关于 pre-mortem
pre-mortem 是帮助团队在项目或重大决策前进行失败预演的技能。它通过设想计划已经失败,系统整理风险、评估优先级并制定缓解措施。
主要功能
- 01设想失败并追溯原因
- 02引导个人与团队提出风险
- 03评估风险的可能性与影响
- 04制定预防和缓解计划
- 05提供画布与会议模板
适用场景
- 01产品或活动上线前检查风险
- 02重大招聘或投资决策前分析
- 03团队规划中识别隐忧
安装
复制下面的命令,在终端中运行。
npx skills add https:/ / github.com/ guia-matthieu/ clawfu-skills --skill pre-mortem安装后在 Claude Code 中输入/pre-mortem确认已加载。
详细介绍
事前验尸
想象你的项目遭遇了惨败,然后逆向推演,找出失败的原因。运用 Gary Klein 的“前瞻性回顾”技巧,在失败发生前及时发现问题。
何时使用此技能
- 在发布产品、发起活动或启动重大计划之前
- 在做出重要决策(招聘、投资、合作)之前
- 启动项目时,找出团队尚未考虑到的风险
- 过度自信时,尤其是所有人都认为计划很棒时
- 在投入资源开展重要工作之前
- 团队规划时,提出大家可能不愿分享的担忧
方法论基础
| 方面 | 详情 |
|---|---|
| 来源 | Gary Klein(1989),认知心理学家,自然情境决策研究先驱 |
| 专家 | Klein 的研究表明,事前验尸能将识别结果成因的准确率提高 30% |
| 核心原则 | “前瞻性回顾”——想象某件事已经发生,能显著提升我们解释其原因的能力。未来显得遥远,过去却显得真实。 |
Claude 负责什么,你来决定什么
| Claude 负责 | 你来决定 |
|---|---|
| 制定制作工作流程 | 最终创意方向 |
| 提出技术方案 | 设备和工具选择 |
| 创建模板和检查清单 | 质量标准 |
| 找出最佳实践 | 品牌和风格决策 |
| 生成脚本大纲 | 最终脚本审批 |
此技能的作用
- 揭示隐藏风险——找出前瞻性规划中看不到的危险
- 允许提出批评——团队成员可以放心表达担忧
- 减少过度自信——抵消规划谬误和乐观偏差
- 在执行前改进计划——趁修复成本还低时解决问题
- 营造心理安全感——“我们在设想失败”比“我觉得这会失败”更让人安心
使用方法
开展事前验尸
Run a Pre-Mortem on this project:
[describe the project, product, or decision]
It's one year from now and this has failed completely. What happened?
对决策进行压力测试
Pre-Mortem this decision:
[describe the decision you're considering]
Assume we made this choice and it went badly. What went wrong?
主持团队事前验尸
Help me facilitate a Pre-Mortem session for my team.
Project: [description]
Team size: [number]
Time available: [duration]
Provide the agenda, questions to ask, and how to synthesize findings.
说明
主持事前验尸时,请遵循以下系统化流程:
第 1 步:做好准备
## Pre-Mortem Setup
**Project/Decision:** ________________________________
**Timeline:** [Imagine we're looking back from X months/years]
**The Scenario:**
"It is [future date]. This project has failed—not just underperformed,
but failed spectacularly. The results are in and they're bad.
We're now conducting a post-mortem to understand what went wrong.
Looking back, it's obvious why this failed..."
**Rules:**
1. Failure has ALREADY happened (this is a fact in this exercise)
2. Everyone must contribute at least one reason
3. No defending the plan (we're explaining failure, not preventing it)
4. Be specific, not vague
5. "I had a bad feeling" is valid (hindsight makes it concrete)
第 2 步:个人生成失败原因(静默进行)
## Individual Brainstorm (5-10 minutes)
Each person silently writes answers to:
"Looking back, this project failed because..."
1. ________________________________________________
2. ________________________________________________
3. ________________________________________________
4. ________________________________________________
5. ________________________________________________
**Prompt Questions:**
- What did we overlook?
- What assumption proved wrong?
- What external event derailed us?
- Who let us down (vendor, team member, partner)?
- What resource ran out?
- What competitor move hurt us?
- What customer behavior surprised us?
- What technical problem emerged?
- What communication failure occurred?
- What timeline assumption was wrong?
