All articles
DevelopmentNovember 12, 202611 min read

How to build an MVP: scope, timeline and what actually drives the cost

A practical guide to scoping a minimum viable product: what belongs in the first release, what drives cost and timeline, and what to ask a contractor before signing.

An MVP is the smallest version of a product that lets a real user complete one valuable scenario from start to finish, so you can find out whether anyone wants it. To build one, pick that single scenario, cut everything that does not serve it, and release to real users as soon as the result is reliable. Cost and timeline depend less on the idea than on how many roles, integrations and platforms the first version has to carry.

What is an MVP, and what is it not?

A minimum viable product is a working product with a deliberately narrow scope. Both words matter: "minimum" means one scenario instead of ten, "viable" means that scenario works properly, with real data and real users.

It is often confused with two other things:

  • It is not a prototype. A prototype is a clickable mock-up that shows how the product will look and helps everyone agree on the screens. Nobody can do actual work in it. An MVP is live, and people use it.
  • It is not a cheap version of the full product. Building every planned feature at low quality gives you a large product that works badly. You learn nothing from it, because you cannot tell whether users leave because of the idea or because of the bugs.

The purpose of an MVP is knowledge: does your audience have this problem, do they solve it with your product, will they come back or pay.

How do you cut the scope down to one core scenario?

Start with a single sentence: who the user is, what they do in the product, and what they get out of it. For example: "A clinic patient books an appointment with a specific doctor online and receives a confirmation." Every feature is then tested against that sentence.

A sequence that works:

  1. Write down every feature you have in mind, without filtering.
  2. For each one, ask: can the user complete the core scenario without it? If yes, it moves to the "later" list.
  3. For what remains, ask: can a manual process replace this at first? Invoices sent by hand, onboarding done on a call, a report assembled in a spreadsheet are fine at low volume, and they show you what is really worth automating.
  4. Check what is left against the sentence once more.

The "later" list is not a bin. It is a backlog you will prioritise by what users actually do after launch.

What goes into the first release, and what can wait?

There is no universal answer, but the typical decisions look like this:

  • Authentication. Needed if users have personal data or history. Keep it simple: a link or code sent by email, or sign-in through one existing provider. Elaborate permission matrices and corporate single sign-on can wait.
  • Payments. Include them if willingness to pay is the thing you are testing. If you are testing whether people use the product at all, or the first customers are few and can be invoiced, defer them.
  • Admin panel. A minimal one: see users and key records, fix a mistake by hand. A full back office comes later.
  • Analytics. Do not defer it. A handful of events along the core scenario, showing where people start and where they drop off, is what the MVP exists for. Polished dashboards can wait.
  • Notifications. Only the ones the scenario breaks without: confirmations and sign-in codes. Newsletters, push campaigns and fine-grained preferences come later.

Some things are not up for discussion: backups, HTTPS, basic protection of personal data, error logging. Users never see them, but an MVP that loses data teaches you nothing.

What actually drives the cost and timeline of an MVP?

Not the number of screens as such. The main drivers:

  • Number of user roles. Each role (customer, provider, manager, administrator) brings its own screens, permissions and test cases. A two-sided marketplace is closer to two products that meet in the middle.
  • Integrations. Every external system (CRM, accounting, delivery, messengers, maps) adds work and uncertainty: its documentation, limits and quirks are outside your control.
  • Payments. Subscriptions, refunds, payouts split between participants and multiple currencies each add logic and edge cases on top of a one-off checkout.
  • Design depth. A clean interface built on a ready-made component library and a custom design system with illustration and motion are different amounts of work.
  • AI features. Calling a model is quick. The time goes into making the answers reliable: preparing data, evaluating quality, handling wrong answers, keeping running costs under control. More on that in where to start with AI integration.
  • Platforms. A web app, mobile apps and a mini app inside a messenger are each counted separately. A responsive web app is usually the most economical start; native apps are justified when you need device capabilities.
  • Data and industry requirements. Migrating existing data and working in regulated fields such as healthcare or finance increase the scope.

That is why we do not name a price before these points are clear. How we get from a brief to a fixed scope is described in how we estimate a project.

What phases does MVP development go through?

  1. Discovery. Understanding the task: who the user is, which scenario is the core one, what already exists, what the constraints are.
  2. Prototype and architecture. A screen flow, a data model and a choice of integrations, so both sides share one picture of what will be built.
  3. Fixing the scope. The feature list, cost and timeline are agreed in writing before development begins.
  4. Development in short iterations. Each ends with a working result you can open and check.
  5. Testing and launch preparation. The scenario is verified on real data and devices, and the environment is set up.
  6. Launch and observation. First users, analytics, fixes, and a decision about what comes next.

Each step is covered in more detail in the stages of web development.

No-code or custom code: which should you choose?

