MVP (Minimum Viable Product): The Complete Guide for Product Managers
An MVP is an experiment, not a cheap version one. How to find the risky assumption, six low-code ways to test it, and a worked example.
An MVP is the smallest thing you can build to test your riskiest assumption — an experiment, not a low-quality version one. Find the assumption that would collapse the idea if false, choose the cheapest test (landing page, fake door, concierge, Wizard of Oz, single feature or prototype), and write success and stop criteria before you begin.
- What an MVP is, and is not
- Find your riskiest assumption first
- Six ways to build an MVP without building much
- How to define your MVP
- A worked example: a study planner for exam aspirants
- MVP, prototype, beta, and MLP: what is what
- A second worked example: a fake-door test
- How to measure an MVP
- Continue, change or stop?
- MVPs inside an existing product
- Common MVP mistakes
“MVP” is possibly the most abused acronym in tech. To some people it means a half-finished product. To others it means version one. It means neither. A minimum viable product is the smallest thing you can build to test your riskiest assumption. That definition changes how you plan, how you build and how you decide what to cut.
What an MVP is, and is not
| An MVP is | An MVP is not |
|---|---|
| An experiment to learn something specific | A cheap, buggy version of the full product |
| Focused on one risky assumption | Every feature, at low quality |
| Built to be learned from, then changed or dropped | A launch you are emotionally committed to |
| Measured against pre-agreed criteria | Judged by whether it feels finished |
The word “viable” matters as much as “minimum”. If the experience is so bad that users leave for reasons unrelated to your idea, you learn nothing. The MVP has to be good enough that a failure genuinely tells you the idea was wrong, not that the execution was.
Find your riskiest assumption first
Every product idea rests on assumptions: that people have this problem, that they will pay, that they will come back, that you can build it. The MVP should target the one most likely to be wrong and most damaging if it is. Ask: “If this one thing is false, does the whole idea collapse?” That is what you test first.
Six ways to build an MVP without building much
Landing page: describe the product, measure sign-ups. Fake door: add a button for the feature and count clicks; show a “coming soon” message. Concierge: deliver the service by hand to five to ten people to learn what they truly need. Wizard of Oz: the user sees an automated product; humans do the work behind it. Single feature: build just the one action that matters. Prototype: a clickable design to test comprehension and appeal before writing code.
How to define your MVP
A worked example: a study planner for exam aspirants
Idea (illustrative): an app that builds a personalised daily study plan for students preparing for a competitive exam.
Riskiest assumption: not “can we build the planning algorithm”. It is “will students follow a plan given to them by an app, and will they pay for it?” If the answer is no, the algorithm is irrelevant.
Cheapest test (a concierge MVP): recruit 15 students through a WhatsApp group. For two weeks, hand-write each student a daily plan in a spreadsheet, send it every morning, and ask each evening whether they followed it. No code at all.
Success and stop rules, set in advance: at least 8 of 15 students follow the plan on 10 of 14 days, and at least 5 say they would pay. If fewer than 4 follow the plan, stop and rethink the idea rather than building the app.
What you learn: whether the behaviour exists, what students actually wanted in the plan (often not what you assumed), and what they would pay. Two weeks and no engineering cost. If it works, you now know exactly what to build; if it fails, you saved months.
MVP, prototype, beta, and MLP: what is what
| Term | What it is | Goal |
|---|---|---|
| Prototype | Mockup or rough demo, often not real | Test the idea and usability quickly |
| MVP | The smallest real thing that tests a core assumption | Learn whether the core idea works |
| Beta | A nearly complete product with limited users | Find bugs and refine before launch |
| MLP (minimum lovable product) | The smallest version people actually enjoy | Win early loyal users |
A second worked example: a fake-door test
Idea (illustrative): a “compare two courses side by side” feature in a learning app. Building it would take about six weeks. The riskiest assumption is that learners want to compare at all.
Test: add a “Compare” button to course pages. Tapping it shows a message: “This is coming soon. Want us to tell you when it is ready?” with a one-tap yes. No comparison feature exists.
Rules set in advance: if at least 4% of course-page visitors tap Compare, and at least a quarter of those leave their contact, build it. If under 1.5%, drop it. In between, run interviews to understand why before deciding.
Result (illustrative): 2,000 visitors in two weeks, 34 taps (1.7%), 6 leave contact. That is close to the drop threshold. Rather than build for six weeks, the team runs five short interviews and learns that learners already compare using a spreadsheet and mostly want prices in one place. The real feature turns out to be smaller than the first idea. That is the value of a cheap test: it changes the idea, not just the decision.
How to measure an MVP
- Behaviour over opinion: what people did (used it, returned, paid) beats what they said they would do
- One primary metric tied to your riskiest assumption, not ten vague ones
- A clear baseline or threshold so the result is a decision, not a debate
- Qualitative follow-up with a handful of users to understand why the numbers look as they do
Continue, change or stop?
| What you saw | What it usually means | What to do |
|---|---|---|
| Clear success against your threshold | The core assumption holds | Build the next smallest step |
| Interest, but not enough action | The problem is real, your solution is off | Change the approach and retest |
| Little interest, weak reasons | The problem may not matter enough | Stop or move to a different problem |
| Mixed signals across segments | Right for some users, wrong for most | Narrow to the segment that responded |
Deciding what each result means before you run the test is what protects you from spinning a failure as a win. It is uncomfortable to write stop rules for your own idea. That discomfort is the point.
MVPs inside an existing product
The same thinking applies to features in a live product: ship the smallest version to a small share of users, measure the one thing that matters, and expand only if it works. A staged rollout with a feature flag is simply an MVP with safety nets. The habit that transfers is not building big things on faith.
Common MVP mistakes
Building too much and calling it minimum. Building too little to be viable. Testing no specific assumption. Not deciding success criteria in advance, so any result can be spun as a win. Falling in love with version one and ignoring what the test says. Confusing an MVP with a launch: the point is learning, not marketing. And skipping the cheap options: most ideas can be tested with a spreadsheet, a form and a WhatsApp group before any code is written.
Related reading: how to prioritize features, the skills guide, and how I think about saying no in what not to build.
Frequently asked questions
What is an MVP in product management?
A minimum viable product is the smallest thing you can build or run to test your riskiest assumption. It exists to generate learning, not to be a cheap first release.
What is the difference between an MVP and a prototype?
A prototype is a mockup or rough demo, often not functional, used to test an idea or usability. An MVP is a real, if minimal, product used with real users to test whether the core idea works.
How do you decide what goes in an MVP?
Identify the riskiest assumption behind the idea and include only what is needed to test it well. Everything else waits. Also decide success and stop criteria before you start.
Can you build an MVP without coding?
Yes. Landing pages, fake-door buttons, concierge services delivered by hand, and clickable prototypes can all test core assumptions with little or no engineering.
Want this sharpened for your specific situation?
I take on a small number of career calls and resume reviews each month.
See how we could work together