Polls that did correctly predict the general election result were the ones that were able to find hard-to-reach people, those who are difficult to contact on the telephone or online, according to Damian Lyons Lowe, founder and CEO of Survation.
"Say you are a 35 year old male and you work a 9-5 job, you are not able to answer your land line because you are at work and you are not able to answer your mobile phone," he told CNBC's Street Signs. "We would try and call that person in the evening, at a time when they might be able to participate in the interview."
Polling data is further complicated by the social media. Data analysts Ontotext released a study last week arguing that there was a clear preference for leaving the EU among Twitter users, based on findings that leave-related hashtags were mentioned 2.6 times more often the remain-related hashtags.
However, Twyman questioned the value of using social media for survey data.
"The reason you get difference from social media polls is social media polls are complete rubbish and are not representative of anything other than the echo chamber of social media," he told CNBC in a phone interview.
"Something conducted on Twitter isn't representative of Twitter users; it's an example of what we call passive polling, where it is a free-for-all and anyone can take part."
Another source for predictions is betting companies. Generally, these companies expect the remain vote to win and have offered much shorter odds for bets on a vote to remain than for a vote to leave. For instance, Reuters reported that bookmaker Betfair had calculated at the weekend a 70 percent probability of a vote to remain (although this was down from a 78 percent probability earlier in the week).
However, just because bookmakers expect a vote to remain does not mean they are a more reliable measure than polls.
"Why are polls different from the betting markets? Because they are measuring different things; polls are only ever intended as a snapshot of public opinion at the time," said Twyman.
"[Polls] can't be and nor should they be used as a prediction. Betting markets are only interested in that prediction," he added. "At the moment, they are showing overwhelmingly remain… because that's where historical precedent goes."
Follow CNBC International on Twitter and Facebook.