[gigatexal]

What I do is write the query that gets me the data to answer the question that I need no matter how crappy. And then using the engine and statistics and maybe diagramming out the relations between tables and the indexes try to refactor it to be better. In the end what matters is that the data returned is right. Optimizing after you have that makes most sense in my head and that's when you can go off and feel confident in doing so ( because your deliverable is done, the correct data returned ) and research more "advanced" sql to get your query better.