re. How big it should be — pick a target customer & one of their real customer needs. Make it big enough to get that customer to switch to your solution for that need, but not much bigger. I’ll usually let the project grow 10–20% to capture low hanging fruit outside the primary target.

re. How to roughly estimate how long things will take — work with your engineers and compare to past projects. If it’s nothing like a past project add on a bunch of time for learning the new area.