视频剪辑编排

| Time | Action | Reason |
| --- | --- | --- |
| 0.0-2.0 | tighten hook cut | remove dead air |
| 2.0-6.0 | add product close-up | make product role clear |
| 6.0-10.0 | insert proof B-roll | reduce claim risk |
| 10.0-15.0 | end card + CTA | clarify next action |

This snapshot is an inferred planning readout for customers who have not connected live platform accounts yet. It is safe to show as a benchmark scaffold because it labels assumptions and asks for concrete first-run signals.
Confidence: planning
Assumptions:
- Primary platform model: Instagram Reels.
- Execution type: document.
- No real performance data was provided; this is a first-launch planning baseline.
Modeled metric readout:
| Metric | Value | Percentile | Improve hint |
| --- | --- | --- | --- |
3 秒留存率 | 47% | 35 / P25-P50 · 中下 | 重写前 3 秒：用反常识钩子、数字冲击或视觉反差开头
互动率 | 3.7% | 38 / P25-P50 · 中下 | 结尾加一个具体提问或下决策的 CTA，别让观众无话可说
收藏率 | 1.5% | 61 / P50-P75 · 中上 | 加"清单 / 对比 / 步骤" 等可参考结构，让内容值得收藏
评论率 | 0.45% | 66 / P50-P75 · 中上 | 放一个"可争议 / 可分享经验" 的小问题在结尾
分享率 | 0.35% | 41 / P25-P50 · 中下 | 让内容有"转发给 XX 看 / 这个你身边人一定用得上" 的触发点
Next signals to collect:
- 3-second hold or first-frame drop-off
- Save rate or collection intent
- Comment reason: objection, curiosity, or price concern
- Share reason: utility, surprise, or identity fit
- CTA intent after the final beat

Next-loop artifact preview:
This pack converts user-provided campaign feedback into the next execution constraints. It does not read live ad-account or platform analytics.

Signals received:
- Modeled overall: 整体 P25-P50 · 中下
- Weakest modeled metric: 3 秒留存率
- Confidence: planning

Keep doing:
- Preserve the first structure that produced explicit audience response or measurable retention.

Fix next:
- Weakest modeled metric: 3 秒留存率

Next test matrix:
- Test one new hook while keeping the winning proof beat fixed.
- Test one stronger visual proof moment by second 3.
- Test one shorter caption layer with fewer than 9 words.

Optimization queue:
- Lock: Preserve the first structure that produced explicit audience response or measurable retention.
- Rewrite: Weakest modeled metric: 3 秒留存率
- Generate: one conservative variant, one bolder hook variant, one proof-first variant.
- Measure: hook hold, save/share/comment signal, and CTA intent before scaling.