20 Claude Prompts for Hypothesis Testing
The full hypothesis-driven consulting sequence in 20 copy-paste prompts. Covers forming and decomposing hypotheses, designing tests with explicit falsification criteria, evaluating evidence strength, revising under contradictory data, and communicating tested conclusions to partners and clients.
Free skills and prompts for Claude and strategy work
Templates for Claude, ChatGPT and Perplexity — from diagnostics to board-ready decks.
What this is
Twenty ready-to-use Claude prompts for hypothesis-driven strategy work. They cover the complete sequence from forming a falsifiable hypothesis at engagement kick-off through testing, mid-engagement revision, and communicating the final tested conclusions to partners and clients.
Hypothesis-driven problem solving is the structural discipline at the core of MBB consulting: start with an answer, design a test, find the evidence, and update the answer. The failure mode is not a lack of effort - it is poor hypothesis quality at the start, confirmation bias in evidence gathering, and conclusions that are never explicitly tested against rejection criteria. These prompts address each failure point directly.
Each prompt enforces a specific discipline: falsifiability testing, MECE decomposition, confirmation and rejection criteria set before evidence arrives, evidence strength scoring, and formal close-out when a hypothesis is rejected. Use them in sequence for a full engagement, or use individual prompts on the specific part of the hypothesis process that is currently breaking down.
These prompts pair naturally with issue framing prompts for structuring the problem before the hypothesis is formed, and with storyline prompts for converting confirmed hypotheses into a presentation argument.
The 20 Prompts
The prompts are grouped into five phases of hypothesis-driven work. Click any prompt to expand it and copy the full text.
Hypothesis Formation
Use these before analytical work begins. They establish the hypothesis that will structure the engagement: the initial answer, its decomposition into testable sub-hypotheses, the priority order for testing, and the quality bar each hypothesis must clear before a workstream is assigned to it.
Draft the Initial Hypothesis
Use when: You have a business question and early context, and need to form an explicit hypothesis before starting analytical work
Output: Primary hypothesis as one sentence, reasoning chain, two alternative hypotheses with what-would-need-to-be-true conditions, and the key assumption
Build a Hypothesis Tree
Use when: You have a primary hypothesis and need to decompose it into testable sub-hypotheses before building a work plan
Output: MECE hypothesis tree with 3-4 second-level sub-hypotheses, 2-3 supporting conditions per sub-hypothesis, a completeness check, and the two highest-stakes branches
Prioritize Hypotheses by Impact and Testability
Use when: You have a list of competing or parallel hypotheses and need to decide which to test first given limited time and resources
Output: Impact-testability scoring for each hypothesis, four-quadrant placement, ranked testing sequence, and risk note for each deprioritized hypothesis
Reframe a Vague Hypothesis
Use when: A hypothesis is too broad, circular, or non-falsifiable and needs sharpening before work begins
Output: Pass or fail verdict on four quality tests (falsifiability, specificity, answering the question, actionability), a rewritten hypothesis that passes all four, and an explanation of each edit
Test Design
Use these to design how a hypothesis will be tested before evidence collection begins. A test plan written upfront sets explicit confirmation and rejection criteria that prevent the team from deciding what the evidence means after it has arrived.
Design the Test Plan for a Hypothesis
Use when: You have a confirmed, well-formed hypothesis and need to design a rigorous plan to test it with evidence
Output: Critical fact, confirmation criteria, rejection criteria, 3-5 evidence sources, required analyses, timeline estimate, and risk of false confirmation
Identify the Critical Fact
Use when: You are designing a workstream and want to identify the single piece of evidence that would most efficiently resolve whether a hypothesis is true or false
Output: Critical fact stated precisely, explanation of why it is decisive, most accessible source, threshold values for confirmation and rejection, and misleading risk
Design a Falsification Test
Use when: You have a hypothesis the team or client is strongly attached to, and want to design a test that genuinely attempts to disprove it
Output: Null hypothesis, strongest existing counterevidence, three tests designed to confirm the null, abandonment conditions, and the cognitive bias being addressed
Build a Workplan from Hypothesis to Conclusion
Use when: You have a hypothesis tree and need to convert it into a project workplan with owners, deliverables, and deadlines
Output: Workstream table with owner, deliverables, due dates, and early-close conditions; critical path identification; recommended first hypothesis; and check-in structure
Evidence Evaluation
Use these once evidence has been collected. They assess whether the evidence is strong enough to support a verdict, whether it covers the full hypothesis tree, and whether the team is missing anything a partner or client will challenge.
