An MVP is not a smaller version of your product — it's the smallest thing that tests your riskiest assumption. Before writing code, name that assumption clearly. Is it that people want this? That they'll pay? That you can deliver it profitably?
Build only what's needed to test that one thing. Sometimes that's a concierge service you run manually, a Figma prototype, or a single landing page. The goal is learning per week, not features per week.
Put it in front of real users fast and watch what they do, not what they say. Instrument the few actions that matter — signup, activation, and whether they come back. One paying customer teaches you more than a thousand visitors.
Then iterate on the truth you find. Kill features nobody uses, double down on the moment users light up, and only scale once you see genuine pull. Launch is not a finish line; it's the start of the conversation with your market.