How to validate a startup idea before building an MVP

By Dmitry Volopianskii — product entrepreneur, consultant and founder of Lectoroom. Prepared September 25, 2026.

To validate a startup idea, identify the assumption that would make the business fail, collect evidence from a specific customer group and test a meaningful commitment. Decide in advance what evidence warrants the next investment. The outcome is a reasoned decision about your next experiment, with uncertainty still attached.

For a B2B (business-to-business) founder, that means connecting a user's problem to a buyer's decision. An enthusiastic demo attendee might never control a budget or get permission to change a workflow. This guide focuses on that gap and gives you a worksheet to resolve it before commissioning an MVP.

A minimum viable product (MVP) is a minimal product used to test a core assumption in practice. Here, “building an MVP” means developing working software; a manual pilot is sometimes also described as an MVP. Detailed feature prioritization comes after the decision this guide addresses.

On this page: decision · buyer map · discovery · pilot · evidence worksheet · FAQ.

How to validate a startup idea by defining the next decision

Replace “Is this a good idea?” with a decision small enough to make from the evidence you can obtain. For example: “Should we invest in a pilot for agency operations teams that prepare recurring client reports?” That question gives the research a boundary.

Write down three things before recruiting anyone:

  • The customer situation you are studying.
  • The assumption you most need to challenge.
  • The investment you will consider if the evidence supports it.

The immediate risk might be demand, access to customers, technical feasibility or delivery effort. Testing the easiest assumption is unhelpful if another one can invalidate the whole proposition. A technically novel product may need a small feasibility experiment alongside discovery; a working prototype does not establish a customer's willingness to adopt it.

Y Combinator's discussion of prioritization and startup ideas identifies building before talking to users as a common founder mistake. The practical application here is to make each piece of work answer a customer or business question, including the work you do before coding.

A running example, not a customer success story

Imagine a founder considering a tool that prepares weekly client reports for small agencies. Everything in this example is hypothetical. We do not yet know that reports are painful, that agencies will pay or that automated reports are the right solution.

The first question is narrower: when an agency prepares a report, what consumes effort that someone is motivated to remove? Data collection, explaining results and obtaining approval are different problems. A product that automates the wrong step can be impressive in a demo and irrelevant in practice.

Who needs the product, who pays and who can block adoption?

A useful B2B idea validation exercise maps the people involved in a change. One person may hold several roles in a small company. In a larger organization, the same proposal can pass through several independent decisions.

RoleWhat to establishEvidence to record
UserWho performs the recurring task?A recent workflow walkthrough
Internal advocateWho will organize a trial?An agreed introduction or scheduled working session
BuyerWho can approve the spend?A discussion of the actual buying process
GatekeeperWho controls access or operational requirements?Identified constraints and an owner who can resolve them

In the agency example, an account manager may dislike report preparation, while the agency owner cares about whether clients understand the outcome. A data administrator might control access. Interviewing only the account manager leaves major assumptions unresolved.

Also map the current alternative. It could be an established reporting tool, a spreadsheet, an analyst or accepting the inconvenience. Record why the team keeps using it. Your competing product is being compared with the total cost of switching, including training and workflow disruption, rather than a feature list alone.

What should customer discovery interviews establish?

Ask people to reconstruct a recent task and its consequences. Then separate what you observed from what you inferred. Rob Fitzpatrick's The Mom Test addresses the problem of biased customer feedback and practical ways to learn from conversations. The following prompts are an original interview outline for the reporting example:

  1. “Walk me through the last report you sent to a client.”
  2. “Where did the information come from, and who checked it?”
  3. “What changed between the first draft and the version you sent?”
  4. “Which part have you tried to improve already?”
  5. “What happened the last time you adopted a tool for this workflow?”

With permission, examine a redacted example. You may discover that the difficult work is interpreting a result, not moving data. That observation should change your proposed test before it becomes a development backlog.

Use a consistent note format: situation, current method, consequence, attempted remedy, decision participants and unresolved questions. Preserve contradictory evidence. A team that already finds reporting easy is valuable information about the boundary of your segment.

How many interviews are enough?

There is no interview count in this guide that certifies a market. You can schedule an initial batch of five interviews to make the work manageable, then decide what uncertainty remains. Five is a suggested work unit, not a research finding or a representative sample.

Keep recruitment sources visible. Introductions from your own network may make access easier while masking the difficulty of reaching unfamiliar buyers. If the business depends on a new acquisition channel, test your ability to find appropriate prospects through that channel separately. Do not infer acquisition economics from a handful of friendly conversations.

What commitment should you test before building an MVP?

