MVP (minimum viable product) development means building a product with only the features needed to test whether users find it valuable, and having real users actually use it. Unlike a proof of concept (PoC), which checks whether something is technically feasible, an MVP runs in production with real users to find out whether they keep using it and whether they will pay for it. This article explains how to cut an MVP down to size, which numbers to decide before launch, and how to decide what to do after launch. The difference between a PoC and an MVP is covered in more detail in “Where PoC and MVP fit.”
What an MVP is meant to test
An MVP is not just a small product; it is a product built to test one hypothesis. Start by writing the hypothesis down in one or two sentences:
- Example: “Front desk staff at dental clinics want to move more than half of the bookings they take by phone to the web.”
- Example: “Dispatchers at security companies spend a full day each month building the deployment roster, and would pay a monthly fee to shorten that.”
If you start from a feature list without a hypothesis, the list keeps growing, and after launch it is unclear what you actually learned. Decide at the same time who the users are (industry, company size, role) and how many companies or people you need to reach a decision. Going deep with a few companies and going wide with many people call for different features and different ways of finding users.
How to cut features
- Pick the one moment where users feel the value. A booking comes in, the roster is ready, the invoice is sent. Build only the flow that leads to that moment
- Map the flow and look for steps you can do by hand. Admin setup, data imports and billing can be done manually by your own team at first without weakening the test
- Sort features into keep, later and never. For features you are unsure about, ask whether the hypothesis can be tested without them
- Fix the time box first. Repeat step 3 until the scope fits the time you have
| Category | Examples | Treatment |
|---|---|---|
| Keep | The flow up to the value moment, login, data storage | Build to production standards |
| Do by hand | Initial setup, data migration, issuing invoices | Your team does it behind the scenes |
| Later | Fine-grained permissions, extra document types, integrations | Decide based on user feedback |
| Never | “Nice to have” features | Keep them on the list only |
Do not cut corners on what is hard to change
Even in an MVP, build the following to production standards. Fixing them later means migrating data and disrupting users.
- Authentication and accounts: the login method and how organisations (tenants) relate to users
- The data model: how each customer’s data is separated, and records that must never be deleted
- Personal data: which fields you store, for how long, and which subprocessors handle them
- Billing: if you plan to charge, how plans are structured and how invoicing works
Screen design, document details and admin usability, on the other hand, can stay simple.
Numbers to fix before launch
So you can judge afterwards whether it worked, decide the numbers and thresholds before you launch:
- Activation: the share of sign-ups who reach the value moment
- Retention: whether people active in week one are still active in week four
- Willingness to pay: people who move to a paid plan, make an enquiry or convert from a trial
- Qualitative feedback: where people got stuck and what they expected (a handful of interviews is enough)
For each number, write down “above this, we continue” and “below this, we fix.” The thresholds can be educated guesses; what matters is writing them down before launch.
Deciding after launch
After a set period of use, say four to eight weeks, judge against the numbers you fixed:
- Continue: the thresholds were met. Add the “later” features in order of how often users ask for them
- Fix: part of the funnel falls short, for example people start but do not stay, or stay but do not pay. Fix where they stall and measure again
- Pivot: the hypothesis itself was wrong. Rewrite it based on what you learned
- Stop: deciding to stop investing is also a result. Learning early keeps the cost down
What to agree when you commission an MVP
- The hypothesis and the decision numbers: share them with your development partner and use them to decide on new features
- The contract type: because features are swapped during development, MVPs are often contracted on a time-and-materials basis, period by period. In Japan, IPA publishes a model contract for agile development
- What may be rebuilt: write down which parts are deliberately kept simple
- Post-launch support: who handles enquiries, how incidents are reported and who watches the numbers
How SHANNON can help
We take on everything from framing the hypothesis to a working prototype, testing with real users and deciding whether to move into full development. See our PoC development and vertical SaaS development pages for how we work.
References
- IPA: Model transaction and contract for information systems (agile development edition, in Japanese)
- Manifesto for Agile Software Development
- The Scrum Guide
FAQ
How long does MVP development take?
It depends heavily on the number of features and whether you integrate with external systems. Fixing the time box first and cutting features to fit it fails less often than fixing the features first. If the plan runs past a few months, the scope has probably not been cut enough.
Is an MVP built to be thrown away?
Not entirely. Parts that are hard to change later, such as authentication, the data model and billing, should be built to production standards. Parts likely to change, such as screens and some workflows, can be kept simple. Agree with your development partner up front on which parts may be rebuilt.
Should an MVP be contracted as fixed-price or time-and-materials?
Because features are swapped in and out during development, MVPs are often contracted on a time-and-materials basis. In Japan, IPA publishes a model contract for agile development. If you choose fixed-price, keep the first fixed scope small and contract the next stage separately.