TEST YOUR STARTUP IDEA IN ONE DAY

Turn a promising idea into a focused customer experiment.

Follow seven steps to define your customer, investigate demand, test an offer and choose the smallest useful MVP.

Finish with evidence, unresolved questions and a clear next action. One day can start validation; confidence grows through repeated tests.

HOW TO READ THIS PLAYBOOK

Foundation: primary sources from startup practitioners and AI researchers.

Method: an original synthesis of customer discovery and experiment design.

Prompts, schedules, thresholds and the café example are original teaching tools. They are not proven benchmarks or actual customer findings.

Source IDs refer to the reference pages at the end.

WHAT ONE DAY CAN ESTABLISH

DEFINE THE DECISION

A useful first-day question is specific: can I find reachable buyers with a recent problem who will take a meaningful next step?

Choose one uncertain assumption that could make the idea fail. Design the day around testing it.

A complete market assessment, repeat purchasing and a scalable acquisition model require more observation.

THE ROOTS OF THE METHOD

Steve Blank emphasizes learning directly from customers. Eric Ries organizes progress around experiments and validated learning. Strategyzer makes assumptions, tests, measures and success criteria explicit.

These approaches offer practical guidance. Their use does not establish a universal success rate or guarantee that a startup will work.

THREE DIFFERENT VALIDATION QUESTIONS

Problem: does a specific customer experience a meaningful difficulty?

Solution: does your proposed approach help them achieve the desired outcome?

Business: will enough customers pay, return and be served economically?

Evidence for one question leaves the others open. A paid pilot still leaves retention and scalable delivery uncertain.

ASSIGN AI A CLEAR ROLE

Use AI to organize sources, compare alternatives, draft interview guides and identify gaps in your reasoning.

Keep a human responsible for source checks, customer conversations and decisions.

Treat generated personas, estimated budgets and simulated customer reactions as hypotheses until supported by observations.

WATCH FOR AGREEABLE ANSWERS

Anthropic researchers reported sycophancy in the language models they tested: responses could align with a user’s beliefs over truth.

This study concerns particular models and evaluations, not every response from every model.

Practical implication: ask for opposing evidence and inspect the underlying sources before trusting a positive assessment.

SET THE RESEARCH RULES

Use this instruction:

Separate sourced facts, interpretations and assumptions. Link each factual claim to an accessible source and date. Mark unavailable information as unknown. Seek contradictory evidence. Never invent customer quotes, prices or interviews. Explain what real-world test could change the conclusion.

STEP 1: DEFINE YOUR CUSTOMER

CHOOSE A REACHABLE FIRST SEGMENT

Describe a group sharing the same job, operating context and buying process. Include who uses the solution, who pays and who can approve a purchase.

Start narrow enough to locate people today.

Demographics matter when they affect the problem or purchase. For business customers, workflow, company size and decision authority may be more useful.

PROMPT: COMPARE THREE CUSTOMER SEGMENTS

My idea is [idea] in [location]. Propose three customer segments. For each, list the job, user, payer, current workaround, buying trigger and a way to reach them. Label assumptions and cite evidence. Rank by observed pain and practical access. Explain missing information and choose one segment for today’s interviews.

EXAMPLE: A CAFÉ CONTENT SERVICE

Broad audience: small businesses needing content.

Initial segment: owner-operated cafés in one district that already post promotions but struggle to prepare them consistently.

User and buyer: the café owner.

Possible trigger: an upcoming menu launch.

Verify these conditions with actual owners before treating this segment as promising.

DEFINE WHO QUALIFIES

For the café example, ask whether the person manages promotions, has prepared one recently and can approve spending. Record how you found them.

Recruit people whose behavior fits your question. Include users of competing solutions and people who declined previous offers.

A convenient group of supportive friends can distort the findings.

STEP 2: FIND A PROBLEM WORTH SOLVING

LOOK FOR RECENT CONSEQUENCES

Describe the last occurrence, how often it happens and its concrete consequence. Look for time spent, missed opportunities, avoidable errors or money already committed to workarounds.

A complaint becomes more informative when linked to behavior.

Keep willingness to pay unresolved until a buyer encounters a specific offer.

PROMPT: RANK THE PROBLEM HYPOTHESES

For [segment], analyze these notes and sources: [inputs]. List recurring problems with frequency, consequences and current workarounds. Reference the supporting note or URL for each. Separate observed behavior from speculation. Rank the problems and explain uncertainty. Identify the single assumption most important to test next.

INTERVIEW AROUND A REAL EVENT

Ask the person to walk through the most recent incident. Explore what made it difficult, what they tried and what happened afterward.

Use follow-ups about steps, costs and trade-offs. Leave room for unexpected answers.

YC’s Eric Migicovsky recommends specific past experiences and listening carefully during early user interviews.

