VettedGaps
All research
Idea validation11 min read

Startup Idea Validation: How Much Evidence Before You Commit

Startup idea validation for founders: which evidence justifies your evenings, savings or a raise, what each signal can't prove, and the kill criteria to set.

Startup idea validation is the check a founder runs before each new commitment of time or money: is the evidence strong enough for that commitment? Size the commitment to the strongest evidence you hold. Complaints from several unrelated people justify evenings of research and interviews. Money already spent on the same job, through freelancers or paid products, justifies a prototype. Quitting a job, spending savings or raising money needs your own buyers paying or committing to pay, measured against a kill criterion you wrote down before the test.

No complaint count or search volume shows that anyone will pay for your product. This page is for indie developers and solo founders building software, usually an add-on for an app marketplace. It covers the decision to commit, not the step-by-step method or idea lists: the method is in how to validate a business idea, and candidates are in startup business ideas.

Key takeaways

  • Name the commitment first, then ask whether your evidence is strong enough for it.
  • Complaints, searches and freelance orders each prove one thing, and none of them tests your price.
  • In our public sample report, one of four functions had direct evidence, so the supported next step was interviews.
  • Write a pass mark and a kill line before each test, and decide what a result between them means.
  • Only buyers paying for your own product, even as a pre-order, justify quitting a job or raising money.

What startup idea validation has to decide

The practical answer to how to validate a startup idea is to do it in stages, because a founder’s bets come in stages: some evenings, then a prototype, then savings or a salary, then other people’s money.

That turns a question you can’t check (“is this a good idea?”) into one you can: does the evidence I have today cover the next thing I am about to spend? Idea validation for startups differs from testing a hobby project mainly because the cost of being wrong rises at every stage. So the answer to how to validate an idea for a startup depends on where you are: a founder with a problem statement and no data has different homework from one with ten interviews and a landing page.

A startup idea validation framework: the evidence-to-commitment rule

Our validator grades every function of an idea (one job the product does) on five evidence levels: E0 not observed, E1 people said it, E2 people search for it, E3 people pay someone to do it, E4 people pay for a product. The scale itself is explained in idea validation, and the method page shows how each signal is collected. We turn those levels into commitments with what we call the evidence-to-commitment rule:

Commitment Evidence that covers it What that evidence leaves open
Evenings of research and interviews E1 from several unrelated people, in more than one source Whether anyone pays to fix it
A landing page or clickable mockup E1 plus E2 for the same job Whether searchers want your version
A working prototype E3 or E4 for the function you build first Whether buyers switch to you
Quitting a job or spending savings Your buyers pre-order or start a paid pilot, above a line set in advance How many more buyers exist
Raising money Paying customers of your product Whether you keep finding buyers at that price

E0 counts neither for nor against an idea, but a function still at E0 after a real search is a reason to cut it from the first version.

Live evidence at each level, and what it leaves open

We read every number below on its live page on 20 November 2026.

Signal Live example What it proves What it can’t prove
Complaints (E1) Card 4136: 307 posts from 258 people, demand score .94 on a 0 to 1 scale Many unrelated people hit the same failure That they would pay you
Search (E2) Sample report, first function: 190 searches a month People look for a fix often enough to measure That searchers are buyers
Freelance orders (E3) Idea 33: 37+ paid Fiverr orders, median order $92 The job has a price and buyers with a budget That software can replace the freelancer
Paid products (E4) Sample report: 3 competitors observed for the first function Buyers already pay software for this job That they will leave it for yours
Your commitment test Sample report plan: 20 merchants asked to commit to a $29 a month pilot Whether your offer gets a yes at your price Retention

Complaints come first because they are free. Every day we read app marketplace reviews, vendor forums, Reddit and one-star Chrome extension reviews, drop praise and support noise, and group posts about the same frustration into pain cards. As of November 2026 the index holds 7,775 cards across 78 marketplaces, all about software, so for a physical product you collect complaints by hand.

Card 4136, dropshippers whose Meta campaigns burn budget with no sales, is one of the top pains on the Shopify App Store. One of them wrote: “At these CPM numbers I can’t see this ever being profitable.” (source) However many people repeat it, that is E1 evidence. The people on the card already lose money to ads, and nothing on the card shows they would spend more on an app.

Search is E2. The sample report sets a bar of 50 searches a month and counts no volume for a function unless a confirmed complaint backs it.

Freelance orders are the first signal where money has moved. Our paid demand catalogue groups 78,710 Fiverr task clusters, and it treats buyer reviews as a conservative lower bound of demand, not a count of orders. On idea 33, a Squarespace export tool, all three signals agree: 10 people who complained, about 2,220 Google searches a month for “squarespace alternatives” and related phrases, and the paid orders above. One complaint reads: “300 a year for a site you barely touch is wild when free options exist.” (source) The idea page still says no paid orders for an exit-specific tool were observed.

