Choosing an MVP type is a decision about what you need to learn, not which tool looks most impressive. This guide compares practical formats and shows how to choose the smallest honest experiment that can answer your next business question.
Which MVP Type Fits Your Idea?
Use a landing page to test a message and an initial commitment, a concierge service to observe delivery and customer needs, and a single-feature product when real use of the technology is essential. A Wizard of Oz approach can test a simple interface while people perform some work behind it. No-code is a building method rather than a separate kind of evidence, and no format can validate every assumption at once.
Start with the uncertainty, not the software
Write down the decision you expect the experiment to inform. Perhaps you do not know whether a particular buyer recognizes the problem, whether users will complete a workflow, or whether a promised result is technically achievable. These are different uncertainties and call for different tests.
A page collecting email addresses can reveal willingness to hear more. It cannot demonstrate that customers will use a product repeatedly. A manual service can expose a valuable workflow but cannot establish that automation will be accurate or economical. A technical demonstration may establish feasibility without proving demand.
The Lean Startup principles emphasize validated learning and experiments rather than treating a product launch as the end of learning. Apply that idea by matching the evidence you collect to the claim you want to make. Avoid describing every positive response as validation.
Also identify the audience. An enthusiastic friend is not equivalent to the person who owns the budget. If your intended customer is an operations manager at a small manufacturer, recruit people with that responsibility rather than an undifferentiated list of technology enthusiasts.
Compare the main MVP formats
The labels below are working categories, not formal standards. Teams often combine them, but each format has a characteristic strength and a predictable blind spot.
| Format | Best question to explore | Evidence it can provide | What remains uncertain |
|---|---|---|---|
| Landing page | Does this audience respond to this promise? | Qualified inquiries or sign-ups | Delivery quality and continued use |
| Concierge service | Does solving the problem help? | Observed workflows and real outcomes | Automation and delivery economics |
| Wizard of Oz | Does this interaction fit the workflow? | Use of an interface with manual support | Fully automated performance |
| Single-feature product | Does the core capability work in practice? | Task completion and repeat usage | Demand for a broader product |
| No-code implementation | Can existing components deliver the workflow? | Behavior in a working service | Limits of the chosen platform |
Think of the table as a filter rather than a ranking. A landing page is not inferior because it contains less code, and a working app is not superior merely because it took longer to build. The better choice is the one that reduces the relevant uncertainty without introducing avoidable cost or misleading participants.
When a landing page is enough
Choose a landing page when your immediate question concerns positioning, audience, or initial interest. Explain who the offer is for, what problem it addresses, and what action a visitor can take. If the product does not exist yet, make that clear before someone submits information or payment.
Use a call to action that matches the commitment you need. Joining an updates list is different from requesting a discovery call, applying for a pilot, or paying a clearly explained deposit. Stronger commitments can provide more relevant evidence, but they also create greater obligations. Do not invent scarcity or present an unavailable product as ready to use.
Interpret conversion in context. Traffic source, visitor suitability, message clarity, and the requested action all affect results. There is no universal sign-up rate that proves a market exists. Track qualified responses and follow up to understand why people acted, rather than treating one percentage as a verdict.
A landing page becomes insufficient when the open question is about the experience after signup. If you need to know whether a report changes a manager's weekly planning, give someone a report and observe what happens. More page variants will not answer a delivery question.
When concierge or Wizard of Oz works better
A concierge MVP delivers the benefit through visible human work. A founder might manually prepare inventory recommendations and discuss them with a shop owner. The interaction makes it easier to discover missing information, exceptions, and the language customers use to describe success.
Paul Graham's essay on doing things that do not scale describes the value of manually recruiting users and providing unusually attentive early service. The point is to learn and help users, not to assume that labor-intensive delivery is already a scalable business.
A Wizard of Oz setup puts an interface in front of some manual operations. A customer submits a request through a form and receives a result, while a person coordinates the work behind the scenes. This can reveal whether the interface and response fit an existing routine before you automate every step.
