How do you validate a startup idea?

Define what could make the idea fail

Write down the assumptions behind the idea before collecting feedback. Who has the problem? How often does it occur? What do they do today? Who pays? Why would the new approach be meaningfully better? Which assumption, if wrong, would end the opportunity?

This prevents validation from becoming a search for compliments. The aim is to find disconfirming evidence early enough to change course cheaply.

Use an evidence ladder

• Weak evidence: people say the idea sounds useful.
• Better evidence: people describe the same problem without being led to it.
• Stronger evidence: people already spend time or money on an imperfect alternative.
• Very useful evidence: people agree to a meaningful next step such as a pilot, pre-order, introduction, data share, or scheduled trial.
• Commercial evidence: a defined buyer accepts the value, price logic, and buying process.

Build the smallest test that answers the question

A landing page can test message and demand, but it cannot prove retention. A manual service can test the outcome before software exists. A prototype can test comprehension and workflow. A paid pilot can test whether the buyer will commit budget.

Choose the test based on the uncertainty, define the decision threshold in advance, and record what happened. Validation is a sequence of better-informed decisions, not a single yes-or-no event.

Keep a decision record, not a folder of compliments

The SBA's market-research guidance distinguishes broad existing research from direct research with the customers a business intends to serve. Use both where appropriate: market statistics provide context, while buyer conversations and observed behavior answer a more specific question. Neither a large market estimate nor one enthusiastic interview proves demand for your offer.

For each test, use the following SavvyTank decision record. It is an original working framework, not an SBA certification, a universal sample-size rule, or a promise that an idea will succeed.

• Assumption: name one customer segment, the problem, the present alternative, and the observable behavior you expect if the assumption holds.
• Test: state the method, how participants will be selected, the time and cost limit, and what would count as counter-evidence. Record when your sample excludes an important buyer type.
• Observation: keep the date, source, relevant context, actual action, and limitations. Separate a requested introduction, a non-binding pilot discussion, a signed agreement, and a payment; these are different kinds of commitment.
• Decision: compare the observations with the threshold you set before the test. Choose to continue, revise, or stop, and state what remains unknown. Test willingness to pay separately from whether the product can deliver its promise.