Written by Simon, founder who shipped 4 products nobody wanted.

Startup Idea Validation Without the Hype: A Founder's Guide to Stress-Testing Before Code or Capital

Ninety percent of startups fail. You've heard that stat so many times it's stopped meaning anything. Here's the part that should sting: most of them fail not because the founders were lazy or incompetent, but because they built something nobody wanted badly enough to pay for. They skipped startup idea validation and went straight to execution. They confused enthusiasm for evidence.

I've done it four times. Four products shipped, four quiet deaths. The pattern was always the same: a compelling idea, a few friends who said "that sounds cool", and a sprint to build before thinking. The cost wasn't just money. It was six to twelve months of compounding assumptions, each one stacking on the last, until the whole thing collapsed under the weight of a customer base that never materialized. If you want to avoid that, validate your idea before you write a single line of code or pitch a single investor.

This guide covers the frameworks, experiments and decision tools that actually work. Not theory borrowed from a business school textbook. Tested approaches you can run in 90 days on a minimal budget.

Why Assumptions Kill Startups (Before the Market Gets a Chance)

Every startup begins as a bundle of assumptions. You assume a problem exists. You assume people want it solved. You assume they'll pay what you need them to pay. You assume you can reach them at a cost that makes the math work. Most founders treat these assumptions as facts and start building on top of them. That's how you end up six months in with a polished product and zero customers.

The riskiest assumption is almost never your idea itself. It's the layer underneath: the mechanism by which the idea becomes a business. A freelance tax tool sounds sensible until you discover the problem is seasonal (January through April only) and nobody wants a $99/month SaaS subscription for a four-month pain point. The idea was fine. The business model assumption was the killer. Identifying your riskiest assumption, not your favourite feature, is where validation starts.

The cost of assumption-driven development compounds fast. A landing page costs $50 and a weekend. A full product costs $50,000 and six months. The difference is enormous, and the only thing separating them is willingness to test before you build.

Section 1: First-Principles Stress-Testing

The Five Questions That Filter 80% of Bad Ideas

Before any framework or tool, you need to answer five questions honestly. Does the problem actually exist in a way that causes people genuine pain? Who experiences it most acutely, and are those people reachable? How are they solving it today, because if the current solution is "good enough", your job is much harder than you think. Why would they switch, given that switching has friction and cost? Can you reach them at an acquisition cost that supports your unit economics?

These aren't rhetorical exercises. Write your answers down. Show them to someone who will argue with you. The places where your answers get vague or circular are exactly where your riskiest assumptions live. If you can't clearly describe how someone solves this problem today without mentioning your product, you don't yet understand the problem well enough to build a solution.

ChatGPT is genuinely useful here as a sparring partner. Not for research in the traditional sense, but for adversarial questioning. Prompt it with: "Here is my startup idea. Argue against it using first-principles logic and identify every assumption I'm making." The output won't be perfect, but it surfaces blind spots you've been unconsciously protecting. Follow that with persona development prompts: describe your target customer in detail and ask it to map what a typical day looks like, where they feel friction and what alternatives they've already tried. You're not outsourcing thinking to the AI. You're using it to pressure-test thinking you've already done.

Section 2: Validation Frameworks That Actually Work

The Lean Startup Loop Applied Before Product

Eric Ries's Build-Measure-Learn cycle is typically described as a product development loop. But the most valuable application is pre-product. Before you build anything, you run minimum viable tests (MVTs) rather than minimum viable products. An MVT is not a stripped-down product. It's an experiment designed to test one assumption with the minimum possible investment. A landing page that explains a product and captures emails is an MVT. A fake door test that shows a "Buy Now" button and records click-through rate before anything exists is an MVT. The goal is signal, not product.

The metrics that matter at this stage are brutally simple. What percentage of people you speak with confirm the problem exists (target: 40% or higher from cold outreach)? What percentage of landing page visitors convert to email signups (target: 10% or higher)? What percentage of pilot users commit actual money (target: 30% or higher)? Everything else is noise. Social shares, Twitter followers, podcast mentions. Noise.

Jobs-to-be-Done: What People Are Actually Hiring Your Product For

The Jobs-to-be-Done framework, developed by Clayton Christensen and Bob Moesta, reframes the question from "who is my customer" to "what job is my customer hiring a product to do". It sounds subtle. It isn't. People don't buy project management software because they want project management software. They hire it to reduce the anxiety of not knowing what their team is doing. That emotional and functional job description changes everything about how you position, price and build.

