I like to develop my code windows as much as possible before collecting any research data. I base my codes off of prior research and my literature review. One of the downfalls of this method is if I spend too much time on scripting before I actually collect data, I might make more work for myself. Theoretical code windows do not always work as expected once we collect our actual research data.
I started collecting video data for my eWorkbooks project, and I realized I needed to make some changes to my OTRs code window. This is a natural part of the process. I use a modified grounded theory approach to my code windows. Rather than starting with a blank slate, I start with my hypothesized codes and then I use my data to generate new codes and to modify what I already made. I really recommend this approach to most educational researchers. One word of caution: Do not spend too much time on scripting and fancy formatting and outputs until your code window is more or less set. I only developed mine as much as I did for purposes of this blog…it was actually more work later to fix some of my links and scripting when I had too much of it.