EXAMPLE: MAKE THE PROBLEM MEASURABLE

Possible finding: an owner spends three hours assembling each promotion and still publishes after the intended date.

Follow up: which tasks consumed the time? What was delayed? What has the owner already tried?

Record the actual response. Do not assume that publishing late reduced sales unless there is evidence for that connection.

KEEP A SIMPLE EVIDENCE LOG

Record:
1. Participant code and segment.
2. Date and recruitment source.
3. Event or behavior described.
4. Supporting note or document.
5. Your interpretation.
6. Alternative explanation.
7. Follow-up needed.

Preserve exact words separately from summaries. Remove identifying or sensitive details before sharing notes with AI.

STEP 3: ANALYZE THE ALTERNATIVES

MAP WHAT CUSTOMERS ALREADY DO

Include direct products, agencies, freelancers, spreadsheets, manual effort and postponing the task.

For each alternative, ask what customers gain, what they sacrifice and why they keep using it.

Switching may require training, trust, migration or a new budget. A useful feature alone may leave those barriers unresolved.

PROMPT: BUILD A COMPETITOR COMPARISON

Research alternatives for [job, segment, location]. Compare target user, promised outcome, pricing unit, limits, setup effort and verified complaints. Use official pages for product facts. Date every price check and link sources. Include manual work and doing nothing. Mark missing data and propose gaps that customer interviews should test.

READ COMPLAINTS IN CONTEXT

A review records one person’s account. Check its date, product version, customer type and whether the issue appears elsewhere.

Repeated complaints can suggest a research direction, but review samples may overrepresent unusually happy or unhappy users.

Verify advertised features with vendors and ask customers whether the claimed gap affects their decisions.

ESTIMATE AN ACCESSIBLE OPPORTUNITY

Illustration only:

200 reachable cafés × 10% paying adoption × HKD 500 per month = HKD 10,000 monthly revenue.

All three inputs are assumptions. Show low and high scenarios and test access, conversion and price separately.

This estimates one initial segment under chosen assumptions. It does not establish total market demand.

STEP 4: DISCOVER WHAT PEOPLE ACTUALLY SAY

COLLECT TRACEABLE CUSTOMER LANGUAGE

Search relevant communities, reviews and discussion threads. Save a URL, date and short note explaining the context.

Look for the problem, attempted solution and consequence. Count independent people rather than duplicated posts.

Then ask qualified customers whether those patterns match their own recent experience.

PROMPT: ANALYZE CUSTOMER EVIDENCE

Analyze only these accessible discussions and interview notes: [inputs]. Group recurring needs, workarounds and objections. For each theme, give participant or source IDs and counterexamples. Keep exact excerpts separate from paraphrases. Explain sampling limits. If a source is inaccessible, say so. Never fill gaps with invented customer language.

LOOK FOR REASONS THE IDEA COULD FAIL

Ask what customers value about their current approach and what would make switching unacceptable.

For the café service, possible obstacles include brand control, unclear sales impact or a preference to create content personally. These are hypotheses.

Actively recruit people likely to disagree and document the conditions behind their answers.

PROMPT: AUDIT YOUR RESEARCH

Review [evidence log]. Identify unsupported claims, leading questions, duplicated sources and customers outside the target segment. Give the strongest alternative explanation for each positive signal. List what remains unknown. Recommend one follow-up observation that would most improve the decision. Do not invent confidence percentages.

STEP 5: TEST WILLINGNESS TO PAY

TURN THE IDEA INTO A SPECIFIC OFFER

Present one understandable offer with a clear deliverable, price, timeline and limits. Ask the person who controls the budget for a concrete next step.

A price discussion helps you learn objections. An actual purchase provides stronger evidence about that offer.

Record discounts, personal relationships and other conditions that may affect the response.

PROMPT: DESIGN THE PRICING TEST

For [customer and outcome], compare subscription, one-time and usage-based pricing. Discuss freemium only if free use has a credible route to payment. Use sourced alternatives where available. Choose one offer to test with a buyer. Define scope, price, objections, a concrete commitment and what the result would leave unresolved.

CHOOSE THE COMMITMENT YOU CAN OBSERVE

Interest: someone reads, likes or compliments the offer.

Intent: a qualified prospect books a relevant conversation or shares required information.

Purchase: a buyer pays for a defined deliverable.

Continued value: the buyer uses it and returns or renews.

Interpret each action in context. Refunds, incentives and personal favors can weaken the signal.

WRITE THE TEST BEFORE RUNNING IT

Specify four things:

Hypothesis: what you believe.
Test: what you will do.
Measure: what behavior you will count.
Threshold: the result that determines your next action.

Also record the audience and time window. Choose criteria before seeing results so you can assess the experiment consistently.

EXAMPLE: A SMALL PAID PILOT

Offer: one week of café promotional content for HKD 500, with a stated delivery date and revision limit.

