Regularly referring to that ~30% spread as "one polling error" made this a lot more understandable than most statistics for many people.