Every channel has one video carrying the rest
Open any competitor with 100 or more uploads and sort by most popular. The shape is almost always the same: one video far above everything, two or three well above the pack, and then a long flat plain of videos that all did roughly the same modest number. It is not unusual for the top video alone to hold 30% to 60% of the channel's lifetime views.
That distribution is the most useful thing on a competitor's channel and almost nobody uses it. Most competitor research consists of watching the newest uploads, because that is what the channel page shows you first. The newest uploads are the least informative videos on the channel: they have not been tested long enough for the algorithm to decide anything about them.
The anchor is different. It is a video the audience voted on, at scale, over months. It survived. Whatever question it answered was a question a large number of people were actually asking, and that question is very unlikely to have stopped existing. That is the asset you are looking for, and it is public.
- The top video often holds 30% to 60% of a channel's lifetime views.
- The newest uploads teach you the least: they have not been tested yet.
- The anchor is a question the audience already voted on, at scale.
- All of this data is public. You need no tool to start.
Finding the anchor in 10 minutes: use the median, never the average
Take the channel's last 30 uploads and write down their view counts. Now find the median, the middle value when you line them up in order. Do not use the average: the anchor itself will drag the average so far up that half the channel will look like it underperformed, which tells you nothing.
Then divide each video's views by the median. That number is the multiplier, and it is the only ranking that matters. A video with 400,000 views on a channel whose median is 40,000 has a multiplier of 10. The same 400,000 views on a channel with a median of 350,000 is an ordinary upload that happened to do slightly better than usual.
One correction that changes the answer more than people expect: normalise for age. A video published 3 years ago had 3 years to accumulate. Divide views by days online to get views per day, and re rank. Anchors that look enormous often turn out to be old and slow, while a video from 4 months ago sitting at half the views is doing three times the daily rate. The recent, fast one is the one worth modelling, because it is winning under the current algorithm rather than the one from three years ago.
- Take the last 30 uploads. Find the median view count.
- Multiplier = video views divided by the median.
- Never use the average: the anchor destroys it.
- Normalise by views per day, not raw views, or old videos win by default.
- Re rank on views per day. The winner often changes.
Reading the multiplier: 3 times is a signal, 10 times is an anchor
A multiplier under 2 is noise. Videos vary, thumbnails vary, publishing days vary, and half the channel will land between 0.7 and 1.5 times the median for no interesting reason at all. Ignore that entire band.
Between 3 and 5 times the median is a signal worth reading. Something in that video connected: usually the topic, occasionally the title framing. If you find three or four videos in that band and they share a theme, you have found a vein rather than a fluke, and a vein is more valuable than a single anchor because it is repeatable.
Above 10 times, you are looking at an anchor, and anchors need to be interpreted carefully. Some are genuine: the video answered a question with real, sustained demand. Some are accidents: a news cycle, an external link, a celebrity mention, a moment. The test is the shape of its traffic over time. A genuine anchor keeps pulling views months later, because search and suggested keep feeding it. An accident spikes and dies inside two weeks, and modelling it gives you a video about something nobody is looking for anymore.
- Under 2 times: noise. Ignore the whole band.
- 3 to 5 times: a real signal. Three of them sharing a theme is a vein.
- Over 10 times: an anchor. Check whether it is genuine or an accident.
- Genuine anchors keep pulling months later. Accidents die in two weeks.
What the anchor teaches, and what it absolutely does not
It teaches you the question. That is the whole prize. Underneath a title like the seven mistakes that kill a new channel there is a question a large number of people typed or half thought, and the video happened to be the thing that met them. The question survives the video. It will still be there next year with a different wrapper on it.
It does not teach you the title. Copying the title format is the most common mistake and the least effective, because the exact phrasing that won is already occupied by a video with 400,000 views and years of ranking signals. You are entering a fight you cannot win against a video that is better positioned than yours will ever be.
It also does not teach you the thumbnail. A thumbnail that worked for a channel with an established audience works partly because that audience recognises the channel. On your channel, the same design is just a stranger's face in a feed.
The move that works is to take the question and change the angle. If the anchor answers what mistakes kill a new channel, your video answers the same question for a specific niche, or with a number, or from the other side, the mistakes that did not matter. Same demand, different door, no head to head fight.
- Copy the question. It is the durable asset.
- Do not copy the title: you lose to the incumbent by design.
- Do not copy the thumbnail: it works partly because of their audience, not yours.
- Change the angle: a niche, a number, or the opposite side of the same question.
