Startup Idea Validation: Test It Before You Build It
Four ways to turn an assumption into evidence: a demand test you run yourself, a clickable prototype, a free scoping session, or a proof of concept. Tell us what you are unsure about and we will tell you which one you need, including when the answer is none of them.
Not sure which one you need?
Describe the idea and the thing that worries you most about it. We come back within 24 hours with the route we would take and what it would cost.
Trusted By Founders
How We Approach Idea Validation
Startup idea validation is not one thing, which is why most of it is wasted
Replacing an assumption with evidence before you spend a build budget finding out. The reason so much of it is wasted is that founders reach for a generic exercise before naming which assumption is actually load bearing.
The wrong question
A survey, a market report or a scoring tool applied to an idea nobody has interrogated yet. Validating the wrong thing feels productive and tells you nothing, which is the worst combination available.
The right question
What would have to be true for this to work, and which of those things am I least sure about? A doubt about demand, about usability, about whether it can be built, and about what to build first are four different problems.
Four doubts, four instruments
Each has a different test, and using the wrong instrument is how founders end up confidently wrong. Most need one of the four below, not all of them, and a fair number need the free one and nothing else.
If you are already past this
If you know your idea works and just want it built, skip all of it and read rapid MVP development instead. More validation in front of a three week build is usually procrastination with a project plan attached.
Start here
Name the specific thing that would make you abandon the idea. That is what needs testing first, and everything below is organised around it.
Four things founders are unsure about, and what settles each
There is no single way to validate an idea, because there is no single kind of doubt. Find the sentence below that sounds like yours and it will tell you which engagement you actually need.
“Can this even be built?”
For a technical unknown: a model that has to hit an accuracy bar, an integration that may not expose the data, a latency or cost ceiling nobody has measured. We build the narrow spike that settles it and report the numbers.
Proof of concept development“Will anyone understand it, or want to use it?”
For a design and usability unknown. Real user flows and a clickable interface with no backend behind it, so you can put it in front of users and a pitch audience and watch where they get lost before any of it is expensive.
UX and prototype design“What should we actually build, and what will it cost?”
For a scoping unknown. We find the one core flow worth building first, cut the rest, and give you a realistic budget, a stack recommendation and a roadmap. No code is written and there is no obligation to build with us.
MVP consultingOr research it properly: product discovery“Will anyone actually pay for it?”
For a market unknown, and the one you should almost never pay an agency for. A landing page, a fake door, a pre-sale or a concierge run will tell you more than a build will, and you can run all of them yourself. We will point you at the right one.
The 11 validation methodsWhat each route costs, how long it takes, what you end up holding
Every route below answers a different question, so they are not alternatives to each other in the way price comparison implies. Two of the five are free or nearly free, and those are the ones most founders should start with.
| Route | The doubt it settles | Time | Cost | What you end up holding | Who does it |
|---|---|---|---|---|---|
| Demand test | Will anyone pay? | Days | A domain and some ad spend | A count of real actions: signups, pre-orders, deposits | You, almost always |
| Clickable prototype | Will they understand it? | 1 to 3 weeks | Scoped, quoted per flow | An interactive interface, and recordings of people failing to use it | A designer, with your users |
| Scoping call | What should I build first? | An hour | Free | A core flow, a budget range and a stack recommendation | A senior engineer and you |
| Product discovery | What should I build, and is my read of the market right? | 2 to 3 weeks | Fixed fee, quoted up front | A written specification any team can quote from | A researcher, an architect and your customers |
| Proof of concept | Can the hard part be built at all? | 2 to 4 weeks | Fixed price, quoted up front | A measured number against a pass mark, and a written verdict | Senior engineers, on your data |
The column that surprises people is the last one. A demand test needs nobody but you, and paying an agency to run one is close to the worst use of a pre-seed budget available. A proof of concept needs senior engineers and your real data, and there is no cheaper version of it that produces a trustworthy number.