Structured JTBD interviews follow a specific pattern. You ask about the last time they experienced the problem, not hypothetically but specifically. You ask what triggered the search for a solution, what they tried first and why it fell short. You ask what a good outcome actually looks like in their life or workflow. The answers reveal the real job, which is almost never what you assumed going in. ChatGPT can help you build these interview guides: describe your target customer and the problem you think they have, and ask it to generate a 15-question JTBD interview script. Then edit it yourself, because the AI will produce generic questions you'll need to sharpen with context.

Design Thinking and Bill Aulet's Disciplined Entrepreneurship

Design thinking adds empathy mapping to the validation toolkit. Before you prototype anything, you map what your target customer thinks, feels, says and does in the context of the problem. The gaps between what they say and what they do are where the real insights live. People say they want to eat healthier. They do not eat healthier. That gap is a business opportunity or a business graveyard, depending on which side of it your product sits on.

Bill Aulet's Disciplined Entrepreneurship framework, particularly the first six of his 24 steps, is the most practical structured approach to early validation I've found. It forces market segmentation before solution design, beachhead market selection before scaling plans and end-user definition before revenue modeling. The key discipline is narrowing. Most founders want to say their market is "everyone who has this problem". Aulet forces you to pick one specific beachhead segment and validate there first. If you can't win a small, specific market, you cannot win a large, diffuse one.

Section 3: The 90-Day Validation Roadmap

Month 1: Problem Validation

The first month has one job: confirm the problem is real and painful enough to act on. You do this through two parallel tracks. First, build a simple landing page (Carrd or Webflow, nothing fancy) that describes the problem and asks visitors to sign up for early access. Drive 200-500 visitors to it through direct outreach, community posting or a small paid traffic test. If you're getting below 5% email conversion, your problem statement isn't resonating. Rewrite it and retest.

Second, run a customer interview sprint. Fifteen to twenty conversations in four weeks is achievable. You're not selling anything. You're asking about their life, their current frustrations and what they've already tried. Use ChatGPT to synthesize your notes after each batch of five interviews: paste anonymized summaries and ask it to identify recurring themes and contradictions. Look for the 40% threshold: if fewer than 40% of people you speak with confirm the problem exists without you leading them to it, the problem may not be painful enough to build a business on.

Month 2: Solution Validation

Month two tests whether people want your specific solution, not just whether the problem exists. The most powerful experiment here is the presale: selling access to a product that doesn't yet exist. This is not fraud. It's the most honest signal you can get. Real money from real people who've heard your pitch and still pulled out a credit card. First Round Capital's research on unconventional validation tactics consistently highlights presale tests as the highest-quality signal founders can generate before building.

Alongside presales, run A/B tests on your sales page copy, test two or three price points through survey tools like Typeform and build toward a waitlist of 100+ engaged signups. "Engaged" means they've responded to at least one follow-up email, not just signed up and gone quiet. The target conversion from visitor to email signup is 10%. If you're hitting that, you have a signal worth acting on. You should also be refining your pricing sensitivity work here: ask directly what they currently pay to solve this problem and what they'd pay for your solution. People round up when asked about hypothetical spending, so discount their stated willingness to pay by 30-40% to get a realistic baseline.

Month 3: Market Validation and the Go/No-Go Decision

The third month is about commitment from real customers and a hard decision. Recruit five to ten pilot customers. These people engage with a manual, non-automated version of your product: a spreadsheet workflow, a done-for-you service, a Notion template with you behind the scenes doing the work. Measure whether they actually use it and whether 30% or more are willing to pay for continued access. That payment commitment is your go signal.

Your go/no-go decision matrix at the end of month three should cover four things. Is the problem real and acute for a specific, reachable segment? Is your solution clearly differentiated from current alternatives? Can you reach customers at an acquisition cost that makes unit economics viable? Have real people committed real money? If the answer to all four is yes, build. If two or three are yes, you have a pivot to consider. If fewer than two are yes, kill it and start again. Killing an unvalidated idea after 90 days costs you 90 days. Building it costs you years. Get started with a structured validation approach before you make that call.

Section 4: Customer Discovery Without the Sales Pitch

How to Structure Interviews That Actually Reveal Truth

The biggest mistake in customer interviews is selling. The moment you explain your solution, you corrupt the data. People are polite. They'll tell you it sounds great. They won't tell you they'd never actually pay for it. Keep your solution out of the first ten interviews entirely. Ask open-ended questions about their current workflow, their last experience with the problem and what they wish existed without prompting them with your answer.

