Most teams don't fail at building an MVP. They fail at deciding what it's meant to prove. This guide covers what an MVP is, how to build one, how long it takes, what it costs, and the mistakes that turn a six-week experiment into a two-year product nobody asked for.
What an MVP Is (and What It Is Not)
An MVP is the smallest thing you can build that lets a real decision be made on real evidence. Not the cheapest version of your product — the fastest answer to your riskiest assumption.
Write down what you'd do if it fails. If the honest answer is "carry on building anyway", you aren't running an experiment, you're shipping slowly.
What an MVP is not:
- A prototype. It's shown to people. An MVP is used by them, with real consequences — real money, real data, real work.
- Phase one of the full build. Phase one assumes the plan is right. An MVP exists to find out whether it is.
- A cheap version. What shrinks is scope, not standards.
- A demo for investors. That's a pitch asset. Different job.
The useful question isn't "what's the minimum we can build?" It's "what's the minimum we can learn from?"
Types of MVP
Four options, ordered by how little you have to build.
No-Code MVP
Assembled from existing tools — forms, automations, a site builder, a database with a front end. Right when the risk is demand rather than feasibility: you're fairly sure you can build it, but not that anyone wants it.
It breaks down on complex permissions, regulated data, audit trails and real volume. The usual trap is success: the tool works, the business grows, and two years later the whole operation runs on something nobody can modify.
Concierge and Wizard of Oz MVP
Concierge: you deliver the service by hand, openly, to a handful of customers. No product at all. Wizard of Oz: the customer sees a working product; behind the interface, people do the work manually.
These are the most underused and often the most informative. They test whether people will pay, and they teach you the real process before you automate it — which matters, because the process you'd automate from a whiteboard is almost never the one that actually runs.
Limits: they don't scale, and can't test anything where the automation itself is the value.
Custom-Built MVP
Built from scratch. Necessary when regulation, integrations with existing systems, data ownership or IP make the other routes impossible — which covers most of fintech, healthcare, logistics and anything selling into a large organisation.
Slower and more expensive, and the only honest option when the thing being tested is the system. See our web app development work.
MVP Development Process Step by Step
Discovery and Hypotheses
Write down the assumption that, if wrong, makes everything else pointless. Usually one of three: nobody wants this, they want it but won't pay, or we can't deliver it at a workable price.
Then write the number that settles it, and the threshold, before you build. "We'll see how it goes" is how projects become permanent. Our product discovery process covers how we run this.
Feature Prioritization
One job done completely beats five jobs done halfway. A user who can't finish anything tells you nothing except that your product is unfinished.
Pick the single path from arrival to value and build it end to end. For everything else, strike it and ask what breaks: if the answer is "the demo looks thinner", it goes; if it's "the user can't finish the task", it stays.
What survives more often than expected: login and roles, error handling, and a way for your team to see and fix data. What rarely survives: settings, notifications, dashboards, and the second user type.
Design and Prototype
Design the flow and the states, not the polish. Empty, loading, error, no permission — users meet these constantly, and a build without them leaves developers inventing answers under deadline.
A clickable prototype tested with five people surfaces more problems in an afternoon than a month of internal review.
Skip the design system, illustration, animation and marketing pages. Don't skip the exact copy on screens where someone commits — payment, submission, approval. That wording is the product.
Development and Testing
Thin but real. An MVP can have fewer features than planned. It cannot have wrong numbers, lost data or broken payments — those don't produce a failed experiment, they produce a wrong answer.
Non-negotiable even in the smallest build: authentication, correct money handling, data integrity, and a record of who changed what.
Automated coverage on anything touching money or data, manual testing for the rest. A full regression suite on an MVP is a sign you're building something else.
Launch and Feedback Loop
Instrument before you launch. Decide which events you record while the code is written, or the first month of data is a guess.
Launch to a group small enough to talk to individually — ten engaged users beat a thousand who signed up and left. Run both halves: numbers tell you what happened, conversations tell you why.

