The Startup Idea Validation Framework: 12 Lenses Before You Build

"Why do so many founders build things no one wants?"

Paul Graham asked that in his 2012 essay on startup ideas. His answer: "Because they begin by trying to think of startup ideas." That approach, he warned, "yields bad ideas that sound plausible enough to fool you into working on them." At Y Combinator, he wrote, "we call these 'made-up' or 'sitcom' startup ideas."

Ask a founder whether an idea has been validated and you'll often hear yes. Dig a little and it can mean a few people were asked and said something encouraging. That's understandable. It just isn't validation.

This guide sets out a framework for doing it properly. It's the same 12 lenses we use in Offer Validation, written up so you can run them yourself. It works for a founder with an idea and for an established B2B company weighing a new service line, a new segment or a price change.

A startup idea validation framework is a fixed set of tests you run every idea through, in the same order, so you can't skip the uncomfortable ones. Ours has 12 lenses in three families: market and demand, commercial and positioning, and risk and defensibility. Grade each on the strength of its evidence, and write your stop condition down first.

Why Most Idea Validation Is Really Confirmation

Most idea validation goes looking for agreement. Real validation goes looking for the conditions under which the idea fails.

Founders are optimists by design. That trait helps when you're building and hurts when you're deciding whether to build. It's how you end up interviewing five friendly contacts, hearing three encouraging answers and calling the idea validated. It's been confirmed. Nothing has tested it.

Tom Eisenmann at Harvard Business School calls one version of this a "false start". In Why Start-ups Fail, published in HBR in 2021, he writes that founders "keep rushing to launch fully functional offerings that don't fit any market needs." The step they skip is "researching customer needs before testing products."

There's a newer version too. Paste an idea into a chat assistant and you'll get a fluent, upbeat analysis in seconds. It feels like research. It's mostly your own framing reflected back, and we've written about why ChatGPT says your idea is brilliant in more detail.

What a Validation Framework Should Do

A validation framework should find the fastest path to a fatal flaw. If the flaw exists, you want to find it before the market does, while it's still cheap.

That means three things. It covers every kind of risk, including the ones you'd rather not think about. It runs in a fixed order, so the risks that can kill the idea outright come first. And it grades how good the evidence is, so a guess can't pass itself off as a fact.

Marty Cagan's four big risks are a useful check on coverage. He names value risk ("whether customers will buy it or users will choose to use it"), usability risk, feasibility risk, and business viability risk ("whether this solution also works for the various aspects of our business"). Any framework that only tests the first one will wave through ideas that customers like and the business can't afford to deliver.

Y Combinator's Jared Friedman has his own version for startup ideas. In this Startup School talk he covers the four most common mistakes and then ten questions to ask about any idea. The clip starts at the questions.

Play Video (Loads YouTube)

The 12 Lenses at a Glance

Each lens asks one question about one kind of risk. They fall into three families. The table gives you the question and the kind of evidence that answers it.

No lens works alone. An idea can look strong on eleven of them and still fail on one that matters, like a regulatory block or a delivery cost that eats the margin. That's the right outcome. A business that can't legally sell its product, or can't deliver it at a profit, isn't viable however good the rest looks.

Lenses 1 to 4: Market and Demand

The first four lenses test whether a real, findable group of buyers wants the result you're promising. If these fail, nothing later can rescue the idea.

1. Market demand and buyer language

Do buyers look for this outcome today, and what words do they use? Look for behaviour: searches, support tickets, forum threads, questions on sales calls. Then compare their words with yours. If you call it "workflow orchestration" and they call it "stopping things breaking at 2am", that gap is a finding in itself.

Paul Graham has a neat test for spotting a real problem in the wild:

One founder in the business automation space learned the buyer-language lesson the long way. After two failed attempts, they wrote about what three years of conversations had finally shown them:

2. Competitor gap

Name the three nearest alternatives, including doing nothing and doing it in-house. Where is your gap real, and where are you telling yourself a story? "Nobody does this" usually means one of two things. Either nobody needs it, or you haven't looked hard enough.

The practical work is dull and quick. Read each alternative's pricing page and note what's included at each level. Read their reviews, and spend most of your time on the middling ones. Five-star reviews tell you what buyers already get. Two- and three-star reviews tell you what they wanted and didn't get, often in their own words. That's where a real gap shows up, and it feeds straight back into lens 1.

If you're an established company, add your own lost-deal notes. The competitor you lost to most often last year is the one your new offer has to beat, whatever the market map says.

3. ICP and buyer clarity

Can you describe your first buyer tightly enough to find them and sell to them at a sensible cost? "UK SMEs" isn't a buyer. The UK government counts 5.7 million private sector businesses at the start of 2025, and 75% of them have no employees besides the owners. "Operations managers at logistics firms with 50 to 249 staff who schedule the fleet in spreadsheets" is a buyer. If you can't name the role, the company type and the tool they use now, this lens fails.