Time is not the same as delay either. Three weeks of discovery in front of a three-week build is only a delay if the build was going to be right, and the whole reason to run it is that nobody knows yet whether it was.
Most validation goes wrong in the sequencing, not the method
Running the right test in the wrong order is the common failure. A proof of concept before anyone has checked whether the product is wanted answers a question that may not matter, and it costs weeks to find that out.
- 01
Whatever kills the idea fastest goes first
Not the thing that is most interesting, and not the thing that is easiest to start. If demand is the assumption that would end the project, no amount of technical certainty helps, and running a proof of concept first just means you learn the expensive thing before the free one.
- 02
Free before paid, days before weeks
A landing page and twenty conversations cost almost nothing and routinely settle the question outright. Only reach for a paid engagement once the cheap test either passed or cannot answer what you need. Most founders arrive having skipped this step entirely.
- 03
Technical risk before scope
If there is a genuine unknown about whether the hard part works, settle that before anyone spends three weeks specifying a product around it. A discovery phase that produces a beautiful specification for something that cannot be built is the most expensive document in this business.
- 04
Then build, and let real users answer the rest
At some point further testing costs more than shipping. If you can build the real thing in three or four weeks, more validation is usually procrastination with a project plan attached, and the MVP is the better instrument.
Four times the order changes
- An investor or a board needs the plan before releasing budget
- Discovery moves to the front regardless of sequence, because its output is the document that unlocks the money.
- You are selling into one named enterprise customer
- Their technical requirements are the demand test. Run the proof of concept against their systems and skip the market work.
- A previous attempt failed and nobody knows which assumption broke
- Start with discovery even if you think you know the answer. The belief that ended the last build is usually still in the room.
- The build is genuinely three weeks
- Skip most of this. Two weeks of validation in front of a three-week build rarely pays for itself unless the technical risk is real.
The four cheapest tests, and the number that counts as a pass
You should almost never pay an agency for these, so here is how to run each one properly, what it costs, and the threshold that separates a signal from a nice afternoon. If one of them answers your question, you have saved yourself an engagement.
Landing page test
A domain and roughly $100 of ads3 to 5 daysOne page describing the product as though it exists, a single call to action, and paid traffic pointed at the exact audience you claim to serve.
What counts as a pass
Above 5% of qualified visitors giving you an email. Below 2% and the positioning is wrong, the audience is wrong, or the problem is not urgent.
Fake door
Free if you already have traffic1 to 2 weeksAdd the button for the feature that does not exist yet. Anyone who clicks gets an honest "not built yet, want to be first?" and joins a list.
What counts as a pass
Enough clicks to justify the build against the same button's competition. This is the only test that measures intent inside real usage.
Pre-sale
A payment link2 to 4 weeksSell it before it exists, at the real price, with a refund promise and a delivery date you can hit. Money changing hands is the least ambiguous signal available.
What counts as a pass
Any paying customer at full price is worth a hundred survey responses. Discounting to close the first sale invalidates the test.
Concierge
Your own timeOngoingDeliver the outcome manually for a handful of customers, with no product at all. You are the software, and you learn exactly which parts are worth automating.
What counts as a pass
Customers who keep coming back and would pay for the manual version. If they will not pay for the outcome by hand, software will not fix that.
Publishing this costs us engagements, which is the point. A founder who runs a landing page test and gets nothing has saved a build budget, and one who gets a queue of signups arrives at a build conversation with evidence instead of hope.
