of the distribution. We report the median: the value at which half the posts perform above and half below. It answers the question creators actually ask — "what can I realistically expect?" — and it is the core reason our numbers differ from figures published by Sprout Social, Buffer, or Hootsuite.

Viral outliers are not removed from the dataset. They exist in the distribution; the median simply insulates the reported figure from being dominated by them.

Matured posts only

A post's view and engagement counts keep climbing for days after publishing. Measuring too early understates settled performance and makes comparisons across posts published at different times misleading. We only include posts that have been live for at least 14 days, so every data point reflects a comparable, finished result.

Dataset scope and coverage

The dataset spans ~508,000 posts from 1,900+ creators across 80+ countries and five platforms over a 12-month window. Coverage is not uniform: the TikTok and Instagram slices are the largest; the LinkedIn and Facebook slices are smaller. Where a finding relies on a platform with limited coverage, we note the sample size in-line.

Retention analytics are available only for the subset of posts where creators granted full analytics access. We report those findings against that subset explicitly rather than projecting them onto the full dataset. For timing analyses, we lead with the US-creator subset where timezone data is cleanest and present global figures as secondary, with the caveat noted.

What we do not do

• We do not average across platforms. TikTok and Instagram are reported separately. A "social media average" would obscure more than it reveals.
• We do not mix matured and fresh posts. All figures use the ≥14-day filter consistently.
• We do not claim causation. Our data shows correlations — for example, posts with zero hashtags earning more median views than posts with seven to ten. We do not claim hashtags cause lower views; other variables likely travel with that pattern.
• We do not suppress thin-sample findings. If a finding is based on a smaller slice, we say so, rather than presenting it with the same confidence as the full-dataset figures.

How to cite this research

You are welcome to cite quso.ai Research findings with attribution. The standard citation format is: quso.ai Research, [report title], quso.ai/research/[slug], [year]. The dataset is published under a CC BY 4.0 license — you may use and adapt the findings with attribution.