4. Outcome claim

Is the result you promise specific, believable and different from the alternatives? Vague outcomes don't convert. Specific ones need proof. In the UK, the advertising rules are blunt about this. Rule 3.7 of the CAP Code says "marketers must hold documentary evidence to prove claims that consumers are likely to regard as objective". If you can't back the headline promise today, you can't put it on the landing page either.

Lenses 5 to 8: Commercial and Positioning

Lenses 5 to 8 test whether the offer can be sold at a price that works, through a channel you can afford, to buyers who'll trust you.

5. Pricing power and positioning

Does the price match the value, measured against what buyers already pay for alternatives? The useful number is what your target buyer actually paid last time, to a competitor or a substitute. A theoretical price someone agreed with in an interview is much weaker. Our guide to testing an offer or price change covers how to test this on a live offer.

6. Channel fit

Does the offer work through the channel you plan to use, and can you afford to reach buyers there? A high-touch service rarely sells through self-serve ads. A low-price tool can't carry a field sales team. And watch for channels you don't control. Paul Graham's warning on this is short:

7. Objection map

What are the three most common reasons a buyer says no, and does the offer answer them before they're raised? Price, switching effort and "we'd need sign-off from IT" come up a lot in B2B. If your interview notes don't include any objections, you probably weren't listening for them.

8. Trust and credibility

Is there enough proof, track record or risk reduction to earn trust at the moment of purchase? A new founder usually scores low here, and that's fine if the plan accounts for it (a pilot, a guarantee, a reference customer). An established company often scores high, which is one of its real advantages. We cover how to make those claims hold up in proof claims on B2B websites.

Lenses 9 to 12: Risk and Defensibility

The last four lenses look for the things that sink otherwise good ideas: a crowded market, a legal problem, a delivery cost that doesn't add up, and bad timing.

9. Market saturation risk

Is the segment crowded, and are prices or features being squeezed? Lots of near-identical offers with falling prices is a warning sign. A market where incumbents exist but serve one segment badly is often a better sign than an empty one.

10. Regulatory and compliance

Do any of your claims or delivery methods create legal exposure? Health, financial and professional claims are the usual suspects, along with data protection. Name the rule that applies, with a source and a date. "We'll sort compliance later" is how a viable idea turns into an unviable one at launch.

11. Delivery viability

Can you deliver the promised result, reliably, at the stated price, without the margin collapsing? This is where services often fail quietly. The pilot works because the founder does it personally. The tenth client breaks it.

12. Stop criteria and timing

The last lens pulls the others together. What would turn a Go into a Change Course or a Stop, and is now the right moment? Timing kills more ideas than people expect. In CB Insights' March 2026 analysis of VC-backed shutdowns, bad timing was cited in 29% of cases where a reason could be identified.

Read the chart with care. These are companies that raised venture money and still failed, so they'd passed some scrutiny already. The mapping to lenses is my reading. CB Insights didn't make it. Still, most of what killed them is visible through the first five lenses and the last one.

Grading the Evidence, A to E

Every answer a lens gives should carry a grade for how good the evidence behind it is. The grade isn't a score for the idea. It's an honest label on what you actually know.

The same scale is on our methodology page, and we apply it to every claim in a Build Verdict.

An early idea will naturally have lots of D and E grades. That's not a mark against the founder. It's a map of what still needs testing. The mistake is treating D and E findings as problems to explain away. Treat them as a ranked list of experiments instead. Your validation isn't finished when a report says Go. It's finished when the D and E grades on the lenses that can kill you have been tested with real buyers.

The Kill Criterion: Write It Down First

A kill criterion is the condition that, if it turns out to be true, makes you stop or change course whatever else the evidence says. You write it before you look at the evidence, because afterwards you'll be tempted to move it.

A kill criterion isn't a risk factor. Risks make success harder. A kill criterion is a line. Annie Duke, the former professional poker player who now writes about decision-making, puts it well in a 2022 article for Behavioral Scientist: "The best quitting criteria combine two things: a state and a date." Her template is simple: "If I haven't done X by Y (time), I'll quit."

For an idea, that might read: "If fewer than three of the first twenty target buyers agree to a paid pilot by 31 March, we stop." For an established company testing a new tier: "If the add-on converts below 5% of renewals shown it in the first 60 days, we drop it." Pick your own numbers. The point is that they're written down and dated.

If you can't write a kill criterion, the idea isn't specific enough to validate yet. Sharpen it until one becomes obvious.

One way to find it is a premortem. Gary Klein, then chief scientist of Klein Associates, described the method in HBR in 2007: "team members assume that the project they are planning has just failed," then write down every plausible reason why. The reason that keeps coming up is usually your kill criterion.

