Here's a stat that should terrify you: 42% of startups fail because there's no market need. Not because the tech didn't work. Not because they ran out of money. Because they built something nobody wanted.
The tragic part? Most of these founders spent 6-12 months building before they discovered the truth.
After 20+ years building products used by over 25 million people, I've watched this pattern repeat endlessly. Smart people. Good ideas. Terrible validation process.
2026 is different. The tools have changed. The playbook has evolved. You can now validate a startup idea in days, not months—if you know which approach to use.
The validation spectrum: from cheap to conclusive
Every validation method sits somewhere on a spectrum. On one end: cheap and fast, but lower confidence. On the other: more investment, but you get real proof people will pay.
The mistake most founders make is staying too long at the cheap end. They collect 50 survey responses saying "yes, I'd use that" and call it validated. Then they're shocked when those same people won't pay £10/month for the real thing.
Let's break down your options from least to most conclusive.
Level 1: Customer discovery conversations
Cost: Free (just your time)
Timeline: 1-2 weeks
Confidence level: Low to medium
This is where everyone should start. No exceptions. Before you build anything, you need to understand the problem space deeply.
But here's where most founders go wrong: they pitch their solution instead of exploring the problem.
The right way to do customer discovery:
- Talk to 15-20 people who fit your target profile
- Ask about the last time they experienced the problem you're solving
- Dig into what they did about it (nothing? bodged together a spreadsheet? paid for a rubbish tool?)
- Ask how much time or money the problem costs them
- Never mention your solution until the very end—if at all
The gold you're looking for: people who are actively trying to solve this problem and spending money or significant time doing so. If you can't find these people, you don't have a business.
Level 2: Landing page tests
Cost: £100-500 (ads + tools)
Timeline: 1-2 weeks
Confidence level: Medium
A landing page test answers one question: will strangers give you their email address (or credit card) based on your value proposition?
The setup is simple:
- Build a single page describing your product (even if it doesn't exist)
- Include a clear call-to-action: join waitlist, pre-order, or request access
- Drive 500-1000 targeted visitors via paid ads
- Measure conversion rate
What the numbers tell you:
- Under 1% email signup: Your positioning or targeting is off
- 1-5% email signup: Interesting but not conclusive
- 5%+ email signup: Strong signal worth pursuing
- Any pre-orders or credit cards: Very strong signal
Warning: email signups are cheap. People will join a waitlist for something they'll never pay for. Pre-orders with refund clauses give you much better data.
Level 3: No-code prototypes
Cost: £0-200/month (tools)
Timeline: 2-4 weeks
Confidence level: Medium
No-code tools like Bubble, Webflow, and Glide let you build functional-ish products without writing code. For certain ideas, this is brilliant.
No-code works well for:
- Marketplaces and directories
- Simple SaaS dashboards
- Content-based businesses
- Community platforms
No-code struggles with:
- Complex user flows
- Custom algorithms or calculations
- Integrations with enterprise systems
- Anything requiring real-time updates
- Products where performance matters
The real limitation: no-code often creates a janky user experience. Users might reject your product not because the idea is bad, but because the execution feels amateur. You're testing two variables at once.
Level 4: AI-assisted MVPs
Cost: Variable (your time + AI tools)
Timeline: 1-4 weeks
Confidence level: Medium to high
AI coding assistants have genuinely changed what's possible for technical founders. You can now ship real code faster than ever.
But let's be honest about the limitations:
AI is excellent at generating boilerplate, solving common patterns, and accelerating development. It's rubbish at understanding user needs, making product decisions, and creating coherent architecture.
If you're a developer, AI tools can cut your build time significantly. If you're not technical, AI won't magically make you capable of shipping production software. The gap between "AI wrote some code" and "users can reliably pay for and use this product" is still substantial.
Level 5: Concierge MVP
Cost: Your time (lots of it)
Timeline: 2-4 weeks to start, ongoing
Confidence level: High
A concierge MVP means manually delivering the service your product would automate. Want to build a bookkeeping app? Do the bookkeeping yourself for 10 clients. Want to build a recruiting platform? Be the recruiter.
Why this works:
- You learn the problem deeply
- You can charge from day one
- You discover edge cases before building
- You build relationships with early customers
Why founders resist it:
- It doesn't scale (that's the point)
- It feels like you're not building a "real" company
- It's exhausting
The concierge approach is underrated. Some of the best companies spent months doing things manually before building technology. You'd be surprised how much you learn.