Rank evidence by what it cost the person to give you
Every signal is worth exactly what the other person gave up to produce it. That single rule sorts the whole ladder below, and it explains why a paying customer outranks a hundred people who said they loved the idea.
| Weight | What they did | What it cost them | Why it ranks there |
|---|---|---|---|
| Strongest | They paid, at full price, before it existed | Money | Nothing else is close. A pre-order with a refund promise is the cleanest evidence available to a pre-product company. |
| Strong | They signed a letter of intent, or gave you a start date | Reputation | In B2B this is the realistic ceiling before a product exists, and it survives contact with an investor. |
| Strong | They kept using the manual version | Time, repeatedly | A concierge customer coming back a fourth time is worth more than any number of first-time signups. |
| Moderate | They gave a work email and opened the follow-up | A little attention | Real but soft. Email capture measures curiosity, and curiosity converts at a rate you cannot predict yet. |
| Weak | They clicked a button for a feature that does not exist | A second | Useful for comparing two options against each other. Almost useless as an absolute number. |
| Noise | They said they would definitely use it | Nothing | Stated intent about hypothetical money. Generous, encouraging, and the single most common reason founders build the wrong thing. |
And four signs you have validated enough
The same answer keeps arriving
By interview six or seven the themes repeat. If conversation nine tells you what conversation four did, further interviews are costing money to confirm something you already know.
You know what you would build tomorrow
If you can name the one core flow and defend cutting everything else, the remaining uncertainty is about execution, and execution is answered by building rather than by asking.
The next test costs more than the build
Against a three or four week MVP, two more weeks of validation rarely pays for itself. At that point the product is the cheapest instrument you have.
You are testing to feel better, not to decide
The honest check is whether a negative result would actually change what you do next. If it would not, you are not validating, you are collecting reassurance.
Four things that feel like validation and are not
Each of these produces a confident answer without producing any evidence, which is worse than having no answer at all, because it is the kind of thing founders quote back to themselves for a year.
An AI tool scoring your idea out of 100
It is scoring your description, not your idea. It has no access to your customers, your market or your pricing, so it can only pattern-match against how similar the wording is to things that worked before. Confident output, zero evidence.
Do this instead: Show the thing to twenty people who would pay, and count how many try to.
A market report saying the sector is worth $50bn
Market size tells you the prize if you win. It says nothing about whether you will, and every failed startup in the category had the same number in its deck. It is a reason to be interested, never evidence that you are right.
Do this instead: Find the number of people who took a real action: a payment, a signup, a booked call.
Asking friends, family or your network
They are answering a different question, which is whether they like you. Enthusiasm from people who are not the customer is the most common reason founders spend a year building the wrong thing feeling encouraged the whole way.
Do this instead: Ask strangers who have the problem, and watch what they do rather than what they say.
A survey where people say they would pay
Stated intent and actual behaviour diverge badly, and the gap is widest exactly where it matters, on price. People are generous with hypothetical money because saying yes to a survey costs them nothing.
Do this instead: Take a pre-order, a deposit or a signed letter of intent. Money is the only honest survey.
Everything we know about validating an idea, written down
Most founders should run a demand test before they hire anyone, including us. These are the guides we would send you anyway, so they are here rather than behind a form.
The methods
Eleven ways to test demand before you write code, and when each one is the right instrument.
Know what you are buying
The words get used interchangeably and they are not interchangeable. Buying the wrong one is the expensive mistake at this stage.
The fundamentals
What validation is for, what counts as a signal, and what to do with the answer once you have it.
And when the evidence says go
Validation is only worth doing if it changes what you do next. When the answer is positive, the work goes straight into a build: the scope is already defined, the technical risk is already measured, and the first version costs less and takes less time than one scoped blind. That build is an investor-ready MVP in 21 days. When the answer is negative, we say so, and you have saved a year.
Explore our other MVP builds
Building something that spans categories? These related MVP development services share the same senior team, fixed timeline, and full code ownership.
Common Questions About Idea Validation
What is idea validation?
How do you validate a startup idea?
How much does idea validation cost?
How long does it take to validate an idea?
What is the difference between idea validation and market research?
Do I need to validate before building an MVP?
Can I validate my idea myself?
What happens after validation?
What is the one thing that would kill this idea?
Tell us the idea and the doubt underneath it. We come back within 24 hours with the route we would take, what it costs, and how long it takes. If the free route would settle it, we will say that instead.
Tell Us What Worries You