Best advice is to launch fast and then iterate. Don't try to hundreds of surveys, user interviews, etc (it doesn't highlight anything about users).

Pre-launch startup goals

• Launch quickly
• Get initial customers
• Talk to customers and get feedback — figure out how to help them
  Do these steps and then iterate.

There is a misconception that if you should build a fully working product.

Many people have fear that they give a product to customers and they don't like it. This fear is usually unproductive.

Software MVP

• Very fast to build
• Very limited functionality
• Appeal to a small set of users

Popular Examples

• AirBnb
  • No payments
  • No map views
  • Must have airbed
  • Only for conferences
• Twitch.tv
  • Justin.tv, only one stream, one streamer, no games, just camera
• Stripe
  • No bank deals
  • Manually called banks, did all the paperwork in the background
  • Very few features, no api

Who are the people who want to use these scrappy prototypes?
If you imagine your hair is on fire now, you would be happy to buy a bucket of water. But you would also buy a brick to smash your head.
Find the most desperate people and sell your product to them.

Maybe I can skip MVP and just talk to people?

You customers are experts in their problems. But they don't have answers about how to solve the problems.
Surveys can help you understand what are the problems but not how to solve them.

You don't start a startup with all the answers. Building startup is about searching the answers.

How to build MVP quickly?

1. Set a specific deadline. 2 weeks / 1 month.
2. Write your spec: all the required features.
3. Cut your spec: think if clients really need these features, get rid of the rest.
4. Don't fall in love with your MVP: fall in love with the client, not your product.

It's better to have 100 customers who love your product than 100k customers who kind of like your product.