Most teams claim to be data-driven right up until the data threatens a favorite idea. Real progress comes from testing constantly, learning quickly, and not treating every failed experiment like a personal tragedy.