Listen specifically for contradictions between what people say they do and what they describe doing. Someone says they "always" track their freelance income carefully, then describes a process of guessing at tax time. That contradiction is more valuable than ten minutes of confirmed problem statements. Use ChatGPT to process interview transcripts: paste the text and ask it to identify contradictions, unspoken anxieties and features that were requested in multiple different ways. Pattern recognition across 15 or more interviews is where the real strategic insights emerge, not in any single conversation.

Avoiding Sampling Bias

The people who will talk to you are not representative of your market. Your friends, your LinkedIn connections and your Twitter followers are all biased samples. They know you, they want to be supportive and they're not typical customers. This is one of the most underappreciated problems in early validation. Go to communities where your target customer already spends time: Reddit threads, Slack groups, industry forums, conference attendee lists. Cold outreach to strangers is harder but the signal is cleaner. A 20% response rate on cold outreach to 100 targeted prospects gives you 20 unbiased conversations, which is more valuable than 50 supportive chats with people who know your name.

Section 5: Common Pitfalls and How to Catch Them

The Echo Chamber, Vanity Metrics and the Feature Trap

Your network is not your market. Build this into every validation decision you make. The echo chamber effect is invisible from inside it: everyone around you has similar worldviews, similar problems and similar tolerances for your solution's shortcomings. Use ChatGPT explicitly to break out of it. The prompt "argue against this business idea using first-principles logic" is one you should run every two weeks during validation. Not because the AI is always right, but because the arguments it generates force you to confront objections you haven't prepared for.

Vanity metrics are the other silent killer. Website traffic does not equal product-market fit. Email signups without follow-up engagement do not equal validated demand. Social media followers do not equal paying customers. Build a simple validation scorecard in Google Sheets: track problem recognition rate, landing page conversion, interview-to-pilot conversion and payment commitment rate. Update it weekly. If three of four metrics are moving in the wrong direction after six weeks, you have a data problem, not a marketing problem.

The feature trap is where founders go when they've heard "no" and misread it as "not yet, but add this feature". Sometimes no means the positioning is wrong and a pivot in framing will fix it. Sometimes no means the problem isn't painful enough and no feature will change that. The signal to kill is when you've run 20+ interviews and fewer than 25% confirm the problem exists without prompting. The signal to iterate is when problem recognition is high but willingness to pay is concentrated in a segment you hadn't prioritized. Segment, don't add features.

Section 6: Tools, Market Sizing and the Decision to Build

The TAM/SAM/SOM Reality Check

Every founder inflates their market size estimates. It's not dishonesty, it's optimism bias. The fix is bottom-up sizing. Instead of starting with "the global market for X is $50 billion", start with "there are approximately 400,000 freelancers in the US who file quarterly taxes, and based on our interviews 30% experience acute pain around compliance". That gives you 120,000 potential customers. At $50/month, that's a $72 million annual market. That's a real, defensible number you can explain to an investor without embarrassing yourself.

ChatGPT is useful for finding data inputs to bottom-up models: ask it to identify reliable data sources for workforce size, industry spend and growth rate for your specific category. Treat its outputs as starting points for your own research, not finished answers. Cross-reference with government labor statistics, industry association reports and public company filings from adjacent markets. The HBS Online guide to market validation has a useful framework for structuring this analysis if you want a step-by-step reference.

Your Final Go/No-Go Scorecard

At the end of 90 days, you need a binary decision. Not "I think there's potential". Not "we just need more time". A clear go, pivot or kill. Score yourself against four criteria: problem validation (40%+ recognition rate in cold interviews), solution validation (10%+ landing page conversion and at least one presale), market validation (30%+ of pilot users willing to pay) and founder-market fit (do you have genuine insight or access advantage in this space). Three of four at target: build. Two of four: pivot the segment or solution framing and retest for 30 more days. One of four: kill it and protect your time for a better idea.

Validation isn't about proving you're right. It's about learning fast enough that being wrong costs you 90 days instead of two years. The founders who build enduring companies aren't the ones with the best ideas at the start. They're the ones who were honest enough to test their assumptions before betting everything on them. Read more on how other founders have run this process and what they found.

Run the experiments. Track the numbers. Make the call. Then build something people actually want to pay for.