A good idea is a sentence. A startup begins when another person makes time, effort or money move to solve the problem.
Describe the problem before you describe the product
Write the customer, situation, job and current cost in one paragraph. For example: 'Independent tutors spend Sunday evenings matching payments to individual lessons and message students manually when something is missing.' That is a claim you can investigate. 'An AI tutor platform for everyone' is a product pitch that hides several untested assumptions.
Ask what people currently do, when they last did it, how long it took and what broke. Past behaviour is better evidence than a polite prediction about what they might try. Talk to people who experience the work. Do not pitch the product in the first minutes. Your goal is to find how they already spend time, money or effort.
Look for action, not applause
People are generous with compliments. They are trying to be kind, they enjoy brainstorming, and they may not want to say no directly. A calendar booking, a data sample, a paid pilot or a repeated attempt to use the prototype is a more costly signal than 'this is cool'. None proves a lasting business by itself, but it narrows the uncertainty.
A landing-page signup only means the promise earned a click. A waitlist is not revenue. A demo is not a daily habit. State which uncertainty each experiment can reduce. If you need to learn whether tutors will pay for reconciliation, test a narrow manual service or a prototype around that task, then observe whether people put a real process through it.
Run an experiment this week
Write down the risky assumptions: the problem happens often, the buyer experiences it, the workaround is painful, your proposed fix fits their workflow, and someone can reach the buyer repeatedly. Rank them by uncertainty and by what happens if you are wrong.
Choose a test that is cheap and ethically honest. A clickable mock can test whether users understand a flow. A manual concierge service can test whether an outcome matters before writing an automation. A small paid pilot can test willingness to budget. Never pretend a prototype has a feature it lacks. Ask clear consent before collecting personal data.
After each conversation, record direct quotes separately from your interpretation. Look for what surprised you and what evidence would change your mind. Talk to more than one kind of potential customer before concluding that the loudest person's use case is the market.
Build only after the next question is clear
When the problem recurs and people already work around it, build the thinnest tool that removes one measurable step. Avoid choosing a technology because it sounds fundable. Add a language model only when handling the real input needs flexible language or an uncertain decision. If users can solve it better with one field and a rule, give them the simple thing.
Set one meaningful first target: hours saved, tasks completed without rework, or whether the user returns next week. Watch a real person use it. Fix the part that prevents repeat use before adding more features.
Starting a company has legal, financial, tax and regulatory details that depend on country and circumstances. This guide is about testing customer demand, not a substitute for qualified local advice on the rest.
FOLLOW THE EVIDENCE
Sources & further reading
Process note: AI tools assisted with research and drafting. Sources are linked so you can inspect the evidence. Unless an essay says otherwise, it does not claim hands-on testing or independent reproduction.
