| name | hypothesis-tree |
| description | Build a structured hypothesis tree with testable branches to guide analysis. Use when someone asks to "build a hypothesis tree", "structure this problem into hypotheses", "create an issue tree", "break this down MECE", or when a defined problem would benefit from structured decomposition before research begins. This skill is optional — not every problem requires a formal hypothesis tree. Skip it when the question is narrow enough that direct research is more efficient.
|
Hypothesis Tree
Build a hypothesis tree that decomposes the client question into testable branches. This skill is a tool, not a requirement — invoke it only when structuring hypotheses genuinely helps focus the research effort.
When to Use vs. Skip
Use a hypothesis tree when:
- The client question has multiple plausible answers and research needs prioritization
- The problem has 3+ distinct dimensions (e.g., market entry involves market attractiveness, competitive dynamics, and client capabilities)
- The team risks boiling the ocean without a structured view of what to investigate
Skip the hypothesis tree when:
- The question is narrow and specific enough that direct research answers it
- The user explicitly wants to go straight to research
- The situation is better served by a simple decision framework or pros/cons analysis
Process
Step 1: Start with the Governing Hypothesis
Write a single, falsifiable statement that represents the best initial answer to the client question. This is not a guess — it is a reasoned starting position that research will confirm, modify, or disprove.
Format: "We believe [specific answer to the client question] because [initial reasoning]."
The governing hypothesis must be:
- Specific enough to be wrong (if it cannot be disproven, it is not a hypothesis)
- Actionable — if proven true, the client knows what to do
- Grounded in at least preliminary logic, not pulled from thin air
Step 2: Decompose into Branches
Break the governing hypothesis into 2-4 supporting conditions — the things that must be true for the hypothesis to hold.
Each branch must be:
- Mutually exclusive: no overlap between branches
- Collectively exhaustive: together they fully support (or disprove) the governing hypothesis
- Testable: each branch ends in a question that research can answer with evidence
Do not fetishize perfect MECE structure at the expense of speed. A pragmatic decomposition that the team can test in days beats an elegant one that takes a week to build.
Step 3: Define What Proves or Disproves Each Branch
For every branch, specify:
- What evidence would support this branch (and where to find it)
- What evidence would undermine it (and where to find it)
- What threshold or benchmark distinguishes "this is true" from "this is false"
Step 4: Prioritize
Not all branches are equally important or equally uncertain. Mark each branch:
- Priority: High / Medium / Low (based on impact on the overall answer)
- Certainty: High / Medium / Low (based on how much we already know)
- Research approach: What data or analysis is needed to test this branch
The research phase should start with high-priority, low-certainty branches — that is where investigation adds the most value.
Output Format
## Governing Hypothesis
[The falsifiable initial answer]
## Hypothesis Tree
### Branch 1: [Name]
- Statement: [What must be true]
- Supports if: [Evidence that would confirm]
- Undermines if: [Evidence that would disprove]
- Priority: [High/Medium/Low] | Certainty: [High/Medium/Low]
- Research needed: [Specific data or analysis]
### Branch 2: [Name]
[Same structure]
### Branch 3: [Name]
[Same structure]
## Research Prioritization
[Ordered list of what to investigate first and why]
Present the tree to the user for review. Adjust branches based on their feedback — they often know which branches are already settled (no research needed) and which are genuinely uncertain.