hypothesis-engine

Hypothesis Engine

Turn your competitive insights into a testable strategic hypothesis. This skill runs the Double Diamond, first diverging to generate many candidates, then converging to the 2-3 worth validating. You leave with a hypothesis written in two formats: an outcome statement and a validation commitment.

Stage: Focus. Best for paths: product and positioning, ops and automation. Time estimate: 75 minutes

Overview

1. Load context from SWOT and battle card output (or ask)
2. Diverge: generate 6-10 hypothesis candidates using thought starters
3. Evaluate: score each on desirability, feasibility, viability (1-5)
4. Converge: select top 2-3 for dual-format development
5. Hand off the winning hypothesis in the FOUNDER JAM HANDOFF block

Context

Get context from a FOUNDER JAM HANDOFF block pasted in the chat (the format is defined in founder-setup; if founder-setup is not installed, the same format is in HANDOFF-FORMAT.md next to this file). If you are running in Claude Code with file access, config.json in the working directory works too. Ask for anything that is missing, one question at a time.

Sherpa check: if mode is missing or empty, ask once, "Is a sherpa with you today?" Default to solo. With a sherpa, add this skill's challenge prompts. Solo, skip them.

Path check: if path is empty, undecided, or anything other than product-positioning, gtm-growth, sales or ops-automation, help the founder choose one of the four paths (see founder-setup, Step 2) before going further. With a sherpa, decide it together. Write the path the founder confirms into path in the block you print, so the next skill does not ask again. Never carry undecided forward once the founder has chosen. If a Founder Jam config.json (one with startup_name and path keys) exists in the working directory, update path there too and leave every other key as it is.

Fields needed: startup_name, problem_statement, target_market, stage.

If a field is missing, ask conversationally:

• "What's your startup called?"
• "What problem does it solve, and for who?"
• "What stage are you at: idea, pre-revenue, or early-revenue?"
• "Do you have findings from the battle card session? Share any key competitive insights."

Step 1: Frame the hypothesis challenge

Before generating candidates, orient the founder:

"A good hypothesis is specific, time-bound, and falsifiable. We're going to use the Double Diamond, first we'll generate more candidates than you need, then we'll score and select. Don't edit yourself in the diverge phase."

Share the two hypothesis formats:

Format 1, Outcome statement:

"If we [change/experiment], then [target customer] will [measurable outcome] by [specific date]."

Format 2, Validation commitment:

"We believe [ICP] will pay $[Y]/month for [specific outcome]. We will validate this by [specific experiment] within [timeframe]."

On the ops-automation path, if the founder is automating an internal or support process, there is no price to test. Use: "We believe [the people who run this process] will [specific behavior, e.g. use the automation for most tasks] to [specific outcome, e.g. save X hours a week]. We will validate this by [specific experiment] within [timeframe]." Do not ask for a price.

A strong hypothesis is:

• Specific (names the change, the target, the metric)
• Time-bound (has a date or timeframe)
• Falsifiable (you can prove it wrong)
• Connected to a mechanism (the "because" or "by" clause)

Step 2: Diverge phase

Generate 6-10 hypothesis candidates. Use the 20 thought starters below to prompt the founder. Do not generate all 20, pick 6-10 most relevant based on the startup context.

20 hypothesis thought starters

Use these as starting points. Rewrite in the founder's own words using the formats above.

1. Pricing Model If we switch from a one-time fee to a monthly subscription, customers will have a lower barrier to purchase and we will see a 20% increase in new customer acquisition within 90 days, because the upfront cost is the primary objection in sales calls.

2. Customer Retention If we implement a personalized onboarding program within the first 30 days of signup, we will reduce churn by 15% over six months, because customers who do not reach their first value milestone within a month are significantly more likely to cancel.

3. Value Proposition If we change our core value proposition to emphasize time savings over cost savings in all marketing materials, we will increase our conversion rate by 10% within 60 days, because our target customers are time-constrained professionals who value efficiency over price.

4. Channel Effectiveness If we shift 50% of our marketing budget from paid ads to content marketing over the next six months, we will generate more qualified leads at a lower cost per acquisition, because our ICP researches solutions through educational content before making purchase decisions.

