How to validate a startup idea: a 7-step process that includes defensibility

To validate a startup idea, prove five things in order: the problem is real and frequent, specific customers feel it, they already spend time or money on it, they will pay you, and the business can survive being copied. Most guides stop at demand. The last step, defensibility, decides whether success is yours to keep.

This guide walks through each step, the evidence that counts, and how to use AI without fooling yourself.

What validation is (and is not)

Validation is collecting evidence that changes your confidence in a business, before you spend months building it. It is not asking friends whether they like the idea, and it is not a polished landing page with a waitlist of people who will never pay.

The stakes are real. In CB Insights' analysis of 431 VC-backed startups that shut down, poor product-market fit was cited by 43%. The term product-market fit itself is usually credited to Andy Rachleff, and popularized by Marc Andreessen, who defined it as "being in a good market with a product that can satisfy that market." Validation is how you find out cheaply whether you are on the path to it.

The 7-step process

Step 1: Confirm the problem

Talk to people who should have the problem, and ask about the past, not the future. Rob Fitzpatrick's book The Mom Test is built around this: how to learn "if your business is a good idea when everyone is lying to you." The core habit is to ask about specific recent events ("tell me about the last time this happened") rather than opinions ("would you use an app that...").

Signals of a real problem:

• People describe it in detail without prompting.
• They have already tried to solve it.
• It happens weekly or daily, not once a year.
• It costs them money, time or risk they can name.

Step 2: Narrow the customer

"Small businesses" is not a customer. "Independent physiotherapy clinics with two to ten practitioners in the US" is. A narrow segment makes interviews comparable, makes outreach possible, and gives you a winnable starting market later, which matters for defensibility.

Step 3: Look at current behavior and spend

The strongest early demand signal is that people already spend money or effort on the problem: a spreadsheet maintained every week, a part-time hire, a tool they pay for and complain about. Behavior beats stated interest. If nobody is doing anything about the problem today, ask why.

Step 4: Map the competitors

Search for the problem, not your product name, and look for four kinds of competitor: direct, indirect, adjacent and platform. Our guide does my startup idea already exist? gives a 30-minute method. Finding competitors is usually good news: Paul Graham argues that a crowded market means "both that there's demand and that none of the existing solutions are good enough."

Step 5: Test willingness to pay

Interest is cheap. Ask for something that costs the customer:

• B2B: a paid pilot, a letter of intent with a price, or a deposit.
• Consumer: a pre-order, a paid beta, or a smoke test where the "buy" button collects payment details.
• Marketplaces: manually match a few buyers and sellers and see if money changes hands.

After launch, a common check for product-market fit is the Sean Ellis test: whether at least 40% of surveyed users would be "very disappointed" if they could no longer use the product.

Step 6: Check defensibility (the step most guides skip)

Most validation guides end at Step 5. That leaves a gap: proving demand is exactly what attracts competitors. Once your idea works, a company with more money, users and distribution can study your product and ship a version of it. That is what happened to Slack when Microsoft bundled Teams into Office 365, and to Snapchat when Instagram copied Stories; our guide on what happens if a big company copies your startup walks through those cases with sources.

The defensibility check asks three questions:

1. Who is the most likely copier? Name a company, not a category. Usually it is the one that already has your customers, data or distribution.
2. What would still make customers choose you after they copy it? Look for network effects, proprietary data, switching costs, owned distribution or owned customer relationships.
3. Could a platform or AI model provider ship this as a feature? If Apple, Google, OpenAI, Meta or Shopify could, you need a reason they would not or could not. See platform risk and AI moats.

You do not need a moat on day one. You need a credible path to one, designed into the product, and an honest view of who could copy you before you get there. Our guide on how to make a startup defensible covers how to build it.

Step 7: Decide with criteria you set in advance

Before you start, write down what would make you proceed, change direction or stop. For example: "At least 8 of 15 interviewees describe the problem unprompted, 3 agree to a paid pilot, and we can name a moat the likely copier lacks." Criteria written afterward tend to bend toward the answer you wanted.

B2B vs consumer validation

The steps are the same, but the evidence looks different.

B2B ideas are easier to validate through conversation, because buyers can describe their workflow, budget and current tools in detail. The risk is mistaking a friendly champion for a buyer. Ask who signs, what budget it comes from, and what they would stop paying for to afford you. On defensibility, look hardest at the software vendors your customer already pays, since they can add your product to an existing contract.

Consumer ideas are harder to validate by interview, because people are poor predictors of their own behavior. Lean more on observed behavior: what people already do, spend and download, and how a small group uses a rough version over several weeks. On defensibility, look hardest at platforms and large apps that already hold your users' attention, since they can ship a copy to an existing audience overnight.

Marketplaces need validation on both sides. Many fail not because no one wants them, but because one side never arrives in sufficient density. Validate in one small market first, which is also where a network effect starts.

How to validate a startup idea with AI

AI tools are useful for the desk-research parts of validation, if you use them adversarially.

Good uses:

• Brainstorming customer segments and the different ways each describes the problem.
• Drafting interview scripts, then checking them against Mom Test rules (past behavior, no pitching).
• Summarizing public reviews of competitors to find repeated complaints.
• Generating a list of search terms for competitor research.
• Arguing against your idea. Ask: "You are a skeptical investor. Give the five strongest reasons this fails, and name the company most likely to copy it."

Poor uses:

• Asking "is this a good idea?" General chat models tend to be agreeable and will usually find something positive to say.
• Trusting company names, funding amounts or market sizes without checking the source. Models without live search can list companies that no longer exist and miss recent launches.
• Treating synthetic "customer interviews" generated by a model as evidence. They tell you what the model thinks customers say, not what your customers do.

Useful prompts:

1. "List 10 distinct customer segments that might have [problem], and for each, the phrase they would type into Google."
2. "Here are 20 reviews of [competitor]. Group the complaints by theme and count them."
3. "Here is my idea. Which existing company could launch it fastest, and what would they still be missing?"

Moat Check automates the defensibility part of this: it runs live web research to find existing competitors, names the most likely copier, and scores the idea on ten tests (pass, partial or fail, with one line of evidence each) described in our methodology. It does not replace customer conversations or willingness-to-pay tests, and it cannot judge your team or timing.

Common validation mistakes

• Pitching instead of listening. Once you describe your solution, people become polite.
• Counting signups as demand. A free waitlist measures curiosity.
• Validating with the wrong people. Friends, other founders and investors are not your customers.
• Stopping at demand. A proven market without a moat is a gift to the next, better-funded entrant.
• Never deciding. Validation without a decision rule becomes procrastination.

A one-page validation checklist

• Problem described by customers in their own words, with recent examples.
• One narrow, reachable segment named.
• Evidence of current spend or workarounds.
• Competitor map including adjacent and platform players.
• At least one commitment that costs the customer something.
• A named likely copier and at least one moat it lacks.
• A written build, change or stop decision.

For examples of how ideas score on the defensibility step, browse the idea library, or read what a moat is for the fundamentals.

Check your startup's moat

Questions

Keep reading

• What is a moat in business?
• Does my startup idea already exist? How to check, and what to do if it does
• Economic moat vs startup moat: what investors mean, and what founders need
• How to make a startup defensible (and build a moat around your business)
• 7 Powers vs moats: how Hamilton Helmer's framework maps to startup moat tests
• The 10 types of moats