# Retention-Led Shorts Development Notes

Use these notes when turning a short-form video bot from a renderer into a growth engine.

## What tends to stop the scroll

Prioritize ideas and scripts with these traits:

1. **Immediate self-relevance** — the first line names a concrete situation the viewer recognizes: “If your brain has 47 tabs open...”
2. **Open loop** — the hook promises a specific resolution: “do this before...”, “why...”, “stop...”, “not...”, “instead...”.
3. **Specificity** — numbers, named platforms/tools, exact timeframes, visible objects, or concrete scenarios beat generic advice.
4. **Tension / contrarian angle** — challenge a common belief without cheap outrage: “You do not need more discipline. You need a smaller first step.”
5. **Concrete payoff** — a prompt, template, before/after, checklist, exact micro-action, or decision rule.
6. **Visual pattern interrupt** — a motion/change in the first 1–2 seconds; avoid static AI-art slideshows.
7. **Conversation trigger** — a specific comment keyword or question tied to the payoff.
8. **Niche authority** — recognizable viewpoint and repeated format, not broad generic life advice.

## Practical scoring signals

Add or reward:

- direct address (`you`, `your`, `du`, `dein`)
- open-loop words (`do this`, `before`, `why`, `stop`, `not`, `instead`)
- specificity markers (numbers, named tools, `minute`, `tabs`, `AI/KI`, platform names)
- tension or myth-busting (`not`, `without`, `stop`, `instead`)
- short hook (roughly <= 14 words when possible)
- visual energy (before/after, screen recording, timer, cards, close-up, fast cut)
- concrete payoff (template, prompt, calendar block, two-minute action)
- comment CTA (`Comment RESET`, `Comment CRITIC`, “Which app...?”)

Penalize or reject:

- slow intros: “In this video...”, “Today I will talk about...”
- generic value: “productivity tips”, “habits”, “mindset” without a concrete mechanism
- static slideshow / abstract background with no visual transformation
- unproven videos over ~90s
- AI-generated fake UI/text, illegible charts, or obvious “AI slop” artifacts

## Testbatch pattern

For early channels, structure tests as:

- 2–3 non-conflicting content pillars.
- 3 ideas per pillar.
- 6 growth candidates around 25–45s.
- 3 monetization candidates around 61–75s if TikTok Creator Rewards is relevant.
- Generate review drafts first, then finalize voiceover for top 3, then build shot plans, then render.

## Example safer pillars for AI/productivity channels

If faith/religion creates brand or policy ambiguity, pause it and test:

- `life_optimization` — systems over motivation, calendars, task triage, small first steps.
- `ai_life_systems` — AI as critic, organizer, assistant, approval-gated automation.
- `attention_design` — doomscrolling, defaults, phone/app environment design, algorithm incentives.

## Voiceover finalization rules

- Keep the hook as the exact first spoken sentence.
- Make 25–45s growth scripts roughly 70–110 words.
- Make 61–75s TikTok reward candidates roughly 140–180 words.
- Use plain spoken language; no academic transitions.
- For AI content, keep the human decision/approval moment explicit.
- For attention design, avoid moral panic; frame changes as defaults and incentives.
- Generate caption beats and production notes before rendering.