Choose a next step that resembles adoption closely enough to answer your question. A newsletter signup might reveal interest in the topic. It does not establish willingness to change a reporting process. For the agency example, a better operational test could be agreeing to prepare one real report together, subject to permission and suitable data access.

Paul Graham's essay Do Things that Don't Scale describes hands-on recruitment and early work that founders perform manually. You can use that approach to learn about delivery before automating it. Record your effort as well as the customer's response: personal attention can produce a result that the proposed software cannot yet reproduce.

Agree on a short pilot brief:

FieldQuestion to resolve
Real taskWhich upcoming report or business event will the pilot support?
ParticipantsWho uses the output and who decides whether to continue?
InputsWhat information is required and who can authorize access?
OutcomeWhat observable change would make the trial useful?
Commercial termsWhat is paid or unpaid, and what decision follows the trial?
Review dateWhen will both sides examine the evidence?

Be clear about what is manual, experimental and unavailable. A paid pilot can provide evidence about this particular offer, but it is not proof of retention or a scalable business. An unpaid pilot can still test workflow adoption; it leaves willingness to pay unresolved.

If buying takes longer than your research window, distinguish “not purchased yet” from “rejected.” Ask for the next legitimate step in that organization's process. A named buyer and scheduled review help clarify the path, but remain weaker evidence than a completed commercial agreement.

How do you turn evidence into a build, revise or stop decision?

Decide how you will interpret results before running the test. Strategyzer's Test Card makes the hypothesis, experiment, measure and success threshold explicit. Extend that discipline by recording alternative explanations and the next decision, so a positive result cannot quietly become permission to build everything.

Copy this worksheet for each important assumption:

FieldYour entry
AssumptionWhat must be true for this offer to work?
Eligible customerWho belongs in the test, and who does not?
Test actionWhat will the customer actually do?
Evidence sourceWhere will the observation or agreement be recorded?
Continue conditionWhat justifies a further, bounded investment?
Revise or stop conditionWhat would change the decision?
Remaining uncertaintyWhat will this test still leave unanswered?

Illustrative planning thresholds, not benchmarks: invite six eligible agencies to discuss one upcoming reporting cycle. Decide beforehand to investigate a manual pilot if two agree to a working session involving the relevant decision-maker and authorized inputs. This condition tests access to a pilot. It does not validate revenue, and declining to provide data may reflect access constraints rather than lack of demand.

After the pilot, make a separate decision about development. Look at whether the output was used, which parts required your intervention, whether the buyer wants the outcome again and under what terms. If each agency needs a different bespoke service, narrowing the offer may be more useful than adding configuration settings.

What belongs in the handoff to an MVP team?

Hand over the tested situation, observed behavior, remaining risk and the next measure. Include the things you still do manually, since hiding them creates misleading expectations about the product. A developer should be able to explain what you expect to learn from the first release.

For planning the next investment, review what drives MVP development cost. Scope should follow the learning requirement. This article's decision worksheet establishes why a build might be justified; detailed feature prioritization is a separate exercise.

Startup idea validation FAQ

Can I validate demand before building any software?

You can investigate a real problem, reach prospective buyers and test an offer through a transparent manual pilot. That can justify a small software experiment. It cannot establish how the finished product will perform, whether customers will keep using it or whether acquisition will remain affordable at a larger scale.

Does a waitlist validate a startup idea?

A waitlist records a particular action in response to a particular message. Its meaning depends on who signed up, why they arrived and what they understood the offer to be. Follow up with an appropriate next step and track participation. Do not treat an email address as a purchase commitment.

Should a B2B pilot always be paid?

Use payment when the question is whether a buyer will spend money on the defined offer. Use an unpaid trial only when its learning goal is explicit, such as testing workflow access or output usefulness. Record willingness to pay as unresolved and agree when the commercial discussion will happen.

Can AI validate my business idea for me?

AI can help draft interview prompts, propose competing explanations and organize notes you are authorized to use. Check its output against the underlying evidence. Generated customer personas and simulated interviews are hypotheses, not customer behavior. They cannot establish a buyer's budget, purchasing authority or willingness to adopt your product.

You have learned enough to make the next decision when you can connect customer evidence to a bounded investment and name what remains uncertain. Keep that reasoning with the project. It will be more useful during development than a slide saying the idea is validated.

Author and source notes

Dmitry Volopianskii is the founder and Product Lead of Lectoroom, which he has developed since 2013. His work includes product strategy, monetization, user journeys and product operations. See the Lectoroom product case study for the scope of that work.

Methodological sources checked September 25, 2026: Y Combinator, Rob Fitzpatrick, Paul Graham and Strategyzer. The agency scenario and decision worksheets are editorial examples, not reported customer research. The author's commercial interest is product consulting; this article contains no affiliate links.