Your first step depends on what you know, what you don't know, and what you need to know.
Build it in the same order that the customer uses it
Start with an online ad, then a sign up page, then the product. That way, each step you take is in the right direction because your customers are with you every step of the way. When they do come with you, you can talk to them so they can guide you. And if your customers don't follow, you've just got one step back to find them.
Build it with the common objects first
Amazon starts every project by writing the press release. Some software companies draw the screens on paper with their clients. That way, the team is collaborating on something that builds common understanding.
Build it to scratch your own itch
It doesn't need to be pretty, easy or sellable. But if you make something you love, then you can find a few more people who love it just as much. That's how GMail, Linux, Virgin Airlines and Ben & Jerry's Ice Cream started.
Build the riskiest parts first
Great idea but it's hard to meet the right customer? Start there. They'll love it, but only when they can really understand it? Start there. Starting with the riskiest part of your idea avoids getting yourself stuck on a path that's not going to work.
Don't build it, do it
Instead of building anything, find a way to deliver the same benefit personally. Then you can learn all the important details about what your customers really want, and use that knowledge to build the first version.