How much should an MVP actually include?

The useful question is not “How little can we build?” It is “What is the smallest version that can teach us something important?” That distinction keeps an MVP from becoming either an overbuilt first release or a disposable prototype nobody can learn from.

Start with the assumption

Every early product is carrying an assumption. People will pay for this workflow. Teams will use this internal tool. Customers will trust this new way to buy something. Your first release should make that assumption visible and testable.

Write the assumption down in one sentence. Then ask what a real person would need to do for the test to mean anything. That action is usually the centre of the MVP.

Separate must-have from useful-later

A must-have is not a feature someone mentioned in a meeting. It is part of the shortest path to the learning you need. Account creation, payment, notifications or an admin view may be essential if the product cannot be tested without them. A dashboard with six filters probably is not.

• Keep the core user action intact from beginning to end.
• Include the trust, payment or operational pieces that make the test real.
• Write down useful-later ideas instead of letting them quietly enter the first scope.

Do not confuse small with disposable

If real users are expected to use the MVP, build the important path with enough care to be dependable. That does not mean designing a complete platform. It means avoiding choices that make every next change a rewrite: clear data boundaries, sensible error states and an interface people can understand.

What can usually wait?

Advanced permissions, elaborate reporting, edge-case automation, multiple integrations and a fully generalised design system often belong later. So do most “while we are here” improvements. They can be valuable, but they rarely help answer the first product question.

Know when it is ready

An MVP is ready when someone outside the team can complete the core action, the team can observe what happened and the result will change what you do next. If the release only produces opinions, it is not yet a useful test.