The volume of social data that is being generated is huge, but if we move on from the ‘big data’ buzzword, the power of capturing the human feeling behind that data, offers exciting new possibilities.
IBM estimates that 90% of data currently stored has been created in the last few years. 73% of organisations are now storing more data than ever before, with close to 40% expecting it to inform strategic decisions.
Extracting information from online content such as search engine requests, blogs, forums and social media platforms by automated techniques, it can be applied to volatile markets, to track shifting public opinion in real-time about brands, investments, and new stories. From there it can be used to trigger better results ￼from marketing campaigns, determine how and what stakeholders want to hear senior executives, and shape the design of new, better products. No surprise then, to see the attraction for businesses which rely on feedback to remain competitive.
Yet while the methods of data analysis grow in sophistication, relying on automation remains fallible. 1 in 3 business leaders don’t fully trust their data, and it is thought that poor quality data has huge costs, $3.1 trillion annually in the United States alone. The reason is that sentiment analysis, despite efforts to sharpen techniques, still lacks depth and context. Successful application requires the logic and reasoning that only humans can apply – essentially, the difference between information and knowledge.
When Google couldn’t catch the flu
One of the most widely-known examples of data analysis’ limitations is the Google Flu episode. Beginning in 2009, the tech giant attempted to track the global spread of influenza from the terms people were entering into its search engine. The theory went that they would be able to detect future flu outbreaks faster than authorities. Despite Google’s confidence in its predictions however, it turned out that the majority of them were quite as accurate as hoped and it actually missed the 2009 swine flu pandemic.
One reason for this is that although Google’s approach was able to track what search terms people were interested in, such as “Symptoms of flu”, it couldn’t tell whether these people were actually suffering from the flu already, were worrying they might have it, or were interested for another reason entirely. If this data had been analysed in conjunction with the more qualitative opportunities that social data offers, a different result it may well have been.
Trading in sentiment
Real-time data, is naturally invaluable, especially when it turns on the opinion of social media influencers like Warren Buffett.
But then there is also the question of how much faith to put in data sourced from Twitter, when one looks at the huge damage caused by a fake Tweet from a hacked Associated Press news feed in 2013 about a supposed attack on the White House. Coverage of the alleged incident swept Twitter, wiping millions off stocks within minutes before it was officially denied. Traders could only watch as complex algorithms designed to assess and apply social sentiment devastated portfolios.
Social sentiment can tell businesses how people are feeling at a set point in time – it can’t instruct them on how best to deal with that information – that requires human involvement.
Is the future of social data human?
At its most useful when there is large volumes of data that needs to be quickly condensed into easily-understandable numbers, the capabilities of of analytic social sentiment software will undoubtedly be enhanced in years to come. But it must be matched with the pattern-recognition and associative thinking to match up the human and the machine.