Learning how to validate a startup idea is the cheapest business education you will ever get. Validation means replacing opinions with evidence: evidence that a specific group of people has a problem that matters, that they already try to solve it, and that they will pay for a better solution. Until you have that evidence, your idea is a guess.
This guide follows the right order: define the problem, check the market, interview customers properly, run simple tests, then decide whether to proceed, change direction or walk away. Most of it can be done in a few weeks alongside a job.
What validation really means (and what it does not)
Validation is not asking people whether they like your idea. Friends, family and even friendly strangers will tell you it sounds great, because being encouraging costs them nothing.
Real validation is based on behaviour, not compliments. It looks like this:
- People describe the problem unprompted, in their own words, with real examples.
- People already spend time or money on imperfect solutions.
- People give you something scarce: time for a follow-up meeting, an introduction to their manager, a place in a pilot, or money.
Y Combinator’s Startup School makes the same point: ask about specific things that have already happened in someone’s life, not hypotheticals about the future. People are poor at predicting their own behaviour, so stories about the past are data. Your goal is not applause — it is evidence you would still believe if a sceptical investor questioned it.
Step 1: Define the problem and the customer in one sentence
Before testing anything, write down exactly what you are testing:
[Specific group of people] struggle with [specific problem] when [specific situation], and it costs them [time, money or stress].
For example: “Independent bookkeepers in the UK struggle to chase late client documents when preparing VAT returns, and it costs them several hours of unpaid chasing every month.”
Be narrow: “small businesses” is too broad to interview or sell to, while “owner-managed cafés with fewer than ten staff in Manchester” is a group you can actually find. Y Combinator’s guidance on evaluating ideas is useful here: judge the problem before the solution, and ask whether it is frequent, urgent and expensive to live with.
Also list your riskiest assumptions — usually that the problem is painful enough, you can reach customers affordably, and they can and will pay. Test the one that would kill the idea fastest, first.
Step 2: Do your desk research before you talk to anyone
Desk research will not validate your idea on its own, but it stops you walking into interviews uninformed. The Start Up Loans Company, part of the British Business Bank, recommends starting with sources such as the Office for National Statistics (ONS) for real figures on potential customers. Useful free UK sources include:
- ONS data for population, business counts and industry trends.
- Companies House to look up competitors and view their filed accounts.
- Review sites and forums — repeated complaints about existing solutions are unmet needs in customers’ own words.
- Competitor websites and pricing pages, to see what is charged and who it is aimed at.
Competitors are not bad news: a market with none usually signals no proven demand. Look for a gap — a group competitors serve badly, ignore or overcharge — and record each competitor’s price, target customer and recurring weakness in reviews.
Step 3: Run customer interviews that tell you the truth
Customer interviews for startups are the highest-value step in this process, and the one founders most often do badly. The classic mistake is pitching: describing your idea, watching the person nod, and leaving with false confidence. A better model comes from Rob Fitzpatrick’s The Mom Test: ask questions so grounded in someone’s real life that even your mother could not lie to you with the answer.
How many interviews do you need?
Start with 10 to 15 people who closely match your customer definition, in 30-minute conversations. You are looking for patterns, not statistics. If the same problem, in the same words, keeps coming up unprompted, you are learning something real. If every conversation points a different way, your customer definition is probably too broad.
What to ask
- “Tell me about the last time you faced [the problem]. What happened?”
- “What did you do about it? What have you tried?”
- “What is the hardest or most frustrating part of dealing with it?”
- “How much time or money does it cost you when it goes wrong?”
- “What do you use at the moment, and what do you dislike about it?”
What not to ask
- “Would you use a product that does X?” — an imaginary future.
- “Do you think this is a good idea?” — an invitation to be polite.
- “Would you pay £30 a month for this?” — hypothetical guesses are worthless.
Listen far more than you talk, and note exact phrases — you will reuse that language on your landing page.
Step 4: Run a landing page test
Once interviews confirm the problem is real, test whether strangers act on your solution. A landing page test — or smoke test — is a single page describing the offer and asking for one action: join a waiting list, book a demonstration, or register for a pilot.
The page needs a headline in your customers’ words, a short description of the outcome rather than a feature list, who it is for, and one call to action. Drive targeted visitors to it: posts in your customers’ communities, messages to interviewees, or a modest test budget on search or social adverts.
Decide what counts as success before you launch, in writing. There is no universal pass mark, but setting the threshold first makes it a test rather than a hope. If visitors arrive but nobody acts, your message or offer is wrong. If nobody visits, you may have a distribution problem — itself a vital finding.
Step 5: Build the smallest useful MVP
A minimum viable product is widely misunderstood as a half-finished product. As Eric Ries defines it in The Lean Startup, it is the version of a product that gives you the maximum amount of validated learning for the least effort. Often, that is not software at all:
- Concierge MVP: you deliver the service manually for a handful of customers, learning what they value before automating anything.
- Wizard of Oz MVP: the customer sees a working front end while you do the work behind the scenes.
- Explainer video or prototype: a video or clickable mock-up of how the product would work, paired with a sign-up or pre-order option.
