How do experienced Haskell developers approach laziness at *design* time?

Question

I\'m an intermediate Haskell programmer with tons of experience in strict FP and non-FP languages. Most of my Haskell code analyzes moderately large datasets (10^6..10^9 thi