Hypothesis Driven 80/20

Use $ARGUMENTS as initial context.

When to use this skill

• Ambiguous problems that need fast direction before full analysis.
• Prioritizing limited team capacity across competing hypotheses.
• Designing a test plan with clear stop or continue criteria.
• Converting broad strategic questions into measurable experiments.

Required inputs

• Governing business question and target metric.
• Time horizon and decision deadline.
• Data constraints and available analysis bandwidth.

Workflow

1. Define one decision question and explicit success metric.
2. Generate 3-7 falsifiable hypotheses using strict format.
3. For each hypothesis define expected signal, disconfirming signal, and kill threshold.
4. Score hypotheses with 80/20 lens: impact, confidence, effort, speed.
5. Select top 1-2 hypotheses for minimum viable tests.
6. Build workplan with owners, deadlines, and decision checkpoints.

Ask-first questions

Ask up to 3 questions before ranking hypotheses:

1. Which single metric determines decision success?
2. What is the latest acceptable date for a go or no-go decision?
3. Which datasets are trusted and immediately accessible?

Assumption policy

• If critical data is unavailable, proceed with transparent assumptions.
• Annotate each assumption with confidence and validation action.
• Avoid merged hypotheses; keep one causal chain per hypothesis.

Output contract

Always produce these sections in order:

1. Context
2. Decision or Recommendation
3. Analysis
4. Risks
5. Next Actions
6. Assumptions

Guardrails

• Hypotheses must follow: "If X, then Y, because Z.".
• Avoid descriptive statements that cannot be disproven.
• Include both confirming and disconfirming signals.
• Stop low-value analysis once kill threshold is reached.

Resources

• references/hypothesis-design.md - Falsifiability rules and signal design.
• references/80-20-prioritization.md - Scoring and sequencing framework.
• templates/hypothesis-plan.md - Decision-ready hypothesis template.
• examples/hypothesis-example.md - Golden example with incomplete inputs.

Keywords

hypothesis driven, 80/20, falsifiable hypothesis, prioritization, test plan, consulting