Of course in the real world experiments are needed to determine causality. I could easily hypothesize that two-year contracts don’t reduce churn, they select for customers who find the product more useful in the first place!