How do you set up experiments or statistical tests to validate assumptions about operational data, and how do you interpret the results?