Level 6: Production-ready MVP
Cost: £12,500-25,000+
Timeline: 1-5 weeks
Confidence level: Highest
Here's the uncomfortable truth: the only validation that truly counts is people paying money and coming back.
Every other method involves some form of speculation. Will people who said they'd pay actually pay? Will people who signed up for the waitlist use the product? Will users who tried a clunky prototype stick around for a polished version?
A production-ready MVP cuts through all that. You put a real product in front of real users and see what happens.
This is what we do with MVP Sprints at Precode. In 5 days, you get a deployed, functional product that real users can access and pay for. Not a prototype. Not a Figma file. Working code.
The speed matters because market conditions change, competitors move fast, and every month you spend building is a month you're not learning from real users.
The validation decision tree
Not sure which approach fits your situation? Work through these questions:
Question 1: Have you talked to 15+ potential customers about the problem?
No → Start with customer discovery. Everything else is premature.
Yes → Continue to question 2.
Question 2: What's your budget?
Under £1,000 → Landing page test or concierge MVP
Under £10,000 → No-code prototype (if it fits) or extended concierge
£10,000-25,000 → Production MVP Sprint
Over £25,000 → Full MVP Sprint with extended scope
Question 3: What's your timeline?
2 weeks or less → Landing page test or customer discovery
2-4 weeks → No-code prototype or AI-assisted build (if technical)
1-2 weeks (with budget) → MVP Sprint
Question 4: How complex is your core value proposition?
Simple (one key feature) → No-code might work
Moderate complexity → Consider concierge first to learn
Technical or complex → Production MVP is likely necessary
Question 5: Do you have technical skills?
Yes → AI-assisted building could work for simpler products
No → No-code, concierge, or outsourced MVP Sprint
When to skip straight to a production MVP
Sometimes the cheaper validation methods aren't just slower—they're actually counterproductive. Here's when to skip ahead:
Your product is the experience. Some products can't be validated with landing pages or prototypes because the core value is in how it feels to use. A janky no-code version might kill an idea that would succeed with proper execution.
You're in a competitive market. If three competitors are already chasing your space, spending 3 months on careful validation means arriving after the window closes. Speed becomes the priority.
You've done this before. Repeat founders with domain expertise don't need the same level of de-risking. You already understand the problem; you need to test the specific solution.
You have paying customers waiting. If you've run a concierge MVP and have customers asking when the real product will be ready, stop validating and start building.
The hidden cost of slow validation
Every week you spend in validation limbo has costs:
- Opportunity cost of your time
- Competitor advancement
- Market changes
- Your own motivation declining
- Potential co-founders and early employees losing interest
I've seen founders spend 6 months talking to customers, building landing pages, creating no-code prototypes—and still not knowing if their idea works. They've validated themselves into paralysis.
The goal of validation isn't certainty. Certainty doesn't exist. The goal is reducing risk enough to justify the next investment of time or money.
The 2026 validation playbook
If I were starting a new company today, here's exactly what I'd do:
Week 1: 15-20 customer discovery conversations. No pitching, just learning.
Week 2: Landing page with pre-order option. £300 in targeted ads. Measure conversion.
Week 3: If conversions look promising, run a concierge MVP with 5-10 customers. Manually deliver the service.
Week 4-5: If concierge customers are happy and willing to pay, commission a production MVP Sprint. Go from manual to automated.
Week 6+: Real users, real feedback, real data. Iterate based on actual behaviour, not speculation.
Total time to real validation: 5-6 weeks. Compare that to the founders who spend 6 months building in isolation before showing anyone.
The fastest path to real user feedback
Here's what I know after 20+ years: talking to users is valuable, but watching users actually use (and pay for) your product is invaluable.
Every other validation method is a proxy. A prediction. An educated guess.
A production MVP gives you the real thing: actual user behaviour, actual retention data, actual willingness to pay.
That's why we built the MVP Sprint the way we did—5 days to a deployed product, fixed price, no ambiguity. Because the sooner you have real software in front of real users, the sooner you know if you're onto something.
The question isn't whether to validate. It's how quickly you can get from idea to real user feedback.
In 2026, the answer should be "weeks, not months."
