Having a big sample size does not help unless it is a representitive sample of who will vote in an election. It only makes the the poll more precise in its inaccuracy. As an extreme example to illustrate the difference consider a poll of 100,000 members of a communist fan club vs a poll of only 1000 residence from households in a district who regulraly vote. The “margin of error” of the 100,000 communist fans will be very very small due to a huge sample size. But they are not a representitive sample of voters so the poll results will be garbage.
It would be easy for a program to generate a random list of voters with the correct demographic ratios.
A 2nd program then datamines their social media, email and browsing history to calculate who they are likely to vote for.
Far more accurate than traditional polls where they have to deal with diminishing land lines, shy voters and those who deliberately lie to the pollster.
Most modern polls are push polls designed to sway or measure public opinion.