Invite 10 qualified owners. Preliminary criterion: two pay within seven days.

Use that result to decide whether to run a larger pilot. Two purchases would provide early evidence for this offer within this sample; general demand remains uncertain.

MEASURE THE WHOLE FUNNEL

Track qualified people contacted, offers actually seen, replies, purchases and cancellations separately.

Example: 2 purchases from 10 qualified offer viewers = 20% observed conversion. With such a small sample, uncertainty is substantial.

If you cannot verify that people saw the offer, report that limitation. Avoid presenting outreach response as purchase conversion.

KEEP EARLY OFFERS CLEAR

State what is available now, what requires manual work and when delivery can happen. If accepting advance payment, explain the scope and cancellation terms before the customer commits.

Record reasons for declining. A rejection may concern timing, trust, scope or budget. Use follow-up questions to understand which explanation the evidence supports.

STEP 6: FIND YOUR ADVANTAGE

SEPARATE POSITIONING FROM DEFENSIBILITY

Positioning explains the customer, outcome and reason to choose your offer.

A defensible advantage concerns what competitors would struggle to reproduce, such as trusted access, hard-earned expertise or a resource you control.

Treat claimed advantages as assumptions. Look for evidence that they improve acquisition, delivery or customer outcomes.

PROMPT: IDENTIFY A CREDIBLE STARTING ADVANTAGE

Using [evidence] and my assets [experience, relationships, distribution, resources], propose one focused positioning statement. Explain why buyers might care and what competitors could copy. Identify any advantage supported by evidence. Design a test of whether it improves access, conversion or outcomes. Mark every speculative claim.

EXAMPLE: TEST ACCESS AND EXPERTISE

A founder with café operating experience and direct relationships with local owners may reach interviews faster and understand production constraints.

Test that proposition: track access, pilot sales and delivery quality.

The same relationships can bias feedback, so include unfamiliar owners in later rounds. Technical tooling alone may be easy for others to adopt.

STEP 7: BUILD THE SMALLEST USEFUL MVP

CHOOSE WHAT THE MVP MUST TEACH YOU

An MVP is shaped by the uncertainty you need to resolve. Define the customer action and outcome you must observe.

A manual service can test delivery. A clickable prototype can explore usability. A landing page can test response to an offer.

Each has limits: a signup leaves actual usage, satisfaction and repeat purchasing untested.

"DO THINGS THAT DON'T SCALE."

START WITH A MANUAL DELIVERY LOOP

Paul Graham describes founders recruiting users individually and providing unusually attentive early service.

For the café example, manually collect the brief, prepare content and review it with the owner. Observe where the process breaks.

Record delivery time and outcomes so later automation addresses a demonstrated bottleneck. The quote appears in S5.

PROMPT: SPECIFY THE SMALLEST MVP

From [evidence and unresolved assumption], propose the smallest MVP to test [behavior]. Prefer existing tools and manual delivery when suitable. Specify one customer journey, essential functions, excluded work, event tracking and success criteria. Estimate a two-week plan with dependencies. Identify what needs a separate feasibility test.

EXAMPLE: ONE COMPLETE CUSTOMER JOURNEY

Owner submits a brief and approved brand assets.

Founder prepares the agreed content, checks accuracy and delivers it for review.

Owner approves or requests the included revision, then decides whether to publish.

Measure turnaround, corrections, actual usage and willingness to buy again. Keep the first cycle small enough to observe closely.

CHECK THE DELIVERY ECONOMICS

Example: HKD 500 pilot revenue minus HKD 100 tools and fees minus 2 hours of labor valued at HKD 150/hour = HKD 100 contribution.

This excludes acquisition, fixed overhead and other costs.

Track real delivery time, revisions and support. Early revenue can coexist with economics that require substantial improvement.

KNOW WHEN A LONGER TEST IS NECESSARY

Enterprise purchasing may require several approvals. Hardware may depend on technical performance. Seasonal demand needs an appropriate observation window.

Use day one to arrange access and define the critical test.

Customer enthusiasm leaves feasibility unresolved. A working prototype leaves demand unresolved. Record progress against each question separately.

RUN THE SPRINT AND DECIDE

PREPARE BEFORE THE CLOCK STARTS

Book potential interview times if possible. Gather available notes, create an evidence log and define your spending limit.

If you have no customer access yet, prioritize recruitment during the sprint and schedule conversations for later.

The one-day format describes focused working time. Customer response and purchasing cycles may extend beyond it.

THE FIRST FOUR HOURS

Hour 1: define the segment and riskiest assumption.

Hour 2: investigate alternatives and source quality.

Hour 3: collect traceable customer signals and prepare interviews.

Hour 4: interview available prospects or recruit them.

Capture evidence as you go. Move planned work when access or new findings change the priority.

