Best practices when starting a new project [closed]

Question

When you first fire up Visual Studio, or even before that (Hopefully) What do you consider best practices when starting a new project?

Different things I have done in the past

• Consider the technology choices
• Attempt to identify the potential bottlenecks/brickwalls
• Throw away the specification
• Ask lots and lots of questions
• Launched straight into code! (who hasn't?)

As programmers we should be methodical in our approach.
What do you do when starting a new project?

Answer 1

Start using version control right away.

It is always frustrating building prototype "throwaway code" that eventually gets thrown away - and then realizing that some of that code would be really really useful to have.

Answer 2

(If applicable) Figure out whether you are building a throwaway prototype, an expandable prototype, or a final production system.

I think that a major problem with a lot of projects is that people don't get this one figured out to begin with. They either over-engineer and the project doesn't deliver on time, or they write throwaway code that later becomes too expensive to throw away.