The Long-Form Playbook

Find videos at 5× or more of their channel's median views — median, not average, from the last 20–30 uploads (1of10 does this automatically). Copy the pattern, not the topic: format + angle + timeframe.

• 1 outlier = luck. 2–3 channels with outliers on the same topic = a real signal.
• Small-channel outliers count most — 200K views on a 20K-sub channel proves the idea, not the audience.
• Window is 2–6 weeks. Be in the first wave, not the third.

No packaging, no filming. Idea → title → thumbnail, locked before a script is written.

• Emotion beats information — in 1of10's 62.6B-view study, titles triggering joy, controversy or anger beat neutral ones. Titles around 30 characters earned ~60% more median views than 70+.
• Thumbnail and title say different halves of one curiosity gap. A third of text thumbnails waste their words repeating the title.
• Calm face, few words — across 11,223 long-form thumbnails, 71% show a face but only 11% are shocked. Median is 4 words; 23% have none.
• Phone test: 2 seconds at phone size — can a stranger say what it's about? Giants test: does it hold its own in a homepage row next to Coffeezilla and How Money Works?
• Make 3 thumbnail concepts per flagship, 2 per roundup. The spare is your 48-hour rescue swap.

YouTube autocomplete for the words viewers actually type. Google Trends for velocity, not volume. News trajectory in hours — a story already peaking is late for long form, so take the second-day angle: the documents, the lawsuit, the money trail.

Every topic is labelled trending (speed + a unique angle) or evergreen (search + suggested shelf life). Roundups ride trending. Investigations must be evergreen or series-able. Then the broad-appeal check: would someone who never watches finance click this? The human story — a scam, a fall, a rigged price — has to be bigger than the niche.

Tear down winners and ordinary videos from the same channels (at least 5 of each) — a pattern only counts if it separates the two.

• First 30 seconds confirm the promise. No logo intro, no preamble. Lose more than ~30% there and the opening failed, whatever happens later.
• Re-hook every 2–3 minutes — a new question, a document reveal, a number that changes the story. Open loops early, pay them late.
• Visual change every 20–30 seconds in talking stretches — clip, document, chart, punch-in.
• Payoffs: first real one early, biggest at 70–80% through — never everything saved for the end.
• Length: 15–25 minutes beats shorter across niches; business/finance runs strongest at 25–40. Length is earned by the story, never padded — % viewed beats total minutes.

Shorts has no equivalent of this. Research which videos YouTube suggests next to your outliers — that panel is your neighbourhood. Your title and thumbnail borrow its language, never copy it.

Plan the chain: what primes this video (the Short, the previous episode), and what the end screen sends them to — ideally yours, same character or theme. A video that starts a session of your videos is worth more than one that ends it.

Every claim in a script traces to a primary source — the filing, the regulator release, the company statement, the person's own post — graded A / B / C like the shorts system. News stories need two outlets plus a primary source. Every story carries a footage flag: real footage, photo + documents only, or text only. Never fake it.

Legal pre-flight: no fraud accusation without a primary document on screen. Alleged / reportedly language. Both sides in, right-to-reply logged. Naming a living person as a fraudster, an ongoing case, or a sponsor conflict triggers an attorney check before filming.

Name the emotion before writing: outrage, awe, schadenfreude, relief, curiosity. Emotion is the strongest predictor in the outlier data — a neutral explainer on a hot topic loses to an emotional story on a smaller one.

The send test, long-form version: who gets sent this, and what does the sender say? What does sharing it say about them — smart, warned, ahead of the news? If there's no answer, label the video a search/suggested play and don't pretend it's a share play.

Count who's covered it and what's missing: the document nobody showed, the victim nobody interviewed, the number nobody did. Your angle is that missing piece, stated in one line. If five big channels already did the obvious version, you need a new angle — or you pass, however hot it is.

For the top 2–3 outliers on your topic: full transcript, pacing, chapter map, where the sponsor sits, and the comments. Top comments show what viewers actually valued. The most-liked objection becomes your both-sides section. The most-asked question becomes your next video.

The only research stage after filming. Fixed reads at 48 hours, 7 days and 28 days, a retention-curve diagnosis, and one packaging rescue rule. It feeds R1 — your own outliers are the best research you own. See section 06.