How we measure

Where the data comes from

All figures come from the official YouTube Data API: public subscriber counts, view counts, video metadata and the per-country trending charts. No scraping, no estimated or modelled view counts.

Channels we follow are re-measured every two hours. Each measurement is stored, which is what allows us to compute velocity and to compare videos at equal age rather than at equal calendar date.

The outlier score

For each channel we take the recent videos of the same class — long-form and Shorts are never mixed — and compute the median of their view counts. A video's score is its views divided by that median.

We use the median rather than the average deliberately. One past viral hit drags an average upward for years: a channel whose last ten videos did around 11,000 views plus one at 2 million has an average of 210,000, above every normal video, and nothing would ever register as a breakout again. The median stays at 11,000 and the next unusual video shows up immediately.

• ·A score of 1.0 means the video performed like a normal upload for that channel.
• ·A score of 5.0 means five times what that channel usually does.
• ·Thresholds such as 2× or 3× are analyst conventions, not YouTube rules.
• ·A median needs at least 3 comparable videos; below that we compute no score.

Comparing at equal age

A six-month-old video has had six months to accumulate views. Comparing it directly with one published yesterday measures age, not performance.

We therefore freeze view counts at fixed milestones — 1, 3, 7, 14 and 30 days after publication — and compare a video with the median of the channel's other videos at the same age. This score only exists for videos we have tracked since publication: it cannot be reconstructed retroactively, so it is absent on older videos, by design.

How country rankings are built

Each month we pull the channels appearing in a country's YouTube trending feed, then rank those attached to that market by subscriber count.

Attaching a channel to a market is less obvious than it sounds. YouTube's country field is empty for 15 to 30% of channels, so requiring it would remove a third of the ranking, including many genuinely local channels. But treating every undeclared channel as local is also wrong: it put Beyoncé first in Brazil and third in the United States. Our rule: an explicit declaration decides; without one, a channel counts as local unless it appears in several markets at once.

Each monthly edition is frozen in our database rather than recomputed on every visit. That keeps pages fast, makes the sitemap's modification dates honest, and preserves the history needed to show month-over-month movement.

What these numbers do not tell you

Channel-relative performance is an analysis method, not a ranking signal. YouTube has never stated that it recommends videos reaching a given multiple of their channel median. The score is useful because it removes the effect of channel size and reveals an unusual audience response — it does not trigger anything.

We only see public data. Click-through rate, audience retention, watch time and revenue are private to the channel owner, and no third-party tool has access to them. Any tool claiming otherwise is estimating.

Since 24 August 2026, a public view counts from the first frame, including autoplay. View counts are an exposure metric, and figures from before that change are not directly comparable with figures after it.

Corrections

If you find an error — in a ranking, a figure or a method described here — write to hello@videotrends.io. We correct and date the correction.