At the heart of the Lean Startup is the Build-Measure-Learn feedback loop. Teams build an experiment, usually in the form of a Minimum Viable Product, measure the results with data, and learn whether their hypotheses are correct. Critically, the loop should be planned in reverse: start by deciding what you want to learn, then determine what you would need to measure to confirm or refute that learning, and only then figure out what to build. This discipline prevents teams from building products nobody needs.
A Minimum Viable Product, or MVP, is the simplest version of a product that allows a team to collect the maximum amount of validated learning about customers with the least effort. Its purpose is to test a core hypothesis about the product with real customers as quickly and cheaply as possible, not to build a polished product. Validated learning is the process of demonstrating empirically that a team has discovered valuable truths about present and future business prospects through experiments, rather than relying on opinions or untested assumptions.
There are several common types of MVPs. A landing page MVP is a single webpage describing the product with a sign-up button or email capture that measures interest before any code is written. A Wizard of Oz MVP presents an experience that appears automated to the user but is operated manually behind the scenes, testing demand before building the underlying technology. A Concierge MVP involves personally delivering the service to each customer by hand, validating that people want the outcome before scaling the operation. Other lightweight tests include smoke tests, such as a "Buy Now" button that tracks clicks to validate willingness to pay, and split tests (A/B tests) that compare two versions of a feature to see which performs better on a specific metric. The underlying principle in each case is the same: test the riskiest assumption first and fail fast if it is wrong.