1. Write down the problem and the evidence for it
Before you talk about features, settle who has the problem, how they handle it today, and what evidence says they would pay for a better way.
- Name one primary customer: a role in a type of organisation (for example, "operations managers at 20 to 200 person logistics firms"), not "small businesses".
- Describe what they do today instead: the spreadsheet, the competitor, the manual process or the outsourced service.
- Record the evidence you already have, such as interview notes, a waitlist, letters of intent, pre-orders or paid pilots, and mark which is strongest. Money committed is stronger evidence than interest expressed.
- Write down the result that would tell you the idea is wrong, before you build anything.
- Look at two or three competitors: what they charge, who they target and what their users complain about in public reviews.
If the evidence list is empty, the MVP is the experiment rather than the product, and it is cheaper to run that experiment before writing code. You can structure it with interviews, a landing page or the AI Dev Planner validation platform.