Twitter polls: useless.

I've seen someone doing twitter polls having someone vs someone else in some sort of strange popularity contest - who is "winning" well, whoever has more fans will "win" that sort of poll.

I've seen polls asking silly things and semi-serous things.

However, twitter polls are not going to yield you results that are worth anything.

The issues are numerous.

You have a limited sample size.

Whatever your current following on twitter is, it is limited in scope the demographics of who you are polling is narrow due to this. Any data collected will be very biased.

You can't make the poll a good one.

A yes/no/other poll is simply not a very good poll. Proper polls conceal what it is that they are polling for, so that you as the poll taker do not know what the poll is about, and, thus are more likely to be honest in the answers. Although there is always a margin for error (something that twitter polls could never account for) you are looking to get answers from the people who agree to take the poll, and also guard against people being non-honest actors. This is why you see polls have similar questions latter on, worded in other ways to see if the answer you gave last time is consistent with the answer you give this time. Its hard to remember a lie, easy to remember the truth. This is also why investigation people ask the same question more then once. This can't happen with a twitter poll.

You can't set any controls in place.

Blind, or double blind methods as well as other controls can not be put into place in twitter polls.

Your question might be formatted badly.

Unless you are using formalized logic to make your question (and even then) you run the risk that your question is a leading one, with a desired result in mind. Proper questions are very hard to make in a decent poll.

"So? Its only twitter."
Yes, but people are getting fired for things they put on twitter, so people do seem to take what is said here seriously, for good or for ill. Thus, polls should be subject to the sort of critical analysis that we would subject public polling. In this case, due to how it functions, I submit we should simply not use twitter to poll anything, as nothing those polls yield will tell us anything much... other then who is popular in a popularity contest, and honestly, I simply do not care about such data. I suppose if you do, then well - have at it, but remember due to the above you can never know for sure who is in fact more popular because you can not block bad-faith actors from voting in strange ways.