5. Sales Funnel If we add a free trial option to our pricing page, we will increase the number of users who progress from consideration to purchase by 25% within 90 days, because customers need to experience the product's value before committing financially.

6. Customer Acquisition If we launch a referral program offering existing customers a one-month discount for each new customer they refer, we will acquire 30% of new customers through referrals within six months, because our current customers have strong networks of peers with the same pain point.

7. Brand Messaging If we reposition our brand as a premium solution focusing on enterprise clients rather than small businesses, we will attract higher-value contracts averaging 3x our current deal size within a year, because our product capabilities align more closely with enterprise requirements.

8. Product-Market Fit If we develop a feature set specifically for [vertical] customers, we will increase adoption in that segment by 40% within six months, because this vertical has underserved needs that our current generic solution does not address well.

9. Service Efficiency If we introduce AI-driven automation to handle routine customer inquiries, we will reduce average response time from 24 hours to 2 hours and decrease support costs by 30% within three months, because the majority of inbound tickets are repetitive and rule-based.

10. User Experience If we simplify our onboarding process from seven steps to three, we will increase the percentage of users who complete setup and reach their first value moment by 35% within 60 days, because friction in onboarding is the primary driver of early drop-off.

11. Partnership Impact If we establish a co-marketing partnership with [complementary product], we will access their customer base and generate 200 qualified leads in the first quarter, because their customers already have the pain we solve and trust that vendor's recommendations.

12. Customer Pain Point If we launch a feature that directly addresses [specific pain point], we will see a 50% increase in usage among affected users and a 20% improvement in retention for that cohort within 90 days, because this pain point is the most frequently cited reason for cancellation in exit surveys.

13. Referral Program If we offer customers a cash incentive of $50 for each successful referral, we will increase our referral rate from 5% to 20% of new customers within six months, because our current customers frequently mention our product to peers but lack a structured incentive to formalize referrals.

14. Customer Loyalty If we introduce a tiered loyalty program rewarding long-term customers with exclusive benefits, we will improve our net promoter score by 15 points and reduce churn among customers of three years or more by 10% within a year, because high-tenure customers feel undervalued compared to new customer promotions.

15. Packaging If we offer a bundled package combining our product with a complementary service at a 15% discount, we will increase average deal size by 25% and reduce competitive pressure within 90 days, because bundled solutions reduce the need for customers to evaluate separate vendors.

16. Feature Usage If we redesign the dashboard to make our most-used feature more prominent and accessible within one click, we will see a 40% increase in daily active usage of that feature within 30 days, because the current navigation buries it and users forget it exists.

17. Geographical Market If we launch a targeted marketing campaign in [city/region], we will acquire 50 new customers in that market within six months, because the demographic and industry mix there closely mirrors our current highest-retention customer segment.

18. Customer Support If we proactively reach out to customers who have not logged in for 14 days with a personalized check-in and help offer, we will re-engage 25% of at-risk users and reduce churn in that cohort by 20% within 90 days, because early inactivity is the strongest predictor of cancellation in our data.

19. Marketing Message If we A/B test two versions of our homepage headline, one focused on outcome and one focused on process, we will identify a clear winner with at least 15% higher conversion within 30 days, because our current messaging has not been validated against user language and may not match how customers describe their own problem.

20. Community Engagement If we build an online community for our users to share best practices and peer support, we will increase product engagement by 30% and improve retention by 15% over six months, because customers who connect with peers around a shared tool develop stronger product habits and feel accountable to a group.

---

For each candidate, write it in the founder's voice using one of the two formats. Aim for 6-10 distinct candidates. Label them H1 through H[n].

Output a diverge board:

## Diverge board: Hypothesis candidates

| # | Hypothesis | Category |
|---|---|---|
| H1 | [hypothesis statement] | [thought starter category] |
| H2 | [hypothesis statement] | [thought starter category] |
| ... | | |

If mode is with-sherpa, after the diverge board add: "Which hypothesis are you avoiding because it would require the most change? That's probably the most important one."

Step 3: Evaluate phase

Score each hypothesis candidate on three dimensions (1-5 scale):

• Desirability (1-5): Do customers actually want this? Is there evidence of demand?
• Feasibility (1-5): Can you realistically execute this given your team size, stage, and resources?
• Viability (1-5): Does this create a path to sustainable revenue? Does it help the business model? For an internal process on the ops-automation path, score the time or cost it saves instead.

