Abhay Pratap Mishra

@abhay • 1 months ago

Pressure-Test a Startup Idea and Set Kill Criteria

Stress-test a startup idea: riskiest assumptions ranked, a sceptic's case, cheap tests with pass/fail numbers and dated kill criteria.

At a glance

Best for: Founders and side-project builders deciding whether an idea deserves the next three months, before they write code or quit anything.
Tested on: Claude · Opus 5.5
You fill in: ideafounder_contextrunway
You get: One sentence Independent physio clinicians pay £49/month to finish SOAP notes and exercise sheets without after-hours typing, instead of typing… (full result below)

Prompt

I'm considering this startup idea and I want it pressure-tested, not encouraged.

Idea: {{idea}}
Who I am and what I bring: {{founder_context}}
Time and money I can commit before deciding: {{runway}}

Do this in order:

1. Restate the idea in one sentence as "[customer] pays [price] to [outcome] instead of [current alternative]". If you can't fill a slot, say which one is missing - that is the first problem.

1. List the 5 assumptions the idea depends on, ranked by how likely they are to be false and how fatal that would be. For each, say what evidence would change your mind.

1. Give the strongest case a smart sceptic would make against this idea in under 120 words. No softening.

1. Design the cheapest test for the top 2 assumptions that I can run within my time and money, each with a specific pass/fail number.

1. Write 3 kill criteria: concrete, dated, measurable conditions under which I should stop. Format: "If by [date/milestone] [metric] is below [number], I stop."

Rules: no "it depends" without saying on what. Don't name competitors you're not confident exist. Flag any number you're estimating rather than knowing.

Fill in

{{idea}}: The idea in a few sentences, including who pays and roughly how muche.g. Speech-to-SOAP-notes tool for independent physio clinics, £49/month per clinician
{{founder_context}}: Your skills, network and unfair advantages (or lack of them)e.g. Backend developer, partner runs a 3-clinician physio practice, no sales experience
{{runway}}: How much time and money you'll spend before decidinge.g. 10 weeks of evenings and £1,500

Example input

idea: A tool for independent physiotherapy clinics that turns the physio's spoken session notes into structured SOAP notes plus a home-exercise sheet for the patient, using speech-to-text and an LLM. £49/month per clinician.
founder_context: I'm a backend developer (8 years, Python). My partner is a physio who runs a 3-clinician practice in Leeds. No sales experience.
runway: 10 weeks of evenings and £1,500

9790

Generated Outputs (1)

3 hours ago

Claude

Opus 5.5

3 hours ago

Generated Output

1. One sentence
Independent physio clinicians pay £49/month to finish SOAP notes and exercise sheets without after-hours typing, instead of typing notes into their practice software and building exercise sheets in a separate exercise-prescription tool or by hand.

Every slot fills, but "instead of" is crowded: the alternative isn't nothing, it's software they already pay for plus general-purpose AI medical scribes.

2. Assumptions, riskiest first

1. Clinicians will pay a separate £49/month rather than use a general AI scribe or wait for their practice software to add one. Likely false; fatal. Changes my mind: 5 clinicians outside your network paying.

1. Clinic owners will accept patient audio going to a third-party AI. UK GDPR treats this as health data, so you'll need a DPIA and patient consent. Medium; fatal to sales. Evidence: owners saying yes after reading your data-handling summary.

1. Drafts are accurate enough that editing is much faster than typing. Medium; fatal. Evidence: timed trials. (I'm estimating typing at ~10 minutes a session; measure it.)

1. Notes plus exercise sheet together beat either alone. Medium; survivable.

1. You can reach clinics beyond your partner's cheaply. Likely false; survivable but slow.

3. The sceptic's case
"AI scribes for clinicians already exist, and practice-management platforms have every reason to build this in. You'd be a thin layer over the same speech and language models, selling health-data processing to cautious small clinics, with no sales experience. Your partner gives you insight into one clinic, not distribution. The likely outcome is a nice tool three friendly clinics use for free."

4. Cheapest tests

• Accuracy (weeks 1-3, ~£100 in API costs): build a rough version for your partner's 3 clinicians and run 30 real sessions with patient consent. Pass: median edit time under 3 minutes, and clinicians still choose to use it in week 3 without being asked.

• Willingness to pay (weeks 3-8, up to £400): a landing page with a one-page data-protection summary, posts in UK physio communities, and direct emails to 60 clinic owners. Offer a founding price of £29/month, paid 3 months upfront. Pass: 5 paying clinicians outside your network.

5. Kill criteria

• If by the end of week 3 median edit time on real sessions is above 5 minutes, I stop.

• If by the end of week 8 fewer than 3 clinicians outside our network have paid, I stop.

• If by week 6 two or more owners reject it over data protection after reading our summary, I pause and pay for specialist advice before spending anything else.

Why it works

The "[customer] pays [price] to [outcome] instead of [alternative]" sentence forces the idea to name its real competitor, which is usually existing habits and tools rather than nothing. Ranking assumptions by likelihood of being false × how fatal stops the model listing generic risks in random order. The sceptic's case has a word cap and "no softening", so it reads as an argument rather than a caveat. Most important, the kill criteria template turns a vague "see how it goes" into dated, numeric stop rules you can hold yourself to, which is what founders most often lack.

When not to use it

Not for ideas you've already validated with paying customers; at that stage you need a growth plan, not a kill test. The model can't know your local market, so treat its competitor and pricing remarks as leads to check. Regulated fields (health, finance, legal) need real specialist advice on compliance; the prompt only flags the risk.

Comments (0)

No comments yet. Be the first to share your thoughts!