先静默进行的原因:
- 避免群体迷思
- 让性格内向的成员也能平等发言
- 在讨论缩小关注范围之前,收集多元观点
第 3 步:分享并归类
## Round Robin Sharing
**Process:**
1. Each person shares ONE reason (no discussion yet)
2. Continue rounds until all reasons shared
3. Capture on whiteboard/doc as shared
**Clustering:**
Group similar failure reasons into categories:
### EXECUTION FAILURES
- ________________________________
- ________________________________
### EXTERNAL FACTORS
- ________________________________
- ________________________________
### RESOURCE CONSTRAINTS
- ________________________________
- ________________________________
### ASSUMPTION FAILURES
- ________________________________
- ________________________________
### PEOPLE/TEAM ISSUES
- ________________________________
- ________________________________
### TIMING/MARKET
- ________________________________
- ________________________________
第 4 步:分析并确定优先级
## Risk Assessment Matrix
For each failure reason, assess:
| Failure Reason | Likelihood | Impact | Detectability | Risk Score |
|----------------|------------|--------|---------------|------------|
| [Reason 1] | H/M/L | H/M/L | Early/Late/Never | ___ |
| [Reason 2] | H/M/L | H/M/L | Early/Late/Never | ___ |
| [Reason 3] | H/M/L | H/M/L | Early/Late/Never | ___ |
**Scoring:**
- High Likelihood + High Impact + Late/Never Detection = CRITICAL
- Medium/Medium/Early = MANAGEABLE
- Low/Any/Early = MONITOR
## Priority Failures
**CRITICAL (Must Address):**
1. ________________________________
2. ________________________________
3. ________________________________
**MONITOR (Watch Closely):**
1. ________________________________
2. ________________________________
第 5 步:制定预防和缓解计划
## Prevention Strategies
For each critical risk:
### Risk: [Failure Reason 1]
**Early Warning Signs:**
- ________________________________
- ________________________________
**Prevention Actions:**
- ________________________________
- ________________________________
**If It Happens Anyway (Mitigation):**
- ________________________________
- ________________________________
**Owner:** ____________
**Review Date:** ____________
---
### Risk: [Failure Reason 2]
[Repeat structure]
---
## Tripwires (Automatic Triggers)
"If we see [sign], we immediately [action]"
1. If ______________, then ______________
2. If ______________, then ______________
3. If ______________, then ______________
第 6 步:纳入计划
## Updated Project Plan
### Original Assumptions (Now Questioned)
| Assumption | Pre-Mortem Concern | Validation Plan |
|------------|-------------------|-----------------|
| [Assumption] | [Risk raised] | [How to test] |
| [Assumption] | [Risk raised] | [How to test] |
### New Safeguards Added
- [ ] ________________________________
- [ ] ________________________________
- [ ] ________________________________
### Monitoring Dashboard
Track these metrics as early warning system:
1. ________________________________
2. ________________________________
3. ________________________________
### Scheduled Check-ins
- Week 2: Review assumption #1
- Week 4: Check metric #2
- Month 2: Full pre-mortem review
示例
示例 1:产品发布事前验尸
背景: 一家 SaaS 公司将在 3 个月后发布新的企业产品。
输入:
“我们将在第二季度发布企业产品。大家都很兴奋。来做一次事前验尸——是什么让我们失败了?”
输出:
情景: “现在是 9 月。我们第二季度的企业产品发布失败了。我们的资金储备已经快耗尽,产品也没有获得市场认可,我们正在考虑转型。出了什么问题?”