The trap: modelling an anchor from the wrong kind of channel
A 10 times multiplier on a channel with 5 million subscribers and a 10 times multiplier on a channel with 8,000 subscribers are not the same discovery. The large channel has a subscriber base that guarantees an initial audience, so its anchor may have won on distribution rather than on the question. The small channel had no distribution to lean on, which means the video had to earn every single view from strangers.
For that reason, the most valuable anchors to study are on small and recent channels. A video doing 10 times the median on a channel with 8,000 subscribers is a nearly pure signal about the topic, because nothing else could have carried it. That is also the channel nobody is analysing, because everyone is busy studying the giants.
The second version of the same trap is age. A channel that has been running for six years accumulated authority in its niche, and its anchor benefits from that. A channel that started nine months ago and already has an anchor is showing you something that works right now, under the current algorithm, with no history to lean on. That is the closest thing to a controlled experiment you will find for free.
This is also where a proper research pass beats browsing. The FalconVid Spy exists to make exactly this sortable: find the channels that are winning without the advantages, look at what their audience actually voted for, and start from what already monetises instead of from a blank page.
- Anchor on a huge channel: possibly distribution, not the question.
- Anchor on a small channel: nearly pure signal about the topic.
- Anchor on a young channel: proof it works under the current algorithm.
- The best anchors are on the channels nobody is studying.
Modelling instead of copying: what FalconVid does with the anchor
Finding the anchor is 10 minutes. Turning it into a channel is where the work used to be, and that is the part the platform removes. You do not start from zero, you start from what already monetises: the Spy and the FalconVid Selection surface the channels and videos worth modelling, and the Chrome extension puts the numbers on the channel page while you browse YouTube normally.
One honest note on the numbers you see there: revenue figures shown by the extension are FalconVid estimates built from public data and RPM benchmarks. They are not YouTube data, not Google data, and carry no approval from either. Use them to rank opportunities, never as a statement of what a channel actually earns.
From the modelled channel, you approve a calendar once and the pipeline does the rest: the AI specialists work in parallel, a researcher, a scriptwriter, a narrator, an editor and sound design at the same time, and the video is ready in up to 30 minutes, published to YouTube, Instagram, TikTok, Rumble and Facebook in up to 63 languages, with the video SEO already assembled. Channel DNA keeps the identity persistent so the tenth video still sounds like the first.
The reason this matters for anchors specifically: one anchor is not a channel, it is one video. What you actually want is to run the same question through 6 or 8 different angles and let the audience pick, which by hand is two months of production. With generations running in parallel, 2 at a time on Starter up to 50 on Scale, testing the vein is a week rather than a quarter.
- Spy and FalconVid Selection to find what already monetises.
- Chrome extension puts the numbers on the channel page as you browse.
- Approve the calendar once, video ready in up to 30 minutes, 5 networks, up to 63 languages.
- Parallel generations turn testing a vein into a week instead of a quarter.
The protocol: 10 channels, 10 anchors, 3 topics
Pick 10 channels in your niche, and deliberately skew the list: 3 large ones for the map, 7 with under 50,000 subscribers or under 18 months old, because that is where the clean signal lives. For each one, take the last 30 uploads, compute the median, compute the multiplier, then re rank on views per day.
Write down every video above 3 times the median. You will end up with somewhere between 15 and 40 videos. Now group them by the question they answer, not by their title. This step is the whole exercise, and it is the one people skip. When you group by question, the same demand tends to show up on 4 or 5 different channels wearing different clothes, and that repetition is the confirmation you were looking for.
Pick the 3 questions that appear most across channels and are not dominated by a single enormous incumbent video. Those are your next three topics. Give each one an angle that is not a head to head fight: a niche cut, a specific number, or the opposite side of the argument.
Then repeat monthly, not weekly. Anchors do not change fast, and running this too often turns into research as procrastination. Once a month, 10 channels, three topics out, is enough to keep a calendar fed with questions the audience has already voted for.
Three topics a month is only useful if three topics a month actually get published, and that is where most research dies. In FalconVid the three questions go straight into the calendar you approve once, the pipeline writes, narrates, edits, captions and publishes them, and a 12 minute video costs 1,008 credits in economy mode, 8,676 in balanced and 26,760 in premium, so a month of modelled topics fits inside a Starter plan when you mix the modes.
- 10 channels: 3 large for the map, 7 small or young for the clean signal.
- Median, then multiplier, then re rank on views per day.
- List everything above 3 times, then group by question rather than title.
- Pick 3 questions that repeat across channels and lack a dominant incumbent.
- Repeat monthly. Weekly is procrastination.