Do not use the format as permission to deceive. Disclose human access to sensitive information, avoid unsupported automation claims, and explain service limitations. If human intervention materially affects privacy, safety, or the buyer's decision, it belongs in the customer's understanding of the service.
For both approaches, record manual effort and exceptions. If a report requires repeated expert judgment, that is useful evidence about the future product and its economics. Do not hide founder labor simply because you are not paying yourself yet.
When to build a single feature or use no-code
Build a narrow working product when the core uncertainty requires actual software performance. A file converter, for example, may need to process representative files correctly inside the user's workflow. A polished explanation cannot establish compatibility or reliability.
Keep the scope narrow without removing necessary protections. Account creation may be unnecessary for a temporary conversion tool, but access control may be essential when users upload confidential business records. Minimum scope means excluding unrelated capabilities, not ignoring the responsibilities created by the chosen workflow.
No-code and low-code tools can support forms, scheduling, databases, notifications, and simple workflows. They can be appropriate for a concierge service, an interface backed by manual work, or a working single-feature product. The implementation tool does not determine which hypothesis you have tested.
Before choosing a platform, check data export, permissions, integrations, usage limits, and how you would move away from it. Model the expected workflow with representative records. A quick build that traps customer data or cannot represent an essential permission rule may create more work than it saves.
A practical selection process with a worked example
-
Name the risky assumption. Why: Teams otherwise default to the format they already know. How: Finish this sentence: “We need evidence that this customer will do this action under these conditions.” Keep it narrow enough to observe.
-
Choose the closest observable behavior. Why: Opinions and actions answer different questions. How: Decide whether you need an inquiry, a completed task, a payment, or repeated use. Record what that behavior does not establish.
-
Choose the least elaborate credible delivery. Why: Extra infrastructure can delay learning. How: Use the comparison table, then identify legal, privacy, safety, and service constraints that rule out shortcuts.
-
Set a decision rule before recruitment. Why: It is easy to reinterpret disappointing results afterward. How: Define what would lead you to continue, change the offer, or investigate further. Treat small samples as directional rather than representative market estimates.
Consider a hypothetical service that helps independent shops prepare reorder lists. The founder has two uncertainties: whether owners will share inventory information and whether the recommendations are useful. A landing page can recruit pilot participants, but it cannot answer the second question.
The founder therefore chooses a concierge pilot. Hypothetically, six owners agree to participate, four provide usable records, and three request another report. Those numbers suggest questions worth investigating; they are not proof of broad demand. Why did two owners decline to share records? Did repeat requests reflect useful recommendations or personal goodwill?
If the manual reports genuinely influence purchasing, the next test might automate one recommendation step. If participants like the reports but never act on them, building a dashboard may be premature. The format changes because the uncertainty changes, not because every startup must climb a fixed ladder from page to app.
Questions About MVP Type Selection
Can I combine two MVP types?
Yes. A landing page can recruit users for a manual service, and a narrow interface can support concierge delivery. Keep the evidence from each stage separate so that an initial signup is not confused with successful product use.
Should I charge during a manual pilot?
Charge when payment is relevant to the question and you can clearly explain what customers will receive. Agree on delivery, cancellation, and refund terms. Free participation may help investigate usability, but it leaves willingness to pay unresolved.
What if competitors already offer full-featured products?
Their feature lists do not determine your experiment. Focus on a specific underserved workflow or customer segment and provide a credible outcome there. If the proposed difference has no importance to the buyer, additional features will not necessarily fix that problem.
When should I stop doing the work manually?
Consider automation when recurring tasks and exceptions are understood well enough to specify, and when manual delivery is limiting a worthwhile service. Retain human review where mistakes would be costly. Automating an unclear process can make its problems harder to see.
Choose the Smallest Useful Test
Choose the format that produces relevant evidence with the fewest unnecessary commitments. Match the experiment to one uncertainty, describe its limitations honestly, and let the result determine the next step.
Next steps: Read what an MVP means, plan how to scope and ship an MVP, or explore the MVP Building hub.
Sources
- The Lean Startup — Principles
- Paul Graham — Do Things that Don't Scale
This guide is general education, not professional business or financial advice.