That’s without joking exactly what they do. Now if that one millennial voter is a different one in each cohort of participants you get huge swings in the survey that are not very predictive of election outcomes, and that error is poorly represented in the margin of error.