Strategic Decision Doc

---
name: strategic-decision-doc
description: Use when structuring a major strategic decision. Triggers on "decision doc", "strategic decision", "decision memo", "should we [major decision]".
---

# Strategic Decision Doc

Most strategic decisions in early-stage companies happen verbally, drift for weeks, and get made by whoever loses patience first. A decision doc forces clarity, surfaces dissent, and creates a record of why you decided what you did — which is useful when, 18 months later, the decision was wrong and you need to learn.

## Required inputs

1. **The decision** — phrased as a question with 2-4 specific options
2. **Context** — 3-5 sentences on what makes this decision live now
3. **Options considered** — including ones you've already dismissed and why
4. **Your current lean** — and how confident
5. **Who needs to weigh in** — and by when
6. **Deadline** — when does this decision lock?

If "the decision" is too vague ("what's our strategy?") push back. Decisions are specific. "Should we open a UK office in Q4 or defer until next year" is a decision. "Should we expand internationally" is not.

## Structure (1 page, ~500 words)

```
# Decision: [Question, phrased specifically]

**Author**: [Founder name]
**Date**: [Date]
**Decision deadline**: [Specific date]
**Decision-makers**: [1-3 people who actually decide]
**Input from**: [Others to weigh in by deadline]

## Context (3-5 sentences)
What's happening that makes this decision live now. Include the cost of NOT deciding.

## What we've already tried / dismissed
1-2 sentences on prior approaches. So we don't re-litigate.

## Options
For each option (2-4 max):

### Option A: [Name]
- **What it is** (1 sentence)
- **Why it might be right** (2 bullets)
- **Why it might be wrong** (2 bullets)
- **Cost / commitment**: [time, $, opportunity cost]
- **Reversibility**: easy / hard / one-way door

### Option B: [Name]
[same structure]

### Option C: Do nothing / wait
ALWAYS include this option. It's often the best one.

## Recommendation
**Recommend Option [X]**.
Rationale: 3-4 sentences. Specific.
Confidence: high / medium / low.

## The dissent (the strongest case AGAINST the recommendation)
Steelman the opposite view. If you can't, you haven't thought hard enough.

## What would change my mind
2-3 specific things that would flip the recommendation. Forces honest framing.

## Implementation if we go with the recommendation
- First 3 actions, with owners
- Decision review date — when we'd revisit if it's working

## Open questions
Things still unresolved. Tag with owner.
```

## Decision quality patterns

### Reversibility test
Before deciding, label the decision:
- **Two-way door** (easy to reverse): decide fast, learn from doing
- **One-way door** (hard to reverse): take the extra week, get the input

Most decisions are two-way doors masquerading as one-way doors. Founders waste time over-deliberating reversible choices.

### The "if we're wrong" test
For the recommendation: what's the cost of being wrong?
- **Cheap to be wrong**: just decide
- **Expensive to be wrong**: write the dissent twice as long, get more input

### The "regret minimization" test
If you imagine yourself 12 months from now:
- Which option will you regret NOT trying?
- Which option will you regret trying?

### Sunk cost test
Are you weighting one option because you've invested in it, not because it's better? Common with team-favored options.

## Anti-patterns

- Three-page decision doc — by then it's a strategy doc, not a decision
- Recommending something without dissent — bias risk
- "Let's gather more data" as the decision — usually a stall, not a choice
- Including 5+ options — usually means you haven't thought hard about which 2 actually matter
- Decision doc that reads like a defense — you've already decided; this is theater

## Tone

- Direct. Recommend something, even if confidence is medium.
- Honest about confidence. "I lean toward A but my confidence is medium" is more useful than fake conviction.
- Time-boxed. Decisions without deadlines drift.

## Output format

Markdown. 1 page when printed. Bold the recommendation, the dissent, and the deadline — those are what get re-read.