Reviews of apps people already installed show where those products fall short. Card 3981 groups 19 posts from 19 people, Shopify merchants hit by Google Merchant misrepresentation suspensions, and several posts are app reviews like this one: “Same misrepresentation issue as another user, but no clear answer as to what specific rule my specific store was failing at.” (source)

How to validate a startup idea before building: three stages

Each stage names the evidence that lets you move on and the result that should stop you.

Stage 1: before you spend your evenings

You need a one-sentence problem and E1 from people who don’t know each other, found in more than one place. Count distinct people rather than posts. Stop if the only evidence is your own experience plus one thread.

Stage 2: before you build a prototype

You need E3 or E4 for the function you would build first, or E2 plus interviews in which people describe the last time the problem hit them. The sample report gives two stop lines here. Kill the idea if an existing app already offers all your functions together with good reviews and active users, or if fewer than 3 of 10 interviewed merchants express strong interest or willingness to pay.

Stage 3: before you quit, spend savings or raise

You need a commitment your buyers can refuse, such as a pre-order or a paid pilot. The sample report’s version asks 20 multi-warehouse merchants to commit to a pilot at $29 a month. If you plan to raise, show investors those same payments, since complaint counts and search volume only show that a market might exist. This is product evidence, not financial advice: your own runway decides how much salary or savings you can risk.

Kill criteria, and results between pass and kill

A kill criterion is the result, written in advance, that ends the idea. The sample report’s commitment test holds two numbers. Its assumption is “At least 5 of 20 merchants will commit to a paid pilot”. Its kill criterion is “Fewer than 2 merchants commit to a paid pilot.”

A result of 2, 3 or 4 commits falls between them. It passes nothing and kills nothing, and it can cost a founder months, because a weak yes feels like progress. Decide what each band means before the test. Our suggestion: under 2, stop. From 2 to 4, change one thing (the buyer, the price or the feature you lead with) and rerun once. At 5 or more, build only what the committed buyers asked for.

Stop signals worth writing down before you start:

  • Every complaint you can find comes from one thread or one person.
  • After a real search, the function you would build first is still at E0 or E1.
  • People describe the problem in interviews, yet none can say when it last cost them time or money.
  • A test lands under the line and you want to move the line.

A worked decision from a Needs proof report

Our sample validation report checks a four-function multi-warehouse stock tool against the Shopify App Store and its 254 pain clusters. Only the first function, comparing stock levels after a failed sync, reached E2, backed by 2 distinct people and 190 monthly searches. They wrote: “overselling on amazon when shopify already sold the last unit is a nightmare to fix” (source) and “Having to email a customer to apologize and cancel their order because the stock didn’t actually exist completely ruins brand trust.” (source) The other three functions were E0. The verdict was Needs proof, with medium confidence.

For a founder, that verdict covers stage 1 and part of stage 2 for one function. It supports interviews and a mockup for that function, not a four-function build, and the report warns the multi-warehouse angle may be narrower than the evidence, because the supporting complaint is about overselling across sales channels.

Idea 33 is one stage further on. Complaints, searches and freelance orders agree, so a cheap prototype is covered. The open question is price: the idea page proposes testing a one-time fee of $92 per site export, close to the median freelance order, and that test is the stage 3 decision.

FAQ

Should I pay someone to validate my startup idea?

Pay only for data you can check. A useful report links the complaints, searches and payments behind each function and says when nothing supports one. What AI validators can and can’t check is in AI tools for validating product ideas. The commitment test at stage 3 is the one step only you can run, because it asks your buyers for money.

Is a waitlist enough to validate your startup idea?

No. A waitlist sign-up costs the buyer nothing, so it measures curiosity. Ask for a pre-order, a deposit or a paid pilot, and set the number that counts as a pass before you publish the page.

How many interviews should I run before I build?

The sample report plans 10 interviews with merchants who have the problem and kills the idea below 3 positive answers. Pick your own numbers, but write both down before the first call. How to run the interviews is covered in how to validate a product idea.

What if the signals disagree?

Trust the one where money moved. Loud complaints with no payment anywhere call for a commitment test before code. Payments with few complaints often mean the job is solved well already, so check competitors first.

Make the startup idea validation call

Startup idea validation is done when your evidence covers the next commitment, and you can say which stage you are at and what result would stop you. To learn how to validate your startup idea against real marketplace data, paste it into our idea validator. It splits the description into functions and grades each from E0 to E4 with links to the evidence. Validation is part of Pro at $15 a month or $150 a year, with 5 validations a day, as listed on our pricing page.

Other ways people search for this

Some people type the question as how startup founders validate ideas before investing time money; the three stages above answer it. Others search how to validate startup idea before building or validate idea startup, and the same stages apply.