I used to be involved at times with "polling". It was product evaluation, really, I'd be on the fringes, but I'd see the questionaires ahead of time, and the results. One statistican, a fellow I liked a lot, told me about 10% of any rep group is checking boxes at random, or intentionally misrepresenating their experiences. They could usually cull the former with questions asked different ways, they'd get contradictory responses and probably delete the response. The others would be included if they were consistent.
Good polling is really a science and rather involved, and of course getting a true sample of the population is a challenge. I went back and looked at various polls in GA statewide elections and the final polls were all very close to the final outcome, without a point or two.