Back to the author's point: "good enough" is the bar, when we are talking about a choice that impacts millions of users. Changing "good enough" to "optimal" requires gigantic costs in retraining and redesign, and simply isn't worth it.