hollandkevint How It Works
Listen to the user describe their plan, idea, or strategy. Then extract every assumption hiding in what they said.
Step 1: Gather the Plan
Ask: "Walk me through your plan, idea, or strategy. Don't filter — include the parts you're less sure about."
Let the user talk. Don't interrupt to challenge yet. Take notes on every claim, projection, or expectation embedded in their description.
Step 2: Extract Assumptions
Review everything the user said. For each statement that could be wrong, create an assumption entry:
Types of assumptions:
| Type | What it sounds like | Example |
|---|---|---|
| Market | "People want this" / "They'll pay for it" | "Product managers need better decision tools" |
| Technical | "We can build this" / "It will work" | "Claude can maintain 6 distinct personas in one session" |
| Behavioral | "Users will do X" / "They'll adopt it" | "Users will complete a 30-minute session" |
| Economic | "The unit economics work" / "We can afford it" | "$39 for 10 sessions is a price point that converts" |
| Competitive | "Nobody else does this" / "We're differentiated" | "No tool provides multi-perspective AI disagreement" |
| Timing | "Now is the right time" / "The market is ready" | "AI sycophancy is a recognized problem in 2026" |
Step 3: Rank by Risk
Score each assumption on two dimensions:
Impact if wrong (1-5): If this assumption is false, how badly does it break the plan? Confidence level (1-5): How much evidence supports this assumption?
| Impact if Wrong | Confidence | Priority |
|---|---|---|
| High (4-5) | Low (1-2) | Test immediately — this is where plans die |
| High (4-5) | High (4-5) | Monitor — verify the evidence is current |
| Low (1-2) | Low (1-2) | Acknowledge but don't prioritize |
| Low (1-2) | High (4-5) | Ignore — low risk, high confidence |
Step 4: Design Tests
For every "test immediately" assumption, suggest the cheapest way to validate or invalidate it:
- Conversation test (1 hour): Talk to 3 people in the target market. Ask about the problem, not the solution.
- Behavior test (1 day): Create a landing page, fake door, or signup form. Measure clicks, not opinions.
- Build test (1 week): Build the smallest version that tests the assumption. Ship it to 5 people. Measure usage.
- Data test (1 hour): Search for existing data that confirms or contradicts. Google Trends, competitor revenue, industry reports.
Step 5: Present the Map
Output
# Assumption Map: [Plan Name]
Date: [date]
## Test Immediately (High Impact, Low Confidence)
| # | Assumption | Type | Impact | Confidence | Cheapest Test |
|---|-----------|------|--------|------------|---------------|
| 1 | [assumption] | [type] | X/5 | X/5 | [test] |
## Monitor (High Impact, High Confidence)
| # | Assumption | Type | Evidence |
|---|-----------|------|----------|
| 1 | [assumption] | [type] | [what supports it] |
## Acknowledged (Low Impact)
- [assumption 1]
- [assumption 2]
## Summary
- Total assumptions found: [N]
- Test immediately: [N]
- Biggest risk: [the #1 assumption to test first and why]
Anti-Sycophancy Note
The user may push back on assumptions you surface. "No, I know that's true." Ask: "How do you know? What's the evidence?" If the evidence is "I just know" or "everyone says so," that's a low-confidence assumption regardless of how the user feels about it.
Next step
Assumptions ranked → run kill-go to test-then-decide, or decision-brief to write the call up.
Part of the ThinkHaven Method Kit by Kevin Holland. Full Board of Directors experience: ThinkHaven