Also assign a Risk Score (1-5): How bad is it if this hypothesis is wrong? (5 = catastrophic, 1 = low stakes)

Output an evaluation table:

## Hypothesis evaluation

| # | Hypothesis | Desirability | Feasibility | Viability | Risk | Total (D+F+V) |
|---|---|---|---|---|---|---|
| H1 | [short label] | 4 | 3 | 4 | 2 | 11 |
| H2 | [short label] | 3 | 5 | 3 | 3 | 11 |
| ... | | | | | | |

Briefly explain scores for each hypothesis (1-2 sentences). Flag any hypothesis scoring below 8 total as unlikely to be worth pursuing.

Sense-check each candidate against the founder's reality:

• Stage (idea): prioritize desirability and feasibility over viability, prove the need first
• Stage (pre-revenue): balance all three, you need a path to paying customers
• Stage (early-revenue): weight viability higher, optimize what's already working

If mode is with-sherpa, challenge low-risk hypotheses: "A risk score of 1 might mean this hypothesis is safe to test, or it might mean you're not testing something that matters. What's the highest-risk hypothesis on this board?"

Step 4: Converge phase

Select the top 2-3 hypotheses based on total score and founder preference. For each selected hypothesis, develop the full dual format:

## Hypothesis card: [H#] - [short name]

### Outcome statement
If we [specific change/experiment], then [target customer] will [measurable outcome] by [specific date].

### Validation commitment
We believe [ICP] will [specific behavior] for [specific outcome]. We will validate this by [specific experiment] within [timeframe].

### Why this one
[2-3 sentences on why this hypothesis is worth testing now, given the competitive landscape and stage]

### Fastest way to be wrong
[1 sentence, what's the fastest experiment that could disprove this?]

### Validation experiment brief
[One paragraph describing the specific experiment: who you'll talk to, what you'll build or test, what signal will confirm or deny the hypothesis]

After all selected hypotheses are developed, ask the founder to pick one primary hypothesis for the prototype session:

"You'll take one of these into the Prototype Launcher. Which hypothesis do you want to test in the next session?"

Step 5: Hand off

When you finish, print the full FOUNDER JAM HANDOFF block (format: founder-setup; if founder-setup is not installed, the same format is in HANDOFF-FORMAT.md next to this file) in a code fence with these fields filled in:

• hypothesis_selected: [winning hypothesis in outcome statement format]
• path: [the path confirmed in the path check, if it was undecided or empty when you received it]

Also fill in any other FOUNDER JAM HANDOFF field you collected during this session, such as mode from the sherpa check or a startup field the block was missing (never add keys the format does not define). Keep every remaining field as you received it, and tell the founder to copy the block and paste it into the next skill. If you are running in Claude Code with file access and a Founder Jam config.json (one with startup_name and path keys) already exists in the working directory, update the same fields there and leave every other key as it is. Never create config.json in another project or overwrite a file that is not a Founder Jam config.

Sherpa-mode questions

When mode is with-sherpa, add these provocative questions at natural moments:

• After diverge board: "Which hypothesis are you avoiding because it would require the most change? That's probably the most important one."
• After scores: "What's the fastest you could know if this hypothesis is wrong?"
• After convergence: "Are you testing a real risk to the business, or confirming something you already believe?"
• Final challenge: "What would it take to prove this wrong in 2 weeks?"

Output summary

When complete:

• Diverge board: all 6-10 candidates (paste-ready table)
• Evaluation table with scores
• Top 2-3 hypothesis cards in dual format
• Winning hypothesis in the updated FOUNDER JAM HANDOFF block

Then say: "Take your winning hypothesis into the Prototype Launcher. Your goal is to build the fastest thing that can confirm or deny it."

Where this runs

Where	What happens
claude.ai or Claude desktop (skill uploaded)	Full flow. Paste your handoff block at the start and copy the updated block at the end. Save any output you want to keep.
Claude Code	Full flow. Can also read and write config.json.

---

Founder Jam curriculum and skills by Drew Burdick · program co-created with Innovate Charlotte · CC BY 4.0 · drewburdick.com/founderjam