Because everyone always runs A/B tests to decide whether to add a feature, but never runs them to decide whether to remove one.