No-code and low-code platforms are a legitimate way to test an idea, not a lesser form of development.

Where they are strong: quick results in standard scenarios (forms, catalogues, simple marketplaces, internal tools), no development team needed to start, and little regret if you throw the result away.

Where they are limited: non-standard logic, behaviour under load, complex integrations, platform fees as you grow, and dependence on the vendor. Leaving a platform usually means rebuilding, not exporting.

Custom code takes more effort up front, but you own the code, the logic can be anything, and growth has no ceiling.

As a rule of thumb: if you are testing demand for something standard and are prepared to discard it afterwards, use no-code. If the value of the product lies in non-standard logic, integrations, data handling or AI, or you already know the first version has to grow into the main product, you need custom web development. And before either option, demand can often be tested with a landing page and a service delivered by hand.

How does AI-assisted development change the speed, and what does it not change?

AI coding tools do speed up writing code: routine screens, forms, data operations, tests, documentation. What used to take most of the time at the keyboard now takes less, so a first version can reach users sooner.

What they do not change:

  • Requirements. Only you and your users know what needs to be built. A vague task produces a fast, vague result.
  • Architecture decisions. The data model, the boundaries between modules, what to buy and what to build: mistakes here are expensive no matter how quickly the code is typed.
  • Testing. Generated code has to be verified like any other code, on real scenarios and edge cases.
  • Responsibility for the result. A person answers for security, data and the product working in production. A tool does not.

So the useful question for a contractor is not "do you use AI" but "who reviews the code and who is responsible for the result".

What should be on the checklist before launch?

  • The core scenario works end to end in the production environment, on a phone and on a desktop.
  • Analytics events fire, and you can see them in a report.
  • Backups are configured, and restoring from one has been tested.
  • Errors are logged, and someone hears about the serious ones immediately.
  • Domain, HTTPS and delivery of transactional email are checked.
  • A privacy policy, terms of use and consent to data processing are in place if you collect personal data.
  • Accounts for hosting, the domain, the repository and the payment provider are registered to your company.
  • Users have a way to contact support, and someone answers.
  • The first users are known by name: you have a list of people to invite on day one.

What mistakes do founders make most often?

  • Features added "just in case". Each one pushes the launch back and blurs the conclusion.
  • Several audiences at once. A product for everyone tests a hypothesis about no one.
  • No success criterion. If you do not write down before launch what counts as a good result, any numbers can be read either way.
  • Polishing instead of releasing. A second week spent on animation brings you no closer to the main answer.
  • Launching without analytics. Impressions replace data, and the next step is chosen blind.
  • "It is only an MVP" as an excuse for unreliability. Narrow scope is fine; lost data and broken payments are not.
  • No budget for life after launch. The first weeks of real use always bring changes. An MVP is the start of the work, not the end of it.

What should you ask a contractor before signing?

  • What exactly is in scope, in writing? And what is explicitly out of it?
  • What would you remove from my list, and why? A contractor who agrees with everything is not thinking about your budget.
  • Who will write the code, and can I talk to them?
  • How, and how often, will I see the result?
  • Who owns the code and the accounts, and when do I get access to the repository?
  • What happens when requirements change along the way?
  • What happens after launch: bug fixes, support, further development?
  • Which of your live products can I open right now?

How to compare the contractors themselves is a separate topic: freelancer, studio or in-house team.

Where to start

Write the one sentence about your core scenario, list the features, and split them into "does not work without it" and "later". With that list, tell us about the task: we will go through the scope together, say what we would cut, and explain what the estimate will depend on.

FAQ

How long does it take to build an MVP?

There is no single figure: the timeline follows the scope, meaning the number of roles, integrations and platforms. An honest date can only be given once the scope is fixed, and the most reliable way to shorten it is to cut features, not to add people.

How much does an MVP cost?

The cost is driven by user roles, integrations, payments, design depth, AI features and the number of platforms. Ask for an estimate broken down by module, so you can see how much dropping a specific feature saves.

What is the difference between an MVP, a prototype and a proof of concept?

A proof of concept checks whether an idea is technically feasible. A prototype shows how the product will look. An MVP is a working product used by real people.

Can I build an MVP without a technical co-founder?

Yes, with a contractor or on a no-code platform. But product decisions must be made by someone on your side, and the code and accounts must be registered to you.

Will the MVP have to be rewritten from scratch later?

Not necessarily. An MVP built on a sound data model and mainstream technology usually keeps growing; versions made on no-code platforms or rushed without any architecture are the ones more often rebuilt.

Should an MVP start as a mobile app or a web app?

More often as a web app: it runs on any device and needs no app store review. A mobile app is justified when the core scenario relies on device capabilities such as the camera, geolocation or offline use.

Shall we discuss your project?

Tell us about it — we'll come back with an estimate and a proposal.

WhatsAppTelegram