Confession time
I was asked to act as a "teller" outside of the local polling station for the council elections. It wasn't too bad; a nice day, some interesting people, very light work. All I had to do was ask people for their polling number, write it down, and then direct them to the correct door. (That polling station covered several wards)
Some people were a bit obnoxious; OK, if you don' want to tell me your number, not a problem. I wasn't asking how they voted, and TBH I couldn't care. I just recorded it as a null so that we still had an understanding of the number voting.
Part of the reason for this is to act as a check; the numbers were shared between the parties, and as long as those numbers more or less equated to the turnout, they were happy. If the number of votes cast was much higher than the number of people passing through the door however....
The parties also use the details to check who of their likely supporters has voted or not. They generally check those details and start calling around later in the day. I actually gave some thought to streamlining the process, and considered that a small mobile device with voice recognition software might actually allow the data to be entered more efficiently and accurately, plus it could be really up to date.
On top of that, I can see some interesting possibilities for the data analysis; combining it with electoral register and various other information, and you might be able to see what time people vote, try and see if there are relationships to social factors etc. I was getting quite excited by the possibilities.
Unfortunately, it is far more likely that people would be worried about this data being misused (and with some justification). I suspect that this is what might hold back many other ideas; and again, I can see why sometimes these ideas are great on paper, but dangerous if in the wrong hands.