生成的失败原因:
执行失败:
- 我们低估了企业销售周期(需要 6~12 个月,而不是 2 个月)
- 我们没有预料到安全审查要求(SOC2 等)
- 产品实际上还没有达到企业级标准(缺少审计日志、SSO)
- 客户支持团队无法满足企业客户的期望
市场/时机: 5. 由于经济状况,企业预算被冻结 6. 两家竞争对手在第一季度推出了类似产品 7. 目标公司处于“观望”状态
内部/团队: 8. 销售团队没有企业销售经验 9. 工程团队的精力分散在发布工作和错误修复之间 10. 营销信息没有引起企业买家的共鸣
资源: 11. 在足够多的交易达成之前,资金储备就耗尽了 12. 无法及时招聘到企业销售代表
风险评估:
| 失败原因 | 可能性 | 影响 | 优先级 |
|---|---|---|---|
| 低估企业销售周期 | HIGH | HIGH | CRITICAL |
| 产品未达到企业级标准 | HIGH | HIGH | CRITICAL |
| 预算冻结 | MEDIUM | HIGH | WATCH |
| 竞争对手发布产品 | MEDIUM | MEDIUM | MONITOR |
| 销售团队缺乏经验 | HIGH | HIGH | CRITICAL |
预防计划:
风险 1:低估销售周期
- 预防:现在就开始与企业客户沟通(提前 3 个月)
- 验证:发布前拿到 5 份 LOI,而不是接到 5 个表示“感兴趣”的电话
- 预警信号:如果第 4 周前还没有 LOI,就暂停发布
风险 2:产品未达到企业级标准
- 预防:聘请企业顾问审查需求
- 行动:在发布前优先实现 SSO、审计日志和 SLA
- 预警信号:如果未能通过安全问卷审查,就推迟发布
风险 3:销售团队缺乏经验
- 预防:聘请一名资深企业销售代表或顾问
- 替代方案:与专注企业市场的经销商合作
- 验证:全面发布前,先完成一个试点项目
更新后的计划:
- 将发布推迟到第三季度(采用现实的时间安排)
- 将 SOC2 认证加入路线图
- 在第一季度招聘企业销售负责人
- 在公开发布前,与 3 家设计合作伙伴达成合作
示例 2:招聘决策预演
背景: 初创公司即将招聘一位关键的高级员工。
输入:
“我们准备招聘一位工程副总裁。年薪 30 万美元,外加股权,是一项重大投入。对这个决定做一次预演——这次招聘会如何失败?”
输出:
场景: “12 个月后。这次工程副总裁招聘彻底失败。我们已经(付出高昂代价后)分道扬镳,团队士气低落,进度也落后了 6 个月。发生了什么?”
失败原因:
人岗不匹配:
- 擅长大公司流程,却无法应对初创公司的模糊性
- 文化不匹配——对于我们的扁平结构来说,过于等级森严
- 技术技能过时(只有旧技术栈的经验)
- 想打造自己的势力范围,而不是交付产品
流程不佳: 5. 我们急于招人,推进得太快 6. 推荐人都是朋友,而非曾向其汇报过的下属 7. 没有测试实际工作能力(没有试做项目) 8. 创始人的个人魅力掩盖了面试中的危险信号
预期不匹配: 9. 我们说要招“副总裁”,实际需要的却是“亲力亲为、善于指导团队的人” 10. 薪酬预期不断提高 11. 他们期待的团队规模和预算超出了我们的能力
外部因素: 12. 其他公司的挽留报价让他们心思不定 13. 他们的配偶不喜欢搬迁 14. 他们把这份工作当作“跳板”
风险评估:
| 失败原因 | 可能性 | 影响 | 优先级 |
|---|---|---|---|
| 不适合初创阶段 | 高 | 严重 | 严重 |
| 背景调查不充分 | 中 | 高 | 严重 |
| 预期不匹配 | 高 | 高 | 严重 |
| 没有试用期 | 中 | 高 | 行动 |
预防计划:
风险 1:不适合初创阶段
- 预防:面试时明确讨论初创公司的混乱状况
- 测试:“告诉我一次你在需求不完整时交付产品的经历。”
- 危险信号:所有案例都来自大公司
风险 2:背景调查不充分
- 预防:联系他们未列出的推荐人
- 最低要求:联系 2 位曾直接向他们汇报过的人(而非朋友)
- 提问:“你会热情地再次聘用他们吗?”
风险 3:预期不匹配
- 预防:签署书面预期说明文件
- 包括:预算、团队规模、第一年目标
- 讨论:“如果预算削减一半会怎样?”
