I try to get anything going as quick as possible so I'm not just looking at code without any reward for a long time: I start by listing the data I will need to

I try to get anything going as quick as possible so I'm not just looking at code without any reward for a long time: I start by listing the data I will need to track, wire-frame sketch some screens for entry and management. This will get me a picture of what data I will need, I want to do with it, and what tools I need to look for. (Since you are aiming for mobile, it might be good to imagine with a phone in your hand and think how would I get X entered on this device and then do Y?) Next, I would get an initial database up and running sorting out my data elements into relatable tables as needed, then get a few sample records input as a test (maybe even straight from the DB command line - if the entry is going to be complex, at least you will have something to work with as you jump around getting stuff figured out.) Next start work on better entry (if entry is cumbersome), also reporting, etc... That's the start, you know what you want as the result... how to get the data where you need it and how to make value from it, one bit at a time, don't be afraid to re-do/refactor as you get things figured out.