A Minimum Viable Product (MVP) is the simplest experimental release that can start the build-measure-learn cycle for lean development. By quickly getting started with the cycle, a startup is able to begin learning what works best and developing the features that are most in demand.
What features should my MVP have?
According to Eric Ries in this video, most entrepreneurs overestimate how many features an MVP needs. His rule of thumb is to take your first idea for an MVP, then cut that in half, and cut it in half again.
Additionally, an MVP isn't always a product. It is sometimes a survey, demo, or very early prototype. The features that explain what the product is to the user are what really matter. By including these, you can begin to get feedback and determine what changes to make.
What is the goal of using an MVP?
There is one main goal of the MVP: start learning immediately. With a non-functioning demo, you can still learn a lot and begin attracting a user base for testing. By not including all features, you can receive very useful negative feedback and get a clear understanding of what features are most important. Users can come up with their own optimal version of features you may have thought of already, or they may suggest features you would never have thought of.
Buffer, a social media post manager/scheduler, had an MVP of just a page with details of the product and what it would cost. The page included no functionality: it was just a preview of what the product would be. Dropbox, a file backup/syncing solution, had a video demo for their MVP. And finally, Zapppos, an online shoe store, simply had photographs of shoes and links to stores where they could be purchased.
All these examples have in common simplicity, ease of creation, and that they allow learning to begin immediately. From their MVPs, the companies quickly figured out they had good potential products and were able to start implementing ideal features based on user feedback.