I learned how to start with a conjecture and say, “On this basis, we expect these metrics to be affected.” How do you measure that something is working? Are you expecting an increase in sales? An increase in conversions? If you look at science—which has existed for much longer than product management—it relies upon the initial conjecture to set up the experiment, that first hypothesis or leap of judgment. Truly data-driven experiments often involve following your intuition, then establishing a feedback loop of some sort to test whether your intuition is correct.
