Somewhere in your channel's public data (the videos you've already posted, the comments people already left, the searches YouTube already suggests) there's a working answer to "what do I make next?" Most creators never look.

Here's the gap that got us looking.

In the 100 MrBeast videos we logged frame by frame, one hook ingredient, a promise of what's about to happen, shows up in 94 of them. Another, danger, shows up in 22. Same channel, same audience, same hundred uploads.

A 72-point gap like that isn't random. It's a pattern that held across a hundred videos in a row. And if a gap that size sits inside one channel's own history, in plain view, there's one inside yours. You don't need his budget to read it. You need the method, pointed at your own data.

Source: the MrBeast library, 100 long-form videos logged for hook type, hook mechanism and retention devices. Not our channel's numbers.

Open your upload history and compare each video's views with your own channel's typical video, not anyone else's. The ones that clear that bar are your local outliers, and outliers repeat for a reason.

It's the same comparison we ran on the MrBeast library, at a bigger scale: count everything, then look at what repeats. It wasn't the premises that repeated. It was the pattern underneath them. Those numbers only mean something because every video was counted, not the ones that felt biggest in memory.

Do the same with your last ten or twenty uploads: which beat your own typical video, and what do they share that the rest don't?

Open the comments on your outliers, and on the videos just below them, and read for repetition, not compliments.

"More of this" once is an opinion. The same request in five people's own words is an echo: something your channel is already asking for that isn't a video yet.

Sort by top comments, read the first fifty on each outlier, and write down anything that shows up more than twice: a request, a question, a complaint.

Honest note: whether repeated comments on a channel's outliers line up with what makes them outliers has not been tested in our data. It's a hypothesis to run, which is why it's part of the test below.

Type the topic of your best outlier into YouTube's search box and stop before you hit enter. What YouTube suggests is built from what people already search for, not what you assume they search for.

If your outlier's topic and a suggestion overlap, you're no longer guessing at demand. You're pointing at it. Suggestions vary by viewer history and country, so check from a private window too.

Honest note: whether an autocomplete match predicts how a future video does has not been tested in our data either.

One signal alone is weak evidence. An outlier can be a fluke. A repeated comment can be five people who talk alike. An autocomplete match can be a coincidence. When two of the three point at the same idea, you're reading a pattern that showed up in independent places.

Direct response works the same way. In 7 tested Russell Brunson sales videos we measured scene by scene, the proof section is the least talking-head part of the video: only 35% presenter, the rest evidence on screen. The presenter doesn't argue you into the claim; the evidence does. Reading your own outliers, echoes and autocomplete is the same move: let the evidence argue for your next video.

The 94% vs 22% gap comes from 100 videos on one very large channel. The comment and autocomplete signals haven't been run through our datasets at all: they're the method we propose, not a finding we've confirmed. That's exactly why the test below runs on your channel, not ours.

Finding the right idea is half the job. The other half is a title and thumbnail that make a stranger want it in about a second. In our count of 7,646 videos, the idea and whether the package made a stranger want it explain far more than any word trick: all 27 kinds of title wording we counted, together, explain about 2 of every 100 slices of the difference inside a channel.

That's what Make Them Watch teaches: how to turn the idea your data found into a package a stranger can't scroll past. Six checks you can score (Seen, For me, Payoff, Open question, Easy, Believed), the routine of twenty titles and three thumbnails, and the first ten seconds written before you film. Chapter 11 starts exactly where this article ends: choosing the idea first.

Part 1 is already in your inbox, free. The whole book is {PRICE: book}, with a 30-day refund and no questions. At checkout you can add 50 Tested Ways to Win the Click for {BUMP}.

[Get Make Them Watch →]({LINK: art20-book})

Xander

P.S. Next video: a real funnel built on reading exactly this kind of signal, the Mini Katana channel teardown.