Most founders validate the wrong thing. They test whether people like the idea, collect a pile of polite maybe's, then spend four months building something nobody buys. I know because I did exactly that in 2023 — burned 11 weeks and about $4,000 on a scheduling tool that 40 interviewees said they'd "definitely use." Six people signed up. Two churned in a week.
The fix wasn't reading the Lean Startup book harder. It was building a validation framework — a sequence of gates, each with a number attached, that tells you whether to keep going or kill the thing. Here's the exact structure I've used since, and the thresholds I actually apply.
Key takeaways
- A validation framework is a decision system, not a research plan: every stage ends in a go / pivot / kill call with a pre-set number.
- Validate in order: problem, then willingness to pay, then solution, then channel. Skipping ahead is the most expensive mistake you can make.
- The 80/20 rule for startups isn't "do less work." It's identifying the one or two assumptions that would collapse the business if wrong — and testing those first.
- Verbal interest is worthless as evidence. A deposit, an email with a credit card behind it, or a signed LOI is not.
- Your framework should fit one page. If it needs a slide deck to explain, it won't survive contact with a busy week.
What a lean startup validation framework actually is
The Lean Startup framework, as Eric Ries laid it out, is built on a loop: Build-Measure-Learn. You turn an idea into something testable, measure how reality responds, learn, and repeat. The output of that loop is what he called validated learning — knowledge backed by evidence rather than opinion.
That's the engine. But an engine is not a steering wheel. What most founders lack is what sits on top of Build-Measure-Learn: a decision layer that says which assumption to test, what number counts as a pass, and what happens when you fail.
So here's the definition I work with:
A validation framework is a one-page map of your riskiest assumptions, ranked by how badly they'd hurt you if wrong, each paired with a test, a threshold, and a pre-committed action. Nothing more. The Build-Measure-Learn loop is how you run each test. The framework is how you decide what to do with the result.
What are the 5 principles of Lean Startup?
The five principles are: entrepreneurs are everywhere; entrepreneurship is management; validated learning beats vanity metrics; build-measure-learn is the core loop; and accounting for innovation requires measuring progress differently than traditional business does. I'd add a sixth from practice — every experiment needs a kill criterion decided before you run it. Without that, you'll rationalize any result as partial success.
The four gates every validation framework needs
I run everything through four gates in strict order. Each one is cheaper than the next, and failing early saves you the most money.
Gate 1: problem validation
The question is not "do people have this problem?" Everyone has problems. The question is: are they already spending money, time, or workarounds on it?
I look for two signals. First, unprompted mention of the problem in conversation — if you have to explain why it's a problem, it isn't painful enough. Second, evidence of an existing patch: a spreadsheet, a freelancer, a $30/month tool they hate.
My threshold: at least 8 of 15 conversations where the person describes the problem in their own words, unprompted, before I mention my solution. Below that, I stop. I killed a B2B idea last year at 6/15 and it was the right call.
Gate 2: willingness to pay
This is where most founders lie to themselves. "They seemed really excited" is not data. Here's what counts, in ascending order of credibility:
- A reply to a pricing email asking for an invoice
- A pre-order or deposit — even $20
- A signed letter of intent from a company
- An actual payment before the product exists
For B2C, I use a landing page with a real checkout. For B2B, I ask for a deposit against the first invoice. My threshold is blunt: 3 paying commitments from 50 qualified conversations. That's a 6% rate, and in my experience it's the floor where a business becomes plausible. Below 3%, the problem isn't urgent enough or you're talking to the wrong people.
Gate 3: solution validation
Only now do you build. And even here, build the smallest thing that can be rejected. A Figma prototype in a live call, a concierge version where you do the work manually, a Wizard-of-Oz build where users think it's automated but you're behind the curtain.
Threshold: 40% of the people who committed at Gate 2 complete the core action (finish the flow, send the message, book the session) without help. If they need you standing over their shoulder, the solution failed.
Gate 4: channel validation
Your product can work and the business can still die because you can't acquire customers profitably. So the last gate: can you reach these people for less than they're worth?
A rough version I use: run three small paid campaigns at $100 each across different channels. If none gets the cost per qualified lead below one-fifth of your intended first-year revenue per customer, the channel is broken, or the product needs longer retention to pay back.
The 80/20 rule for startup validation
What is the 80/20 rule for startups?
Applied to validation, the 80/20 rule means roughly 20% of your assumptions carry 80% of the risk. Most founders spread their testing evenly across every uncertainty, which wastes weeks. Instead, list every assumption your business rests on — pricing, demand, retention, channel cost, technical feasibility — then rank them by impact if wrong × probability of being wrong. The top two or three go in the framework. Everything else waits.
In my own projects, the single highest-risk assumption was almost never technical. It was whether the pain was urgent enough that someone would pay before seeing a finished product. That one assumption usually decides everything downstream.
| Assumption type | Typical risk level | Cheapest test | Rough cost |
|---|---|---|---|
| Problem urgency | High | 15 problem interviews | Time only |
| Willingness to pay | High | Landing page with checkout or deposit ask | $50–200 |
| Solution usability | Medium | Prototype or concierge test | 1–3 weeks |
| Channel economics | Medium | 3 small paid tests | $300 |
| Technical feasibility | Low to medium | Spike or off-the-shelf stack | Days |
Is it true that 90% of startups fail?
The 90% claim
The often-quoted figure is that about 90% of startups fail, though the number varies a lot depending on how you define "startup," "fail," and the timeframe you measure. The direction is right even if the exact number is slippery: most new ventures don't make it, and the reasons cluster around a handful of causes.
What matters more than the statistic is why they fail. The dominant cause is building something nobody wants — a product that solves a problem people don't actually have, or don't have urgently enough to pay for. That's precisely the failure a validation framework is designed to catch, and it's cheap to catch at Gate 1 or 2. The expensive version is catching it after 18 months of engineering.
I'll be honest: even with a framework, I've had two projects fail. One died because the problem was real but the buyers had no budget authority. The other because the channel cost exceeded what customers were worth. Both were caught by the framework — just later than I'd have liked. Neither cost me more than six weeks.
How to run the framework without burning weeks
A practical sequence
- Week 1: List every core assumption. Rank by risk. Circle the top two.
- Week 2: Run 15 problem conversations. Score against the 8/15 threshold.
- Week 3–4: Test willingness to pay. Aim for 3 commitments from 50 conversations.
- Week 5–6: Build the smallest rejectable solution. Measure the 40% completion rate.
- Week 7: Run three small channel tests. Check the cost-per-lead math.
Every gate ends in a written decision: go, pivot, or kill. And here's the part people skip — write the kill criterion before you run the test. The moment you're staring at disappointing results, your brain will invent reasons they're actually fine. Pre-commitment removes that option.
B2B vs B2C differences
For B2B, everything happens slower and in fewer conversations. Ten well-chosen buyer interviews can be decisive. Deposits and LOIs are realistic asks. For B2C, you need volume — hundreds of visitors, not dozens of conversations — and the fastest signal is a paid checkout on a landing page. The gates are identical; the sample sizes and the credibility of the commitment differ.
The mistake that kills more startups than any other
It's treating validation as a phase you complete and move past. It isn't. It's a loop you stay inside, because your market keeps moving. The framework isn't a one-time gate you clear before building — it's the instrument panel you keep watching.
If you take one thing from this: decide your numbers before you go looking for evidence. The framework's whole value is that it takes the decision out of your hopeful hands and puts it somewhere honest.
What's the riskiest assumption in your current idea — and what number would convince you to walk away?
