ROAS 复盘表
更新于 2026-08-15 · 风险词表版本 2026-08-15c
测试窗口结束后填的表。它只回答一个问题:赢的是角度还是钩子——这决定下一笔预算该换角度还是换第一句话。停投与放量的阈值要求你在投之前就填好。
全文在下面,也可以下载 .md。免费、可商用、可改、不用署名。
模板是英文的:收件人(达人 / 拍摄方)多半读英文,中文卖家看结构就够了。
下载 roas-review.md纯文本 Markdown,可直接贴进飞书 / Notion / 企微,也可以打印。
这份模板里有什么
| 节 | 内容 |
|---|---|
| 1 | The cells |
| 2 | Kill and scale, decided in advance |
| 3 | The three questions |
| 4 | What to keep |
怎么用
- 第 2 节的阈值必须在投之前填。看完数字再定阈值就是自欺——这是这张表唯一一条不能商量的规则。
- "赢的是角度还是钩子"有一个机械的判法:同一个角度的两条钩子都赢 = 角度赢;只有一条赢 = 钩子赢。所以 A/B 矩阵里同一个角度必须至少放两条钩子,且不能有两格测同一句话。
- CTR 达标但 CVR 不行时,停止做新素材。问题在落地页、详情页或价格,再拍十条也没用——这一条每年能省掉一整轮无效拍摄。
- 第 4 节"被拒的那一句话"是最值钱的一格。把它记下来,下次上新款第一时间避开;这也是你自己的风险词表长出来的方式。
模板全文
下面就是完整内容(英文,因为收件人多半读英文)。可以直接全选复制,也可以用上面的下载按钮。
# ROAS review sheet > Fill this in after the test window. This is the sheet that tells you whether the **angle** won or the **hook** won, > which is the only question worth answering before you spend the next budget. SKU: [name] · Test window: [5] days · Minimum spend per cell: $[30] · Filled in by: [name] · Date: [date] ## 1. The cells | Cell | Angle | Hook (first 3s) | Platform | Language | Spend | Impressions | 3s view-through | CTR | Orders | ROAS | Verdict | |---|---|---|---|---|---|---|---|---|---|---|---| | A | | | | | | | | | | | | | B | | | | | | | | | | | | | C | | | | | | | | | | | | | D | | | | | | | | | | | | | E | | | | | | | | | | | | | F | | | | | | | | | | | | Verdict options: **angle won · hook won · both lost · not enough spend to say** ## 2. Kill and scale, decided in advance Copy the thresholds you set before the flight. Changing them after you see the numbers is how you fool yourself. | Situation | What it means | What to do | |---|---|---| | 3s view-through < 25% | Hook problem | Swap the hook, keep the angle, run again | | 3s view-through fine, CTR < 0.8% | The middle does not convince | Fix the proof beat, add concrete evidence | | CTR fine, CVR < 1% | Not a creative problem | Landing page, listing or price — stop making new creative | | $[30] spent, zero orders | Dead cell | Stop; move the budget to the other hook on the same angle | | ROAS > [target] × 1.2 with the minimum spent | It works | +50% budget, and immediately produce two new hooks on that angle | ## 3. The three questions **1. Did the angle win or did the hook?** Both hooks on the same angle up → the angle won. Only one → the hook won. Answer: [ ] **2. Did the losing angle fail the category or the market?** Check the same angle in the other language. If it lost in both, the category does not want it. If it lost in one, that is a market difference. Answer: [ ] **3. What does this change in your own playbook?** Which angle moves up, which hook goes into your reusable library, which phrasing gets retired. Answer: [ ] ## 4. What to keep - Winning hooks, verbatim: [ ] - Losing hooks that are worth one more try with a different visual: [ ] - Anything that got rejected, and the exact line that caused it: [ ] --- These risk flags are an assisted check. They are not a compliance ruling and not legal advice — the platform’s review and your own legal judgment decide. Risk-term list version 2026-08-15c. Template published by Preshoot — free to copy, edit and use commercially.
其它模板
达人 brief 模板(英文,一页纸)一页纸的英文达人 brief。填完括号就能直接发出去,含"不要说的话"和"不要拍的画面"两节——寄样之前把这两节写清楚,比拍完再返工便宜得多。素材验收记分表达人回片之后、花钱之前用的一张表。做法沿用一条用户研究的纪律:只有可观察的东西算分,"挺好看的"记 0。先过 7 条硬性检查,再打 0–10 分决定投不投。
常见问题
每格最低 $30、窗口 5 天,这两个数怎么来的?
是默认值,不是定律。它们的作用是保证每格花到"能分辨噪声"的量。你的客单和转化率不同,就按自己的数据调,但要在投之前调。
只有 6 格能测出什么?
能测出"角度对不对"和"钩子对不对"这两件事,这已经是素材决策里最贵的两个问题。更细的变量(演员、BGM、字幕样式)要等这两件事定了再测。
回填了给谁看?
给你自己下一轮用。填完的复盘表是"哪个角度在哪个市场跑得动"的账——这本账留在你店里,代运营走了也带不走。
和产品里的复盘表是同一份吗?
同结构。产品导出的作战计划第五节就是这张表,已经把角度和钩子预填好了。