Outliers are the only reliable signal of what an audience actually wants. The maths takes a minute; the judgement calls are what most people get wrong.
An outlier is a video that got far more views than normal for the channel that published it. To find one, take a channel’s last 10–20 uploads, find the median view count, then divide each video’s views by that median. A video at 3× the median or higher is an outlier worth studying; 5×+ is a genuine breakout.
The comparison must always be against that channel’s own recent videos — never against subscriber count, and never against other channels. 100,000 views is a disaster for one channel and the best week of the year for another.
No tools needed. This works on any public channel:
Worked example: if the last ten uploads have a median of 20,000 views and one video did 200,000, that video is a 10× outlier. The other nine are the control group — that is what makes the comparison mean something.
This is the single most common mistake. One viral video drags an average upward so far that every other video looks like a failure against it — including the videos that were doing perfectly well.
Take view counts of 1k, 2k, 2k, 3k and 500k. The average is about 102,000, so the 2k videos score 0.02× and look catastrophic. The median is 2,000, which correctly says four videos performed normally and one exploded. The median ignores the extreme; the average is ruled by it.
Subscribers are a historical total, not a measure of reach. A 2-million-subscriber channel whose videos now get 30,000 views is in decline; a 3,000-subscriber channel getting 30,000 views is being pushed hard by the algorithm. The same view count, opposite meanings. Only the channel’s own recent videos give you the baseline.
Shorts routinely pull ten to a hundred times the views of long-form on the same channel, because they are served in a different feed to a different audience. Put them in one list and the median becomes meaningless — the Shorts inflate it and every long-form video reads as a flop. Score Shorts against Shorts and long-form against long-form, always.
A day-old video is still being distributed. Compared against videos that have had months to collect views, it will always look like a failure, and a multiplier calculated in that window tells you nothing. Either wait, or compare it against what the channel’s previous uploads had done at the same age — the only fair comparison that early.
If a channel was dormant for a year and has been uploading weekly since March, its lifetime numbers describe a channel that no longer exists. Keep the baseline recent — the last 10–20 uploads, or the last 90 days.
Finding one is the easy half. The value is in the pattern, so for each outlier ask:
One outlier is an anecdote. Five outliers across several channels in the same niche, sharing a trait, is a pattern you can build on.
The arithmetic above is exactly what VideoCore Lens does automatically. It is a free Chrome extension: install it and every video thumbnail on YouTube gets a coloured frame and its multiplier, calculated against that channel’s own comparable recent uploads — Shorts against Shorts, long-form against long-form, with videos under 48 hours old marked NEW rather than given a number they have not earned.
On a channel page you also get filters for 2×+ and 5×+ so the outliers are the only thing left on screen. There is no credit limit and no account needed to see the scores.
If you would rather not install anything, the free Outlier Finder does the same job on the web for any channel or topic.
VideoCore Lens scores each video against its own channel’s baseline while you browse YouTube normally. No credit wall, no account needed for the scores.
Get VideoCore Lens — free Or use the web Outlier FinderAn outlier is a video that performed far better than normal for the channel that published it. It is measured as a multiplier: the video's views divided by the median views of that channel's recent uploads. A video at 3x or more of its channel median is generally treated as an outlier, and 5x or more as a breakout.
Take the view counts of the channel's last 10 to 20 uploads, sort them and find the median (the middle value). Divide the video's views by that median. If the median is 20,000 and the video has 200,000 views, the multiplier is 10x.
A single viral video distorts an average so severely that normal videos appear to fail. With view counts of 1k, 2k, 2k, 3k and 500k, the average is about 102,000 while the median is 2,000. The median correctly describes typical performance because it ignores the extreme value rather than being dominated by it.
No. Subscriber count is a historical total and does not reflect current reach. A channel with two million subscribers averaging 30,000 views is declining, while a channel with 3,000 subscribers hitting 30,000 views is being promoted by the algorithm. Always compare a video against the recent uploads of the same channel.
No. Shorts are distributed in a separate feed and commonly get many times the views of long-form videos on the same channel, so mixing them makes the median meaningless and every long-form video look like a failure. Score Shorts against other Shorts and long-form against other long-form.
Wait at least 48 hours. Before that the video is still being distributed, and comparing it against videos that have had months to accumulate views is not a fair test. The only sound early comparison is against what the same channel's previous uploads had done at exactly the same age.
Yes. VideoCore Lens is a free Chrome extension that shows each video's multiplier against its own channel's recent comparable uploads directly on YouTube, including filters for 2x and 5x outliers on channel pages. VideoCore also has a free web-based Outlier Finder that needs no installation.