A lot of people starting up on the TT have asked me when they should apply for their first NSF grant. The answer to the question has nothing to do with time, but with data. One thing I really didn't appreciate early on was the inappropriately named "preliminary" data. I thought I could come in with a good idea, some supporting literature and a bit of data and it would totally be fine.
Not so much.
What I have learned along the long path strewn with the bodies of fallen proposals is that the bar for proof of concept is much higher than the n00b thinks. The unfortunate reality is that you've got to demonstrate your ability to generate data that support your hypotheses for at least two of your three aims. In many cases, you need to me most of the way to The Answer.
Established PIs can lean on a track record and toss in some suggestive data and off they go, but when starting out it is much less simple. If you were a forward thinking postdoc or had a mentor who knew how to prepare you to battle it out on the TT, chances are you brought some solid data to you lab to start your new life. In this case you may be able to send in a proposal right away.
If, like myself, you've decided to take your new lab in a new direction, you have a different row to hoe. If there's one thing I would love to tell N00b me, it is to get away from the computer and into the lab. I submitted grants in my first year that had exactly zero chance of getting funded because they lacked enough supporting data. Instead of slaving away on my first proposal, I could have been churning out the data to support it. While not crushing, this was a rookie mistake.
So rather than deciding on a particular time to write a proposal, decide when you think you have enough data to assuage reviewer concerns about the feasibility of your project.
The twitter conversation this morning that was the seed for this post is on storify here.