THE NEXT FOUR HOURS

Hour 5: review patterns and contradictory evidence.

Hour 6: define one offer and its test criteria.

Hour 7: prepare or launch the experiment with reachable prospects.

Hour 8: document what you learned and schedule follow-up.

Keep response windows realistic. A test launched today can have a decision date next week.

USE FOUR POSSIBLE DECISIONS

Continue: evidence supports investing in the next test.

Revise: a pattern points toward changing the segment, problem or offer.

Stop: a critical assumption fails under a credible test and no useful revision is apparent.

Inconclusive: access, exposure, sample size or timing prevents a useful judgment.

Name the evidence behind your choice.

READ MIXED RESULTS CAREFULLY

Pain with no purchases: inspect urgency, buyer authority, trust and the offer.

Purchases with little usage: examine whether the promised result is delivered.

Usage with no renewal: investigate continuing need and value.

No responses: check reach and timing before judging demand.

These are diagnostic questions; interviews and behavior determine the answer.

PROMPT: PRODUCE THE DECISION MEMO

Using only [evidence log and results], summarize the segment, problem, offer, exposure, commitments and contradictions. Separate observed facts from interpretations. Recommend continue, revise, stop or inconclusive. Explain the remaining risks. Specify the next test, owner, deadline, cost limit and decision criterion. Avoid unsupported success probabilities.

THE DELIVERABLES CHECKLIST

1. One defined customer segment.
2. One prioritized problem hypothesis.
3. A sourced alternatives comparison.
4. An evidence log with contradictions.
5. A specific offer and experiment.
6. A small MVP scope if justified.
7. A decision memo with follow-up dates.

Keep missing evidence visible. The checklist records your work; it does not certify product-market fit.

TURN THE SPRINT INTO A WORKFLOW

Use this WorkFlow prompt:

Help me test [startup idea] for [customer] in [location]. Break the work into research, interviews, an offer test and an evidence review. Give each WorkStep an owner, deliverable, deadline and decision criterion. Mark assumptions and dependencies.

Save this StackSlide and update it with your actual findings.

SOURCES AND RESEARCH NOTES

S1: CUSTOMER DISCOVERY

Ardent 2: Get Out of My Building, October 8, 2009.

Supports direct observation and customer contact as a way to challenge internal assumptions.

https://steveblank.com/2009/10/08/get-out-of-my-building/

Historical practitioner guidance. This source does not establish a one-day validation guarantee.

S2: LEAN STARTUP

The Lean Startup: Methodology.

Explains validated learning, MVPs and the build-measure-learn cycle.

https://theleanstartup.com/principles

Supports iterative experiments. The time limits and sample criteria in this playbook are original planning choices.

S3: EXPERIMENT DESIGN

Validate Your Ideas with the Test Card, March 5, 2015.

Specifies the hypothesis, test, measurement and success threshold. Also emphasizes multiple experiments.

https://www.strategyzer.com/library/validate-your-ideas-with-the-test-card

Primary practitioner source for the test-design structure.

S4: CUSTOMER INTERVIEWS

How to Talk to Users, July 25, 2019. Official YC lecture.

Guidance on recent experiences, attempted solutions and listening during customer interviews. Discusses principles from The Mom Test.

https://www.youtube.com/watch?v=MT4Ig2uqjTc

The interview wording in this playbook is adapted.

S5: EARLY MANUAL WORK

Do Things That Don’t Scale, July 2013.

https://www.paulgraham.com/ds.html

Source of the earlier quotation and guidance on manual recruitment and early customer service.

Practitioner examples support the approach; they do not prove that every startup should use the same experiment.

"HOW DOES SUCCESS LOOK LIKE? WHAT'S THE THRESHOLD?"

QUOTE CONTEXT

The preceding quotation is Alex Osterwalder’s fourth Test Card question. It asks experiment designers to make the success threshold explicit.

The sentence is preserved as published.

A threshold should guide the next decision and reflect the test’s limits. This playbook’s sample threshold is illustrative.

S6: AI RELIABILITY

Towards Understanding Sycophancy in Language Models, October 23, 2023.

https://www.anthropic.com/research/towards-understanding-sycophancy-in-language-models

Reports sycophancy in tested models. Supports checking AI judgments against evidence. It does not quantify the success of AI-assisted startup validation.

ALL SOURCES AND ATTRIBUTION

S1: Steve Blank, Ardent 2 (2009).
S2: Eric Ries, Lean Startup Methodology.
S3: Alex Osterwalder, Test Card (2015).
S4: Eric Migicovsky, YC lecture (2019).
S5: Paul Graham, Do Things That Don’t Scale (2013).
S6: Anthropic, Sycophancy (2023).

Links appear on the source pages.
Inspiration: supplied @founderjuice screenshots, seven steps. Prompts and examples rewritten and expanded.