Founder talking to users

Purpose

Create a continuous founder-to-user learning system that recovers facts, observes behavior, earns commitment, and changes product or commercial decisions.

Give the founder a decision, the evidence behind it, the strongest contrary case, and the smallest executable next step. Do not produce generic encouragement or a long menu of tactics.

When to Trigger

Use this skill according to the frontmatter description. If the stated issue is a symptom of an earlier broken link, say so and route to founder-debugger or the owning retained skill.

Inputs Needed

• Decision the research must change
• Target segment and recruiting criteria
• Known facts, assumptions, and disputed beliefs
• Recent product events, support tickets, churn, and sales evidence
• Interview candidates across prospects, active users, retained users, failed activations, churned users, and lost buyers
• Interview window and number of conversations
• Artifacts to request: screenshots, spreadsheets, messages, recordings, invoices, or workflow outputs

Ask only for missing inputs capable of changing the decision. For reversible actions, make assumptions explicit and propose a bounded test instead of blocking on perfect data.

Questions to Ask

• Tell me about the last time this happened.
• What triggered it?
• Walk me through what you did, step by step.
• What was hardest, slowest, riskiest, or most expensive?
• What did you try before, and why did it fail?
• Who else was involved and who could approve change?
• What happens if you do nothing?
• What have you already spent in money, time, workflow change, or reputation?
• Show me the artifact or the actual workflow.
• Why did you start, continue, stop, buy, or refuse?

Mental Models

• Past behavior over future intention.
• Specific episodes over general opinions.
• Observation over recollection.
• Problem interview before solution pitch.
• Commitment ladder: time, access, data, workflow, reputation, money.
• Outliers reveal mechanisms; patterns justify decisions.
• Feature requests are compressed jobs.
• Research expires as product and market change.

Use these models as competing lenses. Select the one that best explains the observed behavior, state what evidence would falsify it, and convert it into a decision rather than repeating it as a slogan.

YC Principles

• Founders must talk to users themselves before delegating the learning loop.
• Do not ask whether an idea is good; recover the user's life and behavior.
• Avoid pitching during discovery because politeness corrupts evidence.
• Speak with the narrow target user, not convenient friends or generic respondents.
• Use conversations to recruit first customers and make early users extremely happy.
• Return to users after shipping and observe whether behavior changed.

Paul Graham Principles

• The best startup ideas begin with real problems, often the founders' own or ones they understand deeply.
• Recruit early users manually and over-serve them until the product reflects what they need.
• What founders learned from users is a stronger test than how many conversations they claim to have had.

Garry Tan Principles

• Direct customer contact develops earned product judgment and craft.
• Use transcription and AI synthesis to increase recall, but preserve raw evidence and founder interpretation.
• Never let synthetic personas, generated interviews, or survey summaries substitute for real people doing real work.

Decision Frameworks

Interview evidence ladder

1. Strongest: observed completed behavior, purchase, renewal, churn, or switched workflow.
2. Strong: a recent detailed episode supported by an artifact.
3. Moderate: repeated testimony from correctly selected users.
4. Weak: opinion, preference, feature vote, or hypothetical intent.
5. Noise: compliments, encouragement, demographic generalization, and feedback from non-target users.

Conversation modes

1. Discovery: understand problem, workflow, frequency, severity, and alternatives.
2. Usability: observe a realistic task without teaching the interface.
3. Churn: reconstruct expected value, failure point, and switching decision.
4. Sales discovery: connect pain, authority, timing, budget, and consequence.
5. Feature request: uncover the job and why the current path fails.
6. Retention: understand the recurring trigger and value loop of best users.

Synthesis rule

1. Tag evidence by segment and lifecycle state.
2. Separate quote, observation, inference, and decision.
3. Count mechanisms, not just mentions.
4. Preserve contradictory evidence.
5. End with what changed, what remains uncertain, and the next test.

Step-by-Step Process

1. Write one decision and the beliefs that would change it.
2. Choose participants capable of supplying the required evidence.
3. Recruit founders' users directly; avoid professional respondents when possible.
4. Prepare five core prompts about recent behavior, not a long script.
5. Open with context and permission; do not pitch.
6. Probe one concrete episode until triggers, steps, people, alternatives, and consequences are clear.
7. Ask to see the workflow or artifact.
8. Request the next appropriate commitment.
9. Debrief immediately into facts, quotes, interpretations, contradictions, and decisions.
10. After five interviews, inspect patterns and sampling gaps; continue only if marginal learning remains high.
11. Ship or test the consequence, then return to the same segment.

Checklists

Before

• Name the decision.
• Recruit the correct segment and lifecycle state.
• Remove leading and hypothetical questions.
• Plan which behavior or artifact to observe.
• Assign interviewer and note-taker when possible.

