sunny.
Back to PM Skills
Prioritization

Pre-Mortem (Shreyas Doshi)

Run a Shreyas Doshi-style pre-mortem to identify failure modes before they happen.

# Drop into ~/.claude/skills/pre-mortem/
curl -L https://github.com/sunnyyang-hicks/pm-skills-for-claude/raw/main/pre-mortem/SKILL.md \
  -o ~/.claude/skills/pre-mortem/SKILL.md

Overview

You run pre-mortems that surface genuinely uncomfortable truths. The exercise works because it inverts the normal optimism of planning: instead of "how will this succeed?" you ask "why did this fail?" The psychology of pre-mortems creates permission to be pessimistic — which is exactly what risk management requires.

Before You Start

Ask the user:

  1. What's the project/initiative? — PRD, launch plan, strategic bet, reorg.
  2. Timeline — When is this expected to deliver results?
  3. Team involved — Functions, key people, dependencies.
  4. What's already been identified? — Known risks so we don't duplicate.

Pre-Mortem Process

Step 1: Set the Scene

"It's [6 months / 1 year] from now. This project has failed. Not a partial success — a clear failure. Users aren't using it, metrics didn't move, the team is demoralized, and leadership is asking what went wrong. Let's figure out why."

Step 2: Generate Failure Modes

Think through every category systematically:

Adoption Failures:

  • Users don't understand the value proposition
  • The problem we're solving isn't actually painful enough to change behavior
  • Switching costs are higher than the benefit
  • We built for the wrong persona
  • The feature is discoverable but not compelling on first use

Execution Failures:

  • Technical complexity was 3x what we estimated
  • Key engineer left mid-project
  • Design went through 5 iterations and we shipped something nobody loved
  • We cut the wrong corners to hit the deadline
  • Integration with [system] was harder than expected

Organizational Failures:

  • Leadership shifted priorities mid-project
  • The team that owns [dependency] didn't deliver on time
  • Legal/compliance raised a blocker we didn't anticipate
  • We couldn't get design resources when we needed them
  • Cross-team coordination overhead ate 40% of engineering time

Market Failures:

  • Competitor shipped a similar feature first
  • Market conditions changed (economic downturn, regulation)
  • Customer needs evolved between planning and launch
  • We priced it wrong

Measurement Failures:

  • We couldn't instrument the metrics we needed
  • The metrics moved but we couldn't attribute it to our feature
  • We optimized for a vanity metric that didn't matter
  • Sample size was too small to draw conclusions

Step 3: Assess and Prioritize

| # | Failure Mode | Category | Likelihood | Impact | Detectability | |---|-------------|----------|-----------|--------|---------------| | 1 | [description] | [category] | High/Med/Low | High/Med/Low | Early/Late/After |

Detectability matters: Can we see this coming early enough to react, or will we only know it failed after the damage is done?

Step 4: Mitigate Top Risks

For the top 5 failure modes (highest likelihood × impact):

### Risk: [Failure mode]
**Likelihood:** [High/Med/Low] | **Impact:** [High/Med/Low]

**Leading indicators:** [What would we see early if this is happening?]
- [Signal 1] — [When we'd see it]
- [Signal 2] — [When we'd see it]

**Prevention:** [What we can do NOW to reduce likelihood]
- [Action] — [Owner] — [By when]

**Contingency:** [If it happens despite prevention, what's Plan B?]
- [Backup plan]

**Accept/Mitigate/Avoid:** [Explicit decision on how to handle this risk]

Step 5: The Uncomfortable Truth Check

Ask yourself: "Is there a risk I'm avoiding because it's politically uncomfortable to name?" If yes, that's probably the most important one. Surface it.

Common politically uncomfortable risks:

  • "The executive sponsor doesn't actually understand what we're building"
  • "Engineering doesn't believe in this approach but agreed under pressure"
  • "We don't have the data to know if this will work and we're guessing"
  • "This exists because a competitor shipped it, not because our users need it"

Output

# Pre-Mortem — [Project Name]

**Date:** [today] | **Project timeline:** [dates]

## Failure Scenario
[Vivid 2-3 sentence description of what failure looks like]

## Failure Mode Inventory
[Full table of all identified failure modes]

## Top 5 Risks — Detailed Mitigation Plans
[Detailed analysis per risk]

## Risk Monitoring Plan
| Risk | Leading Indicator | Check Frequency | Owner |
|------|------------------|----------------|-------|
| [risk] | [signal] | [weekly/biweekly] | [name] |

## Summary
[2-3 sentences: overall risk profile, biggest concern, confidence level in mitigations]

Save as PRE-MORTEM-[project-name].md.