How Long Does It Take to Build an MVP
Typical ranges, assuming a focused scope and a client who can make decisions:
- Concierge or Wizard of Oz — days to two weeks
- No-code — two to six weeks
- Custom-built, one user type — 8 to 12 weeks
- Custom-built, regulated or heavily integrated — 12 to 20 weeks
The build itself rarely moves these numbers. What does:
- Integrations. Every external system has its own timeline and you don't control it. Sandbox access alone can take weeks.
- Compliance and security review. In regulated industries this is a phase, not a formality.
- Approvals. A decision that waits four days for a two-minute answer, repeated weekly, adds a month over a quarter.
- Content and data. Routinely treated as trivial, routinely aren't.
The single biggest variable is how fast your side decides. A team that answers questions within a day consistently ships weeks earlier than one that batches them into a fortnightly meeting.
How Much Does an MVP Cost
The range is enormous, and anyone quoting a precise figure before scope is set is guessing. What's useful is knowing which levers move the number.
Indicative ranges for a custom-built MVP, by where the team is:
- Eastern Europe — $25,000 to $70,000
- Western Europe and the UK — $60,000 to $150,000
- North America — $80,000 to $200,000 and up
No-code builds usually land in the low thousands plus ongoing subscriptions. Concierge MVPs cost mostly time.
What moves the number, in rough order of impact:
- Number of user types. Each is a set of screens, permissions and states. One to three roughly doubles the work.
- Integrations. Each external system is scope you don't control.
- Regulation. Audit trails, data residency, consent, reporting — real engineering, rarely in the first estimate.
- Money movement. Payments, payouts and refunds carry correctness requirements nothing else does.
- Design depth. A design system pays off later; for an MVP it's usually deferred.
Costs that aren't in most quotes:
- The admin panel — often close to a third of the work, and rarely demoed
- Authentication, roles and permissions
- Error, empty and partial states
- Infrastructure, environments, monitoring
- The first three months after launch
A better question than total cost: what does one validated decision cost? An MVP at $40,000 that conclusively kills a bad idea is cheaper than a $15,000 one that leaves you guessing.
Common MVP Mistakes
Calling the full product an MVP. The most common and most expensive mistake. If it has every feature from the plan, the "minimum" is decoration.
No hypothesis, so no possible failure. Without a number agreed in advance, every result reads as encouraging.
Skipping the internal side. No admin tools means your team does it by hand. Fine for a month, a hidden salary by year two.
Launching to nobody. A quiet launch isn't a result. You've learned nothing about the idea, only about your distribution.
Treating MVP code as disposable and then keeping it. Almost every MVP that works becomes the production system. Build the parts you'll keep — data model, auth, money handling — properly, and be deliberately rough everywhere else.
MVP Examples
Dropbox tested demand with a video. The product didn't exist; the sign-up list proved people wanted it before anyone built sync.
Airbnb started with the founders' own flat, photographed by hand, payments arranged manually. A concierge MVP in all but name.
Zappos photographed shoes in local shops and bought them at retail when an order came in. Loss-making per sale, and it answered the only question that mattered: will people buy shoes unseen?
None of them built the hard part first. Each found the cheapest honest way to test whether the hard part was worth building.
MVP Launch Checklist
- One sentence stating the assumption being tested
- The number that decides it, written down before launch
- A single user journey, complete from arrival to value
- Empty, loading, error and no-permission states designed
- Authentication and roles working
- Money handling correct and tested, if money is involved
- Admin access for your team to see and fix data
- Analytics events firing before launch
- A named group of first users you can contact
- A date in the calendar to review and decide what happens next
U1CORE is a product digital studio that works like consultants. Partner in Crime for strong businesses whose digital and operations don't yet play at their level. Websites · Product · Branding · Software — see our product strategy work, or book a strategy call.