After

• Record exact quotes and observed actions.
• Tag segment and evidence strength.
• Separate request from underlying job.
• Document contrary cases.
• Make or schedule a decision; do not create a research archive with no consequence.

Red Flags

• Friends, investors, or founders are treated as customer evidence.
• The founder talks more than the participant.
• Questions contain the desired answer.
• The conversation centers on the proposed feature instead of the user's workflow.
• Participants describe what they would do someday.
• No artifacts or behavior are inspected.
• All evidence is blended across incompatible segments.
• AI summaries discard raw quotes or contradictions.
• The company conducts research but cannot name a changed decision.

Common Mistakes

• Selling during discovery.
• Asking users to design the product.
• Counting mentions instead of severity and commitment.
• Ignoring failed and churned users.
• Treating silence as lack of a problem instead of poor recruiting or questioning.
• Over-scripting and missing unexpected mechanisms.
• Delegating all contact to product, sales, or research teams too early.

Metrics

• Decision-relevant interviews per founder
• Share with observed behavior or artifacts
• Evidence-bearing insights per interview
• Recruit-to-interview conversion by segment
• Time from interview to product or commercial test
• Assumptions confirmed, rejected, or narrowed
• Commitments earned
• Repeated mechanisms by segment
• Decisions changed by evidence
• Post-change behavior of the interviewed cohort

For every metric, define the unit, numerator, denominator, cohort, segment, cadence, source event, and owner. Prefer decision thresholds and cohort movement over universal benchmarks.

Example Scenarios

Scenario 1

Users request dashboards: reconstruct the last decision they could not make, the data they assembled manually, and the consequence; build the decision path, not a generic dashboard.

Scenario 2

Prospects praise a prototype but will not pilot: stop demoing, examine urgency, current workaround, authority, and implementation cost; compliments are not demand.

Scenario 3

Churned users say price was too high: inspect activation and realized value first; 'price' often compresses weak fit, missing outcome, or lost trust.

AI Prompt Templates

Template 1

Create a 30-minute interview guide for this decision and target segment. Use past-behavior questions, artifact requests, neutral probes, and one appropriate commitment ask.
End with a decision, owner, deadline, metric, threshold, stop rule, and strongest contrary case.

Template 2

Analyze these raw interviews. Separate facts, quotes, observations, inferences, contradictions, segment patterns, and decisions. Do not turn feature votes into a roadmap.
End with a decision, owner, deadline, metric, threshold, stop rule, and strongest contrary case.

Template 3

Role-play a user interview evaluator: flag every leading, hypothetical, double-barreled, or pitch-like question and replace it with a behavioral question.
End with a decision, owner, deadline, metric, threshold, stop rule, and strongest contrary case.

Related Skills

• founder-mvp
• founder-retention
• founder-pricing
• founder-distribution
• founder-debugger

Further Reading

1. How to Talk to Users — The core YC interview method: target the right people, recover past behavior, avoid pitching, and use commitment as evidence.
2. How to Get Your First Customers — Turns user conversations into first sales through narrow targeting and founder-led outreach.
3. Building Products Users Love — Shows how support and continuous user contact develop product judgment.
4. What I've Learned from Users — Tests whether founders are paying attention by asking what, specifically, they have learned from users.
5. Do Things That Don't Scale — Manual recruiting, over-serving early users, and direct observation as the path to discovering what should later scale.
6. Bringing Numbers to Life Through Qualitative User Research — Combines quantitative anomalies with qualitative research to expose mechanisms and bias.

Source Links

• How to Talk to Users — Gustaf Alströmer. The core YC interview method: target the right people, recover past behavior, avoid pitching, and use commitment as evidence.
• How to Get Your First Customers — Gustaf Alströmer. Turns user conversations into first sales through narrow targeting and founder-led outreach.
• Building Products Users Love — Kevin Hale. Shows how support and continuous user contact develop product judgment.
• What I've Learned from Users — Paul Graham. Tests whether founders are paying attention by asking what, specifically, they have learned from users.
• Do Things That Don't Scale — Paul Graham. Manual recruiting, over-serving early users, and direct observation as the path to discovering what should later scale.
• Bringing Numbers to Life Through Qualitative User Research — Intercom. Combines quantitative anomalies with qualitative research to expose mechanisms and bias.
• Have You Tried Talking to Your Customers? — Intercom. A concrete example of reserving substantial founder time for customer conversations.
• Lyft's Monal Chokshi on Lightweight User Research — Intercom. Audience selection, non-leading questions, listening discipline, and non-verbal evidence.
• Understanding Your Customer Is Key to Product Judgment — Intercom. Explains why judgment is domain-specific, perishable, and built through direct customer experience.
• YC's Essential Startup Advice — Y Combinator. The early-stage order of operations: launch, talk to users, iterate, and delay scaling until people want the product.