Evaluate Evidence Strength
Use when: You have gathered evidence for a hypothesis and want to assess how much weight it actually carries before building conclusions on top of it
Output: Source reliability, relevance, and recency scores per piece of evidence; identification of weak sources driving conclusions; overall verdict (Strong, Moderate, or Weak); and the biggest evidence gap
Map Evidence to Hypotheses
Use when: You have accumulated a body of evidence from multiple workstreams and need to verify each hypothesis has sufficient coverage before drawing conclusions
Output: Evidence-to-hypothesis mapping by tree node, coverage gaps, single-source risks, over-supported branches, coverage summary (Covered, Partially Covered, or Gap), and minimum additional evidence needed
Assess Whether Evidence Confirms or Rejects a Hypothesis
Use when: You have gathered the planned evidence and need to make a rigorous verdict on whether a hypothesis is confirmed, rejected, or inconclusive
Output: Per-evidence assessment against confirmation and rejection criteria, conflict identification, formal verdict (Confirmed, Rejected, or Inconclusive), next steps if Inconclusive, and confidence level
Identify Missing Evidence
Use when: You are preparing to conclude but want to audit for evidence gaps that could undermine the recommendation when challenged by a partner or client
Output: Logical gap analysis, stakeholder gap analysis, alternative explanation gap analysis, and per-gap priority classification (Critical, Important, or Low)
Hypothesis Iteration
Use these when evidence changes the picture mid-engagement. Hypotheses rarely survive their first contact with data unchanged. These prompts handle revision, pivots, formal close-outs, and provisional synthesis at check-in points.
Revise a Hypothesis After Contradictory Data
Use when: Evidence has come in that clearly contradicts the current hypothesis and you need to revise it rather than defend it
Output: Diagnosis of which part was invalidated, retained valid components, revised hypothesis consistent with all evidence, key new assumption, and one high-priority test for the revised hypothesis
Promote a Sub-hypothesis to Main
Use when: A sub-hypothesis has emerged as more important than the primary hypothesis and you need to decide whether to restructure the engagement around it
Output: Assessment on four dimensions (insight value, completeness, scope alignment, risk of abandoning original), and a recommendation to promote, maintain, or run both
Kill a Hypothesis and Capture the Learning
Use when: A hypothesis has been definitively rejected and you need to formally close it and ensure the learning informs the rest of the engagement
Output: Rejection summary, what the hypothesis got right, implications for the primary hypothesis and remaining branches, and early warning signs missed for future pattern recognition
Synthesize Partial Evidence into Provisional Conclusions
Use when: You are midway through the engagement and need to form provisional conclusions from incomplete evidence to give the team direction before all analyses are finished
Output: Provisional verdict per workstream (Leans Confirmed, Leans Rejected, or Open), provisional answer to the engagement question, identification of critical vs. no-longer-critical open workstreams, and the single most important evidence to get in the next 48 hours
Communication
Use these to translate tested conclusions into outputs the client can act on: a presentation storyline built from the verified hypothesis tree, findings slides for individual tested hypotheses, communication of invalidated beliefs to sponsors, and a next-steps structure that converts confirmed conclusions into owned actions.
Write the Hypothesis-Driven Storyline
Use when: Your hypotheses have been tested and you need to convert the findings into a presentation storyline where the structure reflects the testing process
Output: SCQA opening, action slide titles in presentation order for confirmed sub-hypotheses, placement decision for rejected hypotheses, and a recommendation slide title with key actions
Present Findings from a Tested Hypothesis
Use when: You need to write the body of a findings slide for a specific tested hypothesis, with a clear verdict and selected evidence
Output: Verdict statement, 3-5 evidence bullets in findings format, counterevidence acknowledgment, and implication for the primary hypothesis
Communicate an Invalidated Hypothesis to a Partner
Use when: A key hypothesis has been rejected and you need to communicate that to a partner or client sponsor who had a strong prior belief in it
Output: Four-part communication structure: evidence-grounded opening, specific evidence bullets, implication framing, and a direction-setting next step
Draft the Next Steps After Hypothesis Testing
Use when: The hypothesis testing phase is complete and you need to translate confirmed conclusions into a clear, prioritized set of next steps for the client
Output: Three-tier next steps (immediate actions within 30 days, near-term workstreams within 30-90 days, open questions for next generation), and a single decision point for the client within 5 business days
How to use
Find the prompt you need
Each prompt is named for the situation it addresses. The group headings divide prompts by phase of the hypothesis process. Start with Hypothesis Formation at engagement kick-off. Use Evidence Evaluation prompts once workstreams have produced data. Use Communication prompts at the end.
Copy and fill the placeholders
Click Show prompt on the card, then hit Copy. Fill in every {{placeholder}} with your engagement details before pasting into Claude. The placeholders specify exactly what each expects: hypothesis text, evidence items, engagement question, team size, or client context.
Paste into Claude and iterate
Paste the filled prompt into claude.ai and run it. Claude Sonnet 4.6 handles structured analytical reasoning well. For complex falsification tests and multi-branch evidence mapping, Claude Opus 4.8 produces more rigorous analysis.
Tip
For a full engagement from scratch, run prompts 1 through 8 in the first week: form the initial hypothesis, build the tree, prioritize, reframe if needed, design the test plan, identify the critical fact, design the falsification test, and build the workplan. These eight prompts produce the complete analytical agenda before a single slide is written.
When to use each prompt
Not every engagement needs all twenty. Match the prompt to where the hypothesis process is currently breaking down.
What these prompts enforce
Andrew Persh
Founder, Oria
Former McKinsey consultant turned product builder. Andrew founded Oria to help professionals create boardroom-ready presentations without the formatting overhead.