How to build the smallest version of a project that tests the idea and creates value for its first users.
MVP means Minimum Viable Product — but think “smallest useful project”
An MVP is the smallest functioning version that can produce credible evidence. Its purpose is not simply to launch fast or cheaply. It is to learn something important before more time and money are committed.
Start with the riskiest assumption
Before writing a feature list, identify what could make the idea fail. Do people have the problem? Will they trust the solution? Will they pay? Can the core technical mechanism work? The MVP should be designed around the biggest uncertainty, not around a miniature version of the final vision.
An MVP does not have to be software
A landing page can test demand. A clickable prototype can test understanding. A manual service can test whether people value the result before automation exists. A spreadsheet, form or concierge process can answer the business question faster than a full platform.
Build one complete value path
A focused MVP usually needs a clear target user, one main problem, one core action, a result that feels useful and a way to measure whether the result happened. It may also need accounts, payments or administration, but only when those parts are necessary to test the hypothesis.
The mistake is to take the final feature list and make every feature 20% smaller. That produces a broad but shallow project. A better MVP completes one narrow journey well enough that a real user can receive the promised value.
Leave professional-looking extras for later
Complex permissions, advanced dashboards, dozens of settings, multiple languages, perfect onboarding, internal automation and edge-case integrations often make an MVP expensive without improving the experiment. Build them when evidence shows they are needed.
Define success before launch
Finishing on time and on budget is a development success, not proof that the idea works. Decide what user behaviour would count as evidence: a payment, a completed task, repeat use, a booked call, an uploaded file, a second session or a measurable business result.
After launch, do not automatically start building everything that was postponed. Watch what users do, speak to them and decide whether to continue, change direction or stop.
The cost of an MVP depends on what must be proven
The term can describe anything from a simple landing page to a platform with accounts, payments and specialised business logic. The useful question is therefore not “How much does an MVP cost?” but “What is the cheapest credible way to test this specific risk?”
A good MVP starts with the right question, not with code.