- Single-feature product: one core feature, done properly, with everything else left out.
- Piecemeal MVP: the service stitched together from existing tools — forms, spreadsheets, payment links and email.
Choose the type that tests your riskiest remaining assumption. If you are unsure people will pay, a prettier prototype teaches you nothing; you need a test involving money. Our guide to how to start a business with no money in the UK shows how far these low-cost tests can take you.
Step 6: Look for the strongest signals: commitment and pre-sales
Signals have a hierarchy. At the bottom: compliments and likes. In the middle: joining a waiting list, attending a demo, joining a pilot. At the top: commitment with a cost attached:
- Pre-orders or deposits, even small ones.
- Paid pilots at a reduced rate.
- Letters of intent to buy if the product meets agreed criteria.
- Introductions and time — a customer spending their reputation or diary on you.
Start Up Loans guidance makes the same point from the lender’s side: evidence of trading or advance orders outweighs desk research alone. If people will not commit, believe their behaviour, not their earlier encouragement.
Step 7: Do a unit economics sanity check
Validation is not complete until the basic maths can work — honest answers to four questions:
- What will a customer pay? Base this on what they spend now and what pre-sales achieved — not on interviewees’ guesses.
- What does it cost to serve one customer? Include time, tools, materials, delivery and payment fees.
- What will it cost to win one customer? If adverts cost more than a customer is worth over their lifetime with you, the model is broken however much people like the product.
- How often will they buy, and how long will they stay? Frequent, repeat purchases support a very different business from rare one-off sales.
If the price customers accept is below the cost of serving and winning them, you do not have a validated business yet — perhaps a validated problem with the wrong solution, customer or price.
Step 8: Decide — proceed, pivot or stop
Validation only helps if you act on what you learned.
Proceed when the same painful problem came up repeatedly, strangers acted on your landing page, some customers committed time or money, and the unit economics can plausibly work. Build the first real version for early customers, and sort the basics — including whether to trade as a sole trader or a limited company — before revenue starts flowing.
Pivot when the problem is real but your solution, audience or pricing is wrong: target a narrower customer, switch from software to a done-for-you service (or the reverse), or reposition around the outcome customers kept mentioning.
Stop when, after genuine effort, people do not describe the problem unprompted, will not act, and will not commit. You bought that knowledge with weeks and a small budget instead of a year and your savings.
Your validation checklist
- [ ] Problem and customer written in one specific sentence
- [ ] Riskiest assumptions listed, riskiest one tested first
- [ ] Desk research done: ONS and Companies House data, plus a competitor table
- [ ] 10–15 problem interviews completed, focused on past behaviour
- [ ] Landing page tested, with a success threshold set in advance
- [ ] MVP delivered in the simplest form that tests the key assumption
- [ ] At least one commitment signal sought: deposit, paid pilot or letter of intent
- [ ] Unit economics checked using real, observed prices and costs
- [ ] Decision recorded: proceed, pivot or stop — and why
Common validation mistakes to avoid
Asking friends and family. Their job, as they see it, is to support you. Interview strangers who match your customer definition instead.
Building first, validating later. Every week spent building before testing bets that your assumptions are right. Test the riskiest one first, in the cheapest way available.
Treating survey answers as proof. Surveys record opinions about a hypothetical future. Use them to find interviewees — not as demand evidence.
Counting vanity metrics. Page views, followers and free sign-ups are easy to inflate and hard to interpret. Track actions that cost the customer something.
Conclusion
Validating a startup idea is not a formality before the real work — it is the real work. Define a specific problem for a specific customer, check the market with real UK data, and interview people about their past behaviour, not your future product. Then raise the stakes gradually: a landing page, a simple MVP, and finally a request for commitment or payment. When the evidence says proceed, you build with confidence, because you know exactly whose problem you are solving. The same approach applies to product ideas: test demand before you buy stock when you start an online shop in the UK.
FAQs
How long should validation take?
Most founders can complete desk research, interviews and a landing page test within a few weeks of part-time effort. If validation drags on for months without a behavioural test — a page, a pilot or a pre-sale — it has probably become a way of avoiding an answer.
How many customer interviews are enough to validate a business idea?
Start with 10 to 15 interviews with people who closely match your target customer. You have enough when the same problems and phrases repeat and conversations stop surprising you. Interviews alone do not prove people will pay, so follow them with a test that asks for action or money.
Can I validate an idea without spending money?
Yes, in most cases. Interviews cost time, not money, a landing page can be built with inexpensive tools, and a concierge MVP uses your own effort in place of software.
What is the difference between an MVP and a prototype?
A prototype shows how a product might look or work. An MVP is put in front of real customers to test an assumption, and may involve a real transaction. A prototype can be part of an MVP — a clickable mock-up paired with a pre-order offer — but showing a design and asking if people like it is feedback, not validation.
What if nobody wants my idea?
Then validation has done its job, cheaply. First check whether you tested the right customer and problem framing — a pivot to a narrower audience sometimes changes everything. If the evidence stays negative, stopping frees your time and money for a better problem.
2 Comments