风险 4:没有试用期
- 预防:发出录用通知前,进行为期 2 周的付费试做项目
- 替代方案:先担任一段时间的顾问
- 至少做到:共同用半天时间解决一个实际问题
更新后的招聘流程:
- 在面试中加入有关“初创公司混乱状况”的情景题
- 背景调查包括通过 LinkedIn 联系未列出的推荐人
- 书面预期说明文件
- 最终录用前进行为期 2 周的付费试做
- 入职 90 天时进行里程碑检查
清单与模板
预演画布
┌─────────────────────────────────────────────────────────────────┐
│ PRE-MORTEM CANVAS │
├─────────────────────────────────────────────────────────────────┤
│ │
│ PROJECT: _________________________________________________ │
│ │
│ DATE OF "FAILURE": _________________ │
│ │
│ "It is [date]. This project has failed completely." │
│ │
├─────────────────────────────────────────────────────────────────┤
│ │
│ WHY IT FAILED: │
│ │
│ 1. _______________________________________________________ │
│ 2. _______________________________________________________ │
│ 3. _______________________________________________________ │
│ 4. _______________________________________________________ │
│ 5. _______________________________________________________ │
│ │
├─────────────────────────────────────────────────────────────────┤
│ │
│ TOP 3 RISKS TO ADDRESS: │
│ │
│ 1. _________________ Prevention: _______________________ │
│ 2. _________________ Prevention: _______________________ │
│ 3. _________________ Prevention: _______________________ │
│ │
├─────────────────────────────────────────────────────────────────┤
│ │
│ TRIPWIRES: │
│ │
│ If ___________________ → Then ___________________ │
│ If ___________________ → Then ___________________ │
│ │
└─────────────────────────────────────────────────────────────────┘
团队预演引导指南
## Pre-Mortem Meeting Agenda
**Duration:** 45-60 minutes
**Attendees:** Full project team
**Materials:** Sticky notes, timer, whiteboard
### 1. SETUP (5 min)
"It's [future date]. This project has failed—not just missed
targets, but failed spectacularly. We're doing a post-mortem.
With the benefit of hindsight, why did this fail?"
### 2. SILENT BRAINSTORM (7 min)
- Everyone writes failure reasons on sticky notes
- One reason per note
- Minimum 5 per person
- No talking
### 3. ROUND ROBIN (15 min)
- Each person shares one note at a time
- Post to wall, no discussion yet
- Continue until all shared
- Cluster similar items
### 4. DISCUSS & PRIORITIZE (15 min)
- Which risks are most likely?
- Which would be most damaging?
- Dot vote: everyone gets 3 dots
### 5. ACTION PLANNING (15 min)
- Top 3 risks get prevention plans
- Assign owners
- Set tripwires
### 6. CLOSE (3 min)
- Capture in doc
- Schedule follow-up
- Thank team for honesty
预演问题库
## Pre-Mortem Prompts
Use these to stimulate thinking about different failure modes:
### EXECUTION
- What part of the plan was unrealistic?
- Where did quality suffer?
- What dependency let us down?
### PEOPLE
- Who underperformed or left?
- What conflict derailed us?
- What communication failed?
### RESOURCES
- What resource ran out?
- What was more expensive than planned?
- What took longer than expected?
### MARKET/EXTERNAL
- What changed in the market?
- What competitor move hurt us?
- What customer behavior surprised us?
### ASSUMPTIONS
- What "fact" proved wrong?
- What did we overlook?
- What seemed obvious in hindsight?
### TIMING
- What should we have done sooner?
- What external event affected us?
- What deadline was unrealistic?
预演与事后复盘
## When to Use Each
### PRE-MORTEM (Before)
✓ Project hasn't started yet
✓ Major decision pending
✓ Team is overconfident
✓ High-stakes initiative
✓ Want to find problems early
### POST-MORTEM (After)
✓ Project has completed (success or failure)
✓ Want to capture lessons learned
✓ Building institutional knowledge
✓ Improving future processes
### KEY DIFFERENCE
Post-Mortem: "What happened?" (Limited by actual events)
Pre-Mortem: "What COULD happen?" (Unlimited imagination)
Research shows Pre-Mortem generates 30% more failure reasons.
Why? "Prospective hindsight" makes imagined failure feel real.
技能边界
此技能擅长的方面
- 规划音频制作工作流
- 提供技术指导
- 制定质量检查清单
- 提出创意方法
此技能无法做到的事
- 取代音频工程专业知识
- 做出主观的创意决策
- 直接访问或编辑音频文件
- 保证商业成功
参考资料
- Klein, Gary. “Performing a Project Premortem”《哈佛商业评论》(2007)
- Klein, Gary. “Sources of Power: How People Make Decisions”(1998)
- Kahneman, Daniel. “Thinking, Fast and Slow”(2011)——包含 Klein 的研究
- Heath, Chip & Dan. “Decisive”(2013)——预演的应用
相关技能
技能元数据(内部使用)
name: pre-mortem
category: strategy
subcategory: risk-analysis
version: 1.0
author: MKTG Skills
source_expert: Gary Klein
source_work: "Performing a Project Premortem" (HBR 2007)
difficulty: beginner
estimated_value: $1,000 risk assessment consultation
tags: [risk-management, planning, decision-making, Klein, prospective-hindsight]
created: 2026-01-25
updated: 2026-01-25