• by Sean Hargrave , Staff Writer, September 17, 2014

Social media is great for a lot of things -- but as the Scots prepare for tomorrow's historic vote, the one thing that social media is not very good at is made very clear once again. It's election time and so it falls on a few social media "buzz"-measuring companies to prove the worth of their wares by predicting the outcome of the result.

Today's lengthy article in The Daily Telegraph really doesn't even merit repeating a URL for. It claims 10m conversations on Facebook show a small swing in favour of a "Yes" vote. However, the paper's been sold a pup, as anyone with half a brain can instantly see. The "Yes" prediction is made because of the ten million conversations analysed, which were talking about tomorrow's referendum, around two million were in favour of independence. Just under two million were favouring maintaining the United Kingdom.

To the minds of those looking to back up that their software can magically predict events based on Facebook conversations, this is proof enough. Slightly more people appear to favour independence. Therefore it has to be a Yes vote?

But hang on. What about the circa six million other conversations that were analysed? Are we going to just discard them because they don't fit in with what some social measurement company is trying to prove?

This has been the problem with polls throughout the campaign. Lately they tend to leave out the "don't knows." This is a fatal flaw because the accepted wisdom at the start of the campaign was that there were two entrenched views -- "Yes" and "No" -- and each had roughly a quarter, or maybe a third, of the vote. It was always the third of the vote in between them which would swing the election. 

Such uncertainty doesn't make for good headlines, however, and so pollsters have found they make the biggest headlines and get the most coverage for their companies by leaving out the waverers and sticking with the decided. The result is laughable. Nobody has called it more than one or two percent either way and then admitted there is at least a three percent inaccuracy in all results. In other words, our figures are meaningless -- they predict nothing, but please say our name when you read out or print "in the latest poll by x company...."

Social media analysis only obviously looks at the mainly young, liberal demographic who are active in social. So it's skewed to begin with. When you discard the majority of the posts that show no leaning to either side, but still come out with a prediction, you really are grasping at straws.

Data analysts know a few better tricks than looking at Facebook for predictions. In an excellent presentation a few months back by SapientNitro, Adobe Summit delegates were shown not just the data around the vote, but what people were doing that data. If you want to know which way something is going, follow the money; follow the bookies and their punters to see where people who put their money where their mouth is are placing their bets.

Back then, as now, the bookie markets -- which can be accessed via odds checkers to compare the entire market on one page -- clearly think a No vote is more likely. Some are offering 4-1 on a two-horse race -- yes, that's right, they will give you four pounds for the one pound you lay down if Scotland votes Yes. Conversely, if you would need, on average, to put five pounds down to win a single pound back for predicting a No vote.

I'm sure a lot of brand managers and their agencies have held some suspicion over "buzz" and how social media sentiment can be used to predict the future and so dictate strategy. 

Let's hope tomorrow's vote shows up social media buzz forecasting for what it is -- snake oil that can come back to bite you if you think it speaks truths beyond the site it was measured on.

As I write these words, news has just arrived that BetFair is already paying out on a No vote. It could be a bit of publicity seeking. It certainly proves the point, though.

Although social media, in the minds of social metrics lovers, has said it will be a Yes, a company is forking out "a six figure sum" before it has to because it see the results as foregone conclusion.

Social says "Yes" and BetFair pays out a fortune on "No" are both headlined at The Daily Telegraph. Which do you believe?