> "It’s really about science."

Where all samples are unique and uncontrolled, no matter what you do, any consistent results are basically meaningless.

Once you've got samples size big enough to be statistically relevant, well, your growth problem is probably already solved.