The Order to Run the Lenses

Run the lenses that can kill the idea outright first, and the expensive analysis last. Here's the order I'd use:

1. Write the kill criterion (lens 12). Half an hour, before anything else.
2. Test demand, buyer and legality (lenses 1, 3 and 10). Interviews about past behaviour, a named buyer, and a check of the rules that apply. If any of these comes back E, stop and test it before going further.
3. Map the alternatives and the crowd (lenses 2 and 9). Who else does this, at what price, and how many of them are there?
4. Check the money (lenses 5 and 11). Price against what buyers pay today, and your real cost to deliver. Use actual numbers. Hoped-for ones don't count.
5. Then the go-to-market lenses (lenses 4, 6, 7 and 8). Outcome claim, channel, objections and trust.

A worked example

Here's how the order plays out on a made-up case. It isn't a client, just an illustration. Say a 60-person IT support firm wants to sell a fixed-price "AI readiness review" to its existing clients and to new firms of a similar size.

The kill criterion comes first: "If fewer than four of our 30 largest clients agree to a paid review by the end of next quarter, we drop it." Then demand and buyer. The account managers pull every ticket and email from the last six months that mentions AI, and the team calls ten clients to ask what they've done about it so far. The buyer turns out to be the finance director worried about data leaking into chat tools, rather than the IT lead. That changes the outcome claim and the price anchor before any marketing is written.

Lens 10 matters here because the review touches data protection, so the team checks what it can and can't promise on that front and writes it down with a source. Lenses 2 and 9 come next: how many local competitors already sell something similar, and at what price. Then lens 11, which is often the surprise. If each review takes a senior engineer three days, the price has to carry that, or the offer fails however keen the clients are.

Only then does the firm write the landing page and brief the sales team. By that point, each claim on the page has a grade next to it, and the weakest one has a test planned.

The common failure is doing it the other way round. Weeks of market sizing feel like progress, because they produce big numbers. Facing whether a buyer will pay your price through a channel you can afford is less fun and far more useful.

Kevin Hale gave one of the clearest talks on this for YC's Startup School, walking through how founders should evaluate ideas before they commit. It's from 2019, and the questions haven't dated.

Play Video (Loads YouTube)

Using the Framework for an Established Company's New Offer

The lenses are the same for an established B2B company. What changes is where the evidence comes from and which lenses are hardest.

You have advantages a founder would love. Existing customers you can ask. Support tickets and lost-deal notes that show real behaviour. A sales team that hears objections every week. And brand trust, which makes lens 8 far easier.

You also have risks a founder doesn't. A new tier can pull existing customers down from a higher price. A new service can stretch the delivery team that keeps current clients happy (that's lens 11 again). And internal enthusiasm is harder to argue with when the idea came from the board. The kill criterion matters even more here, because stopping a project with a sponsor is harder than stopping one founder's side project.

Talking to existing customers has a catch too. They already like you, so they're kinder than the market will be. Eric Migicovsky's Startup School talk on how to talk to users is the best short guide I know to getting honest answers anyway. It includes the three common interview errors from The Mom Test.

Play Video (Loads YouTube)

If the offer is concrete (who it's for, what they get, what it costs), you're past idea validation and into offer validation. Our guide to what market validation is walks through the six steps for that stage.

Your Pre-Build Checklist

Before you commit to a build or a launch, run this check against the evidence you have now. Each question maps to a lens. If you can't answer one, that's an E grade, not a gap to fill later.

One warning. Validation doesn't end at a Go. A Go means the evidence justifies the commitment you're about to make. It doesn't mean the idea will work. The list of D and E findings is your first roadmap in disguise. Build to answer those questions.

And sometimes the honest answer is to stop. That isn't a failure of the method. We explain what a Stop means, and how it differs from a Change Course, in why many startup ideas should be stopped.

Where ThriveFinity Fits

Our Offer Validation service runs these 12 lenses for you, against live sources, and grades each one A to E.

• Offer Signal (free, under an hour): describe your idea or offer in plain English and get a directional Go, Go With Conditions, Change Course, Stop or Wait, with all 12 lenses scored and three costed next moves. It's automated and unsigned. Try the free Offer Signal.
• Build Verdict (£149, within 24 hours): a named verifier grades all 12 lenses, runs a multi-persona stress test, sets stop criteria and signs the verdict. It carries our Accuracy Commitment, and the fee is credited to a Decision Brief booked within 30 days.

We're not the right choice if you haven't found a problem yet. The lenses need something concrete to test: a buyer, an offer and a price. If you're still exploring, do the interviews first. And we can't replace them. A verdict tells you where the evidence is weak and what to test next. Only buyers committing money tells you they'll pay. If you'd like to compare us with other tools before deciding, our validation tools comparison lays out prices, methods and where each one is the better fit.

Pranav Unni

Founder · ThriveFinity Connect on LinkedIn →

Pranav Unni is the founder and lead verifier of ThriveFinity. He reads and signs every paid deliverable personally, and writes about go-to-market decisions for established B2B companies.