Leave out anything that does not help test your most important assumptions or deliver the core value under investigation. Start by deciding what you need to learn, then choose the least costly experiment that can produce useful evidence. The right scope depends on whether you mean an experiment to test a hypothesis or a product released for real users: the first can be far smaller than working software; the second still needs to be valuable, usable, and feasible.
Start with the uncertainty, not the feature list
Ask: what consequential uncertainty should this release or experiment resolve? Write down the hypothesis, what observation would count as evidence, and what decision that evidence could change. Strategyzer recommends planning experiments backward from what a team needs to learn and measure (Strategyzer’s guide to testing business ideas).
As an Amazon Associate I earn from qualifying purchases.
For example, “People want this” is too broad to guide scope. A more useful question identifies a particular customer, need, and behavior the team can observe. Be clear about whether you are testing that a problem matters, that a proposed solution makes sense, or that users can complete a particular task. Those are different questions and may need different tests.
Rank assumptions by risk and evidence
List the assumptions that could make the idea fail, then assess how important each is and how strong the supporting evidence is. Give priority to consequential assumptions backed by little evidence. Strategyzer’s assumption-mapping approach helps teams separate high-importance, low-evidence beliefs from those that are less urgent to test (Strategyzer on assumption mapping).
#1 Best Overall
- Desirability: Do the intended customers have the problem, and do they want this kind of solution?
- Feasibility: Can the team deliver the experience or service with the available technology, people, and resources?
- Viability: Can the business model support delivering it?
- Adaptability: Can the idea adjust as customers, conditions, or evidence change?
Make each assumption precise enough to test. Separate beliefs about customers and their needs from beliefs about the proposed value proposition; otherwise, a negative response to a solution can leave it unclear whether the problem or the solution was wrong. Strategyzer’s value proposition guidance distinguishes these customer-side assumptions from the proposed offering (Strategyzer’s Value Proposition Canvas overview).
Choose the experiment before deciding what to build
Working software is only one possible test, and it is often not the cheapest one. Match the method to the uncertainty: a landing page can reveal responses to a proposition; a storyboard or video can communicate a proposed experience; a clickable or working prototype can test interaction or technical feasibility; and a Wizard of Oz service can test demand or workflow while people perform tasks manually behind the scenes. Strategyzer describes a range of experiment formats for testing business ideas (Strategyzer’s experiment guide).
These methods answer different questions. A landing-page response does not establish that people will use a finished product. A prototype interaction does not by itself show that the business is viable. State what a test can establish—and what it cannot—before treating its result as a verdict.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Use customer-preference exercises or split tests when the question is which of several propositions or priorities resonates more. Strategyzer discusses customer exercises and split tests for exploring preferences (Strategyzer on testing customer preferences). Stated preferences are useful evidence, but they are not the same as observed behavior.
Rank #3
Features to defer when they do not change the test
Once the learning goal and method are clear, defer work that neither improves the evidence nor enables the core proposition to be understood and experienced. Depending on the test, that may include:
- Visual polish beyond what is needed to communicate the offer credibly.
- Secondary workflows that are not part of the behavior being tested.
- Broad integrations that do not affect the target user’s test experience.
- Automation when a manual process can safely and usefully deliver the same test.
- Rare edge cases that are outside the experiment’s intended users or conditions.
This is a decision framework, not a universal checklist: a workflow, integration, or edge case belongs in scope if removing it would invalidate the test or break the core value being offered.
Rank #4
- Divergent Series Four-Book Mixed Set: Divergent, Insurgent, Allegiant, Four
Keep enough product for the result to mean something
“MVP” is used in two ways. An experiment-focused MVP can be a lightweight test rather than a product. A product-focused MVP is something users can choose to use or buy, understand how to use, and that the team can deliver with available resources. Marty Cagan of Silicon Valley Product Group sets out that product-oriented standard in his discussion of the minimum viable product.
If you are testing an actual product experience, cutting too deeply can make a negative result uninterpretable: people may reject the rough or incomplete experience rather than the underlying proposition. Keep the smallest coherent experience that lets target users understand the value and perform the behavior under test, while ensuring the team can deliver it.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.A practical way to decide what stays
- Write the learning question. Name the uncertainty that could change the product or business decision.
- List and rank assumptions. Separate customer, product, technical, and business-model beliefs; prioritize those that matter most and have the weakest evidence.
- Define useful evidence. Specify what you need to observe and what result would support, weaken, or leave the hypothesis unresolved.
- Select the smallest suitable experiment. Choose a format capable of producing that evidence, rather than defaulting to a build.
- Remove nonessential scope. Defer features that do not affect the evidence or the minimum coherent value users must experience.
- Interpret and decide. Record what happened, what remains uncertain, and whether to continue, reshape the idea, pivot, or run another test.
Common mistakes that weaken an MVP test
Assuming the first build is the best test
Building working software can be slow and expensive compared with other discovery methods. Teams may also become attached to a prototype because of the effort invested, making it harder to abandon or change. Silicon Valley Product Group discusses these discovery pitfalls and the limits of product-as-prototype approaches (SVPG on product discovery).
Using a method that cannot answer the question
An A/B test can show which option produced a different behavior, but usually cannot explain why. A usability test can reveal whether participants complete a task, but does not alone establish that they value or will buy the solution. Surveys can elicit opinions, but asking customers what to build is not a substitute for testing behavior. Use methods in combination when the decision requires evidence about both what people did and why.
Combining an uncertain problem with an uncertain solution
If both the customer problem and the proposed solution are untested, a poor response does not show which assumption failed. Where possible, investigate whether the customer problem and need are real separately from whether the proposed solution works for them.
Recommended Free Tools
Treating “minimum” as a feature-count target
There is no universal maximum number of features that makes something an MVP. Minimum means the smallest experiment capable of useful learning—or, for a released product, the smallest coherent